Riichi Book I

A Mahjong Strategy Primer

for European Players

Daina Chiba

Riichi Book I

A Mahjong Strategy Primer for European Players

Version ﬀ

Daina Chiba

Permission is granted to copy, distribute, and/or modify this document un-

der the terms of the Creative Commons Attribution NonCommercial 3.0

Unported License, available at http://goo.gl/R9sy.

The original form of this book is L

X source code. Compiling this L

source has the eﬀect of generating a device-independent representation

of a manuscript. The L

X source for this book is available from http:

//riichi.dynaman.net/.

About the Author

Daina (pronounced like “diner”) Chiba is a political scientist and an ap-

plied statistician working as a lecturer at University of Essex in England.

He has been playing riichi mahjong for almost 20 years. He can be reached

at daina.chiba@gmail.com.

Preface

When I moved to England in 2013, I was pleasantly surprised to

learn that riichi mahjong (modern Japanese mahjong) is quite popu-

lar in Europe. In the past two years, I have had the pleasure of play-

ing riichi in London, Guildford, Kent, Oxford, Aachen, Copenhagen,

Prague, and Vienna, along with players from Austria, China, Czech

Republic, Denmark, Estonia, Finland, France, Germany, Italy, Japan,

the Netherlands, Poland, Russia, Slovakia, Sweden, the UK, and the

United States.

European players have been remarkably successful in organiz-

ing tournaments open to anyone who plays the game. These tourna-

ments — held at least once a month somewhere in Europe — are run

by local mahjong players in each country under the auspices of the

European Mahjong Association (EMA).

Founded in 2005, EMA has

been doing a fantastic job in maintaining common rule sets,

keep-

ing a player ranking system, and doing many other useful things to

promote the playing of mahjong across Europe.

Although I have come across a few good players in Europe, I came

to realize that a lot of players here are not very well-versed in the

basic principles of competitive mahjong strategies. Of course, play-

ing competitively is not the only way to enjoy the game. I am also

http://mahjong-europe.org/

EMA’s oﬃcial rule book for riichi mahjong is available online at http:

//mahjong-europe.org/portal/images/docs/Riichi-rules-2016-EN.pdf

(last revised in 2016). At the time of writing this book, EMA is in the process

of revising the rule book. Explanations of EMA rules in this book are based

on the revised rules. New rules will come into eﬀect from April, 2016.

iv PREFACE

not claiming that I know the magic formula to win because there is

no such thing. Nevertheless, there is a set of basic principles worth

learning for any aspiring players. I believe the level of sophistication

among European players could be much improved if these princi-

ples are more widely shared. Unfortunately, however, learning re-

sources currently available for non-Japanese audience are somewhat

limited.

I have thus decided to write a book on riichi mahjong strategies

for European players, primarily with beginners and inter mediate play-

ers in mind. I then ended up splitting the book into two volumes;

Book I is intended for beginners and intermediate players (Tenhou

rank of or below), while Book II is meant for more advanced

players. The two books are not intended for complete novices who

do not know how to play riichi mahjong.

The target reader is any-

one who has played riichi mahjong before and wants to improve their

skills further.

I have three main goals in preparing these books. First, I will in-

troduce a set of English terminology of riichi mahjong. “In beginning

was Word,” scripture tells us. Knowing the names of particular tile

combinations, situations, and strategies will allow us to be conscious

about them and to be able to talk about them with our fellow players

after the game.

My second goal is to introduce the principles of tile eﬃciency.

Book I and Book II both cover tile eﬃciency, but at diﬀerent levels.

There are already a few English books for beginners. There are also several

excellent blog posts on technical details about mahjong strategies. However,

there appears to be a huge gap between these two sets of resources. Introduc-

tory books do not cover strategies extensively, whereas blog posts tend to be

too advanced even for intermediate players.

If you want to learn how to play riichi, I’d recommend Barr (2009).

Book I oﬀers an introduction to tile eﬃciency, covering very basic

mechanisms only. I plan to cover more advanced materials in Book

II. My third goal is to introduce a set of simple strategies regarding

critical judgements such as whether or not to call riichi, whether to

push or to fold, and whether or not to meld.

A lot of the materials covered in the books were introduced to

me through the writings of a notable Japanese mahjong player and

manga author, Masayuki Katayama. Mr. Katayama is an accom-

plished riichi player and arguably the best mahjong manga author in

the world. Some of the strategies introduced here are unabashedly

stolen from Mr. Katayama’s masterpiece manga storybook Utahime

Obakamiiko ( ). I strongly encourage you to read

it yourself if you read Japanese, although I realize that you would not

be reading my book if you understood Japanese.

Another Japanese author whose work has been inﬂuential in the

writing of Book I is Makoto Fukuchi. Mr. Fukuchi is also a distin-

guished riichi player and the best-selling author of mahjong strategy

books. A part of the exposition of t he ﬁve-block method in Chapter

4 is based on Mr. Fukuchi’s skillful explanation in his books.

I am also indebted to a lot of friends I have acquainted with through

mahjong in Europe. Philipp Martin has read an early draft of the

book and provided me wit h valuable comments and encouragement.

I am also grateful to Gemma Sakamoto, who has been hosting a monthly

mahjong get-together in London. Finally, my thanks go to Ian Fraser,

one of the founders of the UK Mahjong Association. Without the ef-

forts of Ian and his team, I would not have been able to get to know

so many fellow players in the UK and in Europe.

The cover photo (© Katarína Mózová) is from the 2015 Bratislava

vi PREFACE

Riichi Open Tournament. I thank Katarína and Riichi Mahjong Slo-

vakia (especially Matej Labaš) for giving me their permission to use

it.

After I made the book publicly available in January 2016, a lot

of people have given me feedback on various aspects of the book.

Based on their feedback, I corrected some terminology inconsisten-

cies and typos. In particular, I thank David Clarke, Aaron Ebejer,

Nicolas Giaconia, Grant Mahoney, Ting, and Chris Rowe for their

valuable inputs.

Daina Chiba

London, UK

10 January, 2016

(updated on 10 April, 2017)

vii

Plan of the book

To improve your mahjong skills, you need not only to learn the

theories but also to practice what you learn by playing lots of games,

preferably with players who are stronger than yourself. Before the

advent of online mahjong platforms, however, doing so was not very

easy if you live outside of Japan.

Thanks to the recent development of online mahjong platforms, it

is now feasible for you to play hundreds or thousands of games with

serious opponents while living outside of Japan. On these websites,

you can easily ﬁnd fellow players to play with 24/7. Most platforms

keep the record of all t he games players have played, and a replay

function would allow you to reﬂect on your past plays. You can also

take a look at player statistics data, which gives you important clues

as to what skills you need to work on.

I thus recommend you practice mahjong skills by playing online

while you study the strategy principles with this book. You do not

need to wait until you ﬁnish reading everything covered in the book

before you start playing. Go ahead and play games ﬁrst, then come

back to the book and study the relevant parts of the book.

This book is divided into four parts. Part I provides an introduc-

tion to an online mahjong platform called Tenhou ( ). The website

is in Japanese, but I will walk you through the account registration

process and show you how to play games in Chapter 1. There al-

ready exist several excellent online resources that explain how to play

Tenhou, including:

• Arcturus’s Tenhou Documentation

http://arcturus.su/tenhou/

viii PREFACE

• Complete Beginner’s Guide to Online Mahjong (Osamuko)

http://goo.gl/F5sJvO

• Playing Online: Tenhou (Reach Mahjong of New York)

http://goo.gl/Oc1eNe

If you have already read either of the three before, you can skip Chap-

ter 1 of this book, for there is not much new information there for

you. Chapter 2 explains some advanced features of Tenhou, which

you can also skip when you read this book for the ﬁrst time.

Parts II and III are the “meat” of the book. Part II covers basic

tile eﬃciency t heories that allow you to maximize the speed and/or

hand value of your hand. After introducing basic terminology in

Chapter 3, I discuss the ﬁve-block method in Chapter 4 and provide

some tips on how to pursue several yaku in Chapter 5. Part III cov-

ers strategy principles, including score calculation methods (Chap-

ter 6), riichi judgement (Chapter 7), defense judgement (Chapter 8),

melding judgement (Chapter 9), and so called “grand strategies” to

win a game (Chapter 10). Finally, Appendices include a chapter on

etiquettes for oﬄine playing (Chapter A) and another chapter on fur-

ther readings (Chapter B).

Numbers and letters shown in this color as well as each entry in

the Contents section below are hyperlinked; clicking on one will take

you to the pertinent page.

Contents

Preface iii

I Playing Riichi Online 1

1 Introduction to Tenhou ( ) 3

1.1 Why play online? . . . . . . . . . . . . . . . . . . . . . 3

1.2 Setting up an account . . . . . . . . . . . . . . . . . . . 4

1.3 The main page . . . . . . . . . . . . . . . . . . . . . . . 10

1.4 Playing a game . . . . . . . . . . . . . . . . . . . . . . 13

1.5 Troubleshooting . . . . . . . . . . . . . . . . . . . . . . 26

1.6 Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2 Advanced features of Tenhou 30

2.1 Rank and rating . . . . . . . . . . . . . . . . . . . . . . 30

2.2 Four rooms . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3 Reading t he statistics . . . . . . . . . . . . . . . . . . . 37

2.4 Viewing games . . . . . . . . . . . . . . . . . . . . . . 43

II Basic Tile Eﬃciency 46

3 Riichi mahjong basics 48

3.1 Learning strategies . . . . . . . . . . . . . . . . . . . . 48

3.2 Basic building blocks . . . . . . . . . . . . . . . . . . . 50

3.3 Complex shapes . . . . . . . . . . . . . . . . . . . . . . 65

3.4 Waits . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

x CONTENTS

3.5 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4 The ﬁve-block method 84

4.1 Finding a redundant tile . . . . . . . . . . . . . . . . . 85

4.2 Alternative conﬁgurations . . . . . . . . . . . . . . . . 88

4.3 Selecting tile blocks . . . . . . . . . . . . . . . . . . . . 97

4.4 Building a block . . . . . . . . . . . . . . . . . . . . . . 104

5 Pursuing yaku 110

5.1 How to get sanshoku . . . . . . . . . . . . . . . . . . . 111

5.2 How to get ittsu . . . . . . . . . . . . . . . . . . . . . 120

5.3 How to get pinfu . . . . . . . . . . . . . . . . . . . . . 125

5.4 How to get honitsu . . . . . . . . . . . . . . . . . . . . 128

5.5 How to get toitoi / chiitoitsu . . . . . . . . . . . . 134

III Strategy Principles 138

6 Scoring 140

6.1 Three steps in score calculation . . . . . . . . . . . . . 140

6.2 Basic scoring . . . . . . . . . . . . . . . . . . . . . . . . 142

6.3 Advanced scoring . . . . . . . . . . . . . . . . . . . . . 153

6.4 Scoring tables . . . . . . . . . . . . . . . . . . . . . . . 161

7 Riichi judgement 163

7.1 To riichi or not to riichi? . . . . . . . . . . . . . . . . . 163

7.2 Insta-riichi . . . . . . . . . . . . . . . . . . . . . . . . . 166

7.3 When not to riichi . . . . . . . . . . . . . . . . . . . . . 176

7.4 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . 188

8 Defense judgement 189

8.1 To push or to fold? . . . . . . . . . . . . . . . . . . . . 189

CONTENTS xi

8.2 Defense basics . . . . . . . . . . . . . . . . . . . . . . . 193

8.3 Defense against riichi . . . . . . . . . . . . . . . . . . . 204

8.4 Defense against open hands . . . . . . . . . . . . . . . 208

8.5 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . 215

9 Melding judgement 216

9.1 To meld or not to meld? . . . . . . . . . . . . . . . . . 216

9.2 Melding choice: examples . . . . . . . . . . . . . . . . 219

9.3 Calling kan . . . . . . . . . . . . . . . . . . . . . . . . . 225

9.4 Miscellaneous tips for melding . . . . . . . . . . . . . 231

10 Grand strategies 235

10.1 What do do in South-4 . . . . . . . . . . . . . . . . . . 235

10.2 What to do by South-3 . . . . . . . . . . . . . . . . . . 244

10.3 Tables for induced point diﬀerences . . . . . . . . . . 247

Appendices

A Manners for oﬄine playing 250

A.1 Dealing tiles . . . . . . . . . . . . . . . . . . . . . . . . 250

A.2 Drawing and discarding . . . . . . . . . . . . . . . . . 252

A.3 Calling . . . . . . . . . . . . . . . . . . . . . . . . . . . 254

A.4 Winning a hand . . . . . . . . . . . . . . . . . . . . . . 255

B Further readings 260

B.1 Books on riichi mahjong . . . . . . . . . . . . . . . . . 260

B.2 Online resources . . . . . . . . . . . . . . . . . . . . . 262

Part I

Playing Riichi Online

Chapter 1

Introduction to Tenhou ( )

1.1 Why play online?

Playing mahjong online is an excellent way to practice your mahjong

skills. You don’t need any mahjong equipment to play; you don’t

need to coordinate with your mahjong friends to ﬁnd the time and

place t hat work for the all four of you. You can simply open your

computer and access one of many online mahjong platforms. As long

as you have an internet connection, you can play mahjong any time,

anywhere, and for any length of time.

Another advantage of online playing is that you can easily keep

the record of your playing history and obtain detailed statistics from

all the games you play. Analyzing these statistics will help you iden-

tify what skill sets you need to work on. You can also show your

game record to your friends and ask for their opinions about partic-

ular choices you’ve made in a game.

(Tenhou) is arguably the

most popular online mahjong plat-

form in the world. As of December,

2015, there are over three hundred

thousands active players on Tenhou.

A lot of professional mahjong play-

ers from Japan now play Tenhou.

There are also some Tenhou players who have later become profes-

sional after practicing their skills on Tenhou. It has become a common

To be exact, it has 304,534 active players and 3,566,353 registered players as of

20 December, 2015.

4 CHAPTER 1. INTRODUCTION TO TENHOU ( )

understanding among players in Japan that your rank and rating on

Tenhou are one of the most reliable indicators of your mahjong skill

levels. To get you started, this chapter explains how to set up an ac-

count on Tenhou and provides some basic operation manual.

1.2 Setting up an account

One of the challenges for European players in setting up an ac-

count on Tenhou would be that almost everything is written in Japanese.

However, you will only need a minimal level of Japanese to get by,

and this chapter will walk you through the process.

First, go to the Tenhou webpage (http://tenhou.net/).

Scroll down and click either the PLAY button (to play in a pop-up

window) or a link just below the button (to play in the current win-

dow) that reads .

Click here to play

in a pop-up

window.

Click this link to

play in the

current window.

1.2. SETTING UP AN ACCOUNT 5

Then, on the next page (either in a pop-up window or in the cur-

rent window), you’ll see something like the following:

Flash version Web version

The bottom line will initially read LOADING... / , but

in a few seconds it will change into »Flash | Web

β . Then, click on the Flash link if you are

accessing from a ﬂash-capable device such as your PC; alternatively,

click on the Web β link if you are accessing from a

smart phone or tablet. If it doesn’t change into »Flash

| Web β within 10 seconds or so, you may want

to click on the link right next to LOADING, which will

prompt the browser to reload the page. Clicking on either of the

Flash/Web links will take you to the log-in entrance of

Tenhou. Explanations below are based on the Flash version.

6 CHAPTER 1. INTRODUCTION TO TENHOU ( )

Registration ID ﬁeld (likely to be blank at ﬁrst)

When you ﬁrst visit this page, the ID ﬁeld right next to the

ID button is likely to be blank, as shown in the picture above.

This is because you haven’t registered an account. In order to create

an account, click on the ID (New ID Registration) button

on the left.

A pop-up message will show up, warning you that whatever ID

that is currently shown in the ID ﬁeld (if any) will be overwritten

with a new ID and that you may want to copy and paste the current

ID (if any) into some text ﬁle or similar. Do so if you do see an old

ID in the ID ﬁeld, just to be safe. If the ID ﬁeld is blank, just click the

Yes button, which will open yet another pop-up message.

1.2. SETTING UP AN ACCOUNT 7

It is telling you the following:

• You can create a player ID for free, and doing so is necessary if

you want to earn a rank (kyu / dan) and rating.

• Some characters or character combinations are not allowed in

player names.

• Once you register, you cannot register another account for a

given period (7 days).

• If you don’t play for 180 days, your ID may be deleted.

• A player name must have 1-8 characters.

Type in a player name you’d like to have (8 characters or fewer)

into the blank ﬁeld at the bottom and click OK. You cannot change

your player name later, so choose wisely. If the player name you type

in is already taken by another player, it gives you an error message,

as follows:

This player name is taken.

Click OK, and type in another name. If successful, you’ll see a

new message asking you to conﬁrm that you want to register an ac-

count with the player name provided.

8 CHAPTER 1. INTRODUCTION TO TENHOU ( )

Click Yes and you’ll see another message as follows:

The 19-digit alpha-numeric code that starts with “ID” (shown in

white on a black background) is your unique player ID (it is ID12345678-

ABCDEFGH in the picture above). I suggest you save your ID in a

text ﬁle or something so that you don’t lose it. They cannot re-issue

your player ID (unless you have a paid membership and hold a rank

of or higher).

Clicking OK will take you back to the log-in entrance page, but

this time you should see your player ID in the ID ﬁeld.

1.2. SETTING UP AN ACCOUNT 9

Your player ID

Choose (male voice) or (female voice)

Can choose

You can make several choices before entering the main page. First,

you can choose male or female voice (for pon / chii / riichi, etc.) by

clicking on the button right next to the ID ﬁeld. You can choose a

diﬀerent gender each time you log in to the main lobby. Second, you

can choose (premium) or (economy) version.

The premium version has better graphics, so I suggest you choose

the premium version.

If you are happy with your choices, you can enter the main page

by clicking on the OK button on the right.

10 CHAPTER 1. INTRODUCTION TO TENHOU ( )

1.3 The main page

Here is what the Tenhou main page looks like when you ﬁrst log

on in. The right half of the main page shows your statistics (currently

all the ﬁelds are blank because you haven’t played any games), and

the left half shows the games you can play and some other features.

Menu Cancel (to cancel a reservation)

Main tabs

Sub tabs

In the second line of the left hand side, you see three numbers. In

the example above, they are 1857, 915, and 118 (the numbers will be

diﬀerent on your screen). These numbers show that 1857 players are

currently online, 915 players are waiting, and 118 players are about

to ﬁnish their games.

Below these three numbers, there are six main tabs, which read

, β, β, , , and . The tab is the main

lobby where we play games ( reads dan-i-sen in Japanese; it

means ranking matches). Under the tab, there are four sub-

tabs, which read , , , and , corresponding to four

1.3. THE MAIN PAGE 11

diﬀerent rooms. At ﬁrst you can only play at tables in the room.

Let’s ﬁrst go to the tab, and choose the sub-tab.

Making reservations

In each of the four rooms (i.e., , , , and ), there are

12 diﬀerent variants of riichi mahjong games you can choose from.

Games in the left column (under tonpusen) are East-only

games,

and games in the right column (under tonnansen)

are more standard East-South games that have both East and South

rounds.

Closed tanyao

Open tanyao

With red ﬁves

Fast

Three-player

Fast (3p)

East-only East–South

1 player waiting

44 players playing

Games in the ﬁrst row ( kuitan nashi) are unusual games

where open tanyao (All Simples) is not allowed; you have to have a

concealed hand to claim tanyao.

There is no red ﬁve in these games,

either. Open tanyao is allowed in all the other games. Games in t he

In a special circumstance where no player gets 30000 or more points by the end

of East-4, the game continues into the South round.

Just like East-only games, when there is no player who has 30000 or more points

by the end of South-4, the game continues into t he West round.

kuitan means “open tanyao” and kuitan nashi means “without kuitan” in

Japanese.

12 CHAPTER 1. INTRODUCTION TO TENHOU ( )

second row are more standard games with open tanyao, but they do

not have red ﬁves. Games in the third row have three red ﬁves. This

is arguably the most standard type of riichi mahjong game played in

Japan as of now. Games in the fourth row have the same r ule as those

in the third, but the time limit on each action is more strict. Games in

the ﬁfth and sixth rows are three-player games, where open tanyao

and red ﬁves are both allowed.

The set of numbers delimited by a colon in each cell represent

the numbers of players currently waiting and playing the game, re-

spectively. For example, the ﬁrst row in the left column shows 3:24,

which means that 3 players are waiting in queue after signing up for

a game, and 24 players are currently playing East-only, closed tanyao

games. As it happens, East-South games with red ﬁves are usually

the most popular on Tenhou, followed by East-only, fast games.

To sign up for a game, click on the (reservation) button in

the corresponding cell. You can make as many reservations as you

want; you will be given a seat at a table that ﬁrst becomes available. If

you make multiple reservations, other reservations will be automat-

ically canceled when you start playing at another table. To cancel all

the reservations at once, click on the (Cancel) button at

the bottom right of the left-hand side of the main page. The cancel

button becomes active (clickable) only after you make a reservation.

1.4. PLAYING A GAME 13

1.4 Playing a game

Once a slot becomes available for

you, you will be taken to a game ta-

ble along with three other players. A

black pop-up screen (see right) will

appear. The game will start in 10 sec-

onds (if all t he four players click on

the OK button, the game will start

immediately). Each player is ran-

domly assigned to East, West, South, or North. In the example below,

my initial seat wind is North ( ).

The Tenhou interface is quite intuitive so you won’t need much

instruction. Once a hand begins, tiles are dealt automatically. You

also automatically draw a tile when your turn comes. In each turn,

click on the tile you want to discard.

Each action is timed. At a standard (non fast) table, you have 5

seconds to discard a tile. In addition, you are given a total allowance

of 10 seconds in each hand. That is, even when you use up the 5

14 CHAPTER 1. INTRODUCTION TO TENHOU ( )

seconds allocated to you in a particular turn, you will be given the

maximum of additional 10 seconds (minus the seconds you have al-

ready used up in previous turns in the hand). For example, when

you use 5 + 4 seconds in the ﬁrst turn, t he remaining allowance re-

duces to 10−4 = 6 seconds in this hand. Therefore, the next time you

use up the ﬁrst 5 seconds, you will be given only 6 more seconds. The

allowance will increase by 1 second (up to 10 seconds) each time you

make your discard choice in less than 1 second. The allowance will

revert to 10 seconds when the next hand begins. At fast tables, each

action must be done in 3 seconds, with a total allowance of 5 seconds.

1.4.1 Calling / melding

When a call becomes available, a box with a call name will show

up to prompt your reaction. The call prompts are written in Japanese.

The good news is that they are relatively simple and easy to guess

from the context. It would be enough to memorize the following

eight mahjong words in Japanese.

1. riichi [r´I:tS]

You can call riichi when you have (1) a closed ready hand, (2) at

least 1000 points left, and (3) at least one turn left to draw. When all

of the t hree conditions are met, a translucent box that reads

in white letters will pop up in your tur n.

If you want to riichi, you must click on the box ﬁrst, then click

on the tile you want to discard. Once you click on the box, you

cannot call it oﬀ. Clicking on the box also makes it impossible

to discard a tile that does not make the hand ready. In the above ex-

1.4. PLAYING A GAME 15

ample, tiles other than ıȷﬁ” will become unclickable once you

click on the box. If you do not want to call riichi, just click on

the tile you want to discard.

2. ron [r´ON]

A box will pop up whenever you can legitimately declare

ron on an opponent’s discard. In other words, chombo is made im-

possible on Tenhou. For example, when you are furiten, a box

will not pop up because you cannot legally ron with a furiten hand.

Whenever your hand is in a furiten status, it is indicated with a

(furiten) sign in small translucent letters below your hand

that looks like: . If you don’t click on the box in time

(i.e., in 5 or 3 seconds + allowance), it is assumed that you pass.

3. Pass (do nothing)

Whenever a box pops up, another box that reads (pass)

will accompany it.

Click on the box immediately if you don’t want to declare ron

on a discard. You would not want to pause for too long because t hat

can look suspicious. A box will also pop up when other calling

actions become available.

4. tsumo [ts´umo]

A box will pop up when you can legitimately declare tsumo

with your draw.

16 CHAPTER 1. INTRODUCTION TO TENHOU ( )

5. pon [p´ON]

When calling pon becomes available, a box will pop up right

above the tiles in your hand with which to call pon. A box will

also pop up. If you want to call pon, mouseover the tiles in your

hand with which to call pon. Then the candidate tiles will stick out,

as follows:

Click on them to call pon. If you click on the box or don’t do

anything in time, it is assumed that you pass.

6. chii [tS´I:]

Chii calls are done in a similar way. When it becomes available,

a small sign that reads will pop up right above the tiles in your

hand with which to call chii.

When you have multiple sets of tiles with which to chii, as is the

case in the above picture, mouseover the candidate tiles to choose. In

the picture above, the left player discarded a ‚ and you can chii it

with either ¸˛or ˛‹. If you want to chii it with ¸˛, mouseover

the ¸ then the ¸ and a ˛ will stick out so you can click on them. If

you want to chii it with ˛‹, mouseover the ‹ then a ˛ and the

‹ will stick out so you can click on them.

1.4. PLAYING A GAME 17

7. kan [k´2N]

Calling kan on a discard is similar to calling pon. To build a melded

kan by extending a melded pon, you need to mouseover the melded

pon until a small sign that reads appears below the pon. To call

a concealed kan, mouseover the four tiles you want to kan then the

tiles will stick out, accompanied by a small sign that reads below

them. Click on them to call kan.

8. Kyushu Kyuhai

When you have nine diﬀerent terminals and honors after the ﬁrst

draw in an uninterrupted ﬁrst set of turns, you can declare an abortive

draw. When this becomes available, a box that reads will

pop up. Click on it if you want to declare an abortive draw. If you

wish to continue with the hand, just click on t he tile you want to dis-

card.

Multiple boxes

Sometimes you have multiple choices as to what to do with a

given discard of your opponent. In the following example, you have

a ready hand waiting for —-ı, and the left player discarded a ı.

You will be given the following three choices:

Ron Pass

Chii

• Call ron

• Call chii

• Pass (do nothing)

18 CHAPTER 1. INTRODUCTION TO TENHOU ( )

To call ron on the discarded ı, click on the (ron) box that pops

up above your hand. If you want to do nothing, click on the

(pass) box right next to the box. Alternatively, if you want to

call chii, mouseover the two tiles you want to chii with (in t his case

‰) and click on them.

1.4.2 Buttons

The buttons at the bottom right corner allow you to toggle on/oﬀ

some calling-related features. Each feature is turned oﬀ at the begin-

ning of a new hand.

Auto call win

Auto discard draw

No call Picture ( )

Sound ( )

(Auto call win)

If you turn this on, you will automatically win a hand when possi-

ble without clicking on or boxes. In other words, the option

of passing is unavailable when this is turned on. Keep in mind that

this can be problematic at times when you intend not to win your

hand from a particular opponent or on a particular tile. When t his

is turned on, the word is shown in white; when it is turned

oﬀ, it is translucent. In the picture above, it is turned on.

(Auto discard draw)

If you turn this on, you will automatically discard whatever tile

you draw. Turn this on when you have to go to toilet or somewhere

but don’t want to quit the game entirely. When you riichi, this feature

is automatically (and implicitly) turned on. In the picture above, it is

turned oﬀ.

1.4. PLAYING A GAME 19

(No call)

If you turn this on, you will not be prompted to call chii, pon, or

kan. This feature is useful for hiding information about your hand’s

tile composition from your opponents. If you pause every time some-

one discards a certain tile you can call, your opponents might be able

to guess what pairs of tiles you have and don’t have. Drawing a de-

duction from such time lags constitutes an important skill in Tenhou.

However, in order not to disadvantage players waiting to call chii /

pon too much, time lags will also occur randomly (i.e., even when no

one can call pon / chii on the discarded tile).

(Picture) and (Sound eﬀect)

You can change the appearance of the tiles and/or mat or resize

the window with the Picture button. You can turn on/oﬀ t he sound

eﬀect (for riichi, chii, pon, etc.) with the Sound button.

20 CHAPTER 1. INTRODUCTION TO TENHOU ( )

1.4.3 Scoring

When a player wins a hand, the score will be calculated auto-

matically. A scoring board will pop up that shows the hand, dora

(and ura dora if riichi was declared), yaku names and the associated

number of han, minipoints, and the total hand value.

In the example above, the left player dealt into my hand that is worth

60 (fu; minipoints) 13 (han) = 32000 (points). Yaku names will

be shown in Japanese along with han counts. Table 1.2 at the end of

this chapter lists all the yaku names Tenhou recognizes.

1.4. PLAYING A GAME 21

1.4.4 Indicators

The black rectangular board in the middle of the screen provides

information about the proceeding of the game.

Number of

riichi bets

Counter

We can see that this is South-4, there is 0 counter and 0 riichi bet,

and it is the North player’s turn. The West player is leading (44000

points), followed by the East player (26800), the North player (17400),

and the South player (11800). Player’s rank (kyu / dan) is shown

right next to their points.

If you mouseover the middle board, you will see the current point

diﬀerences between you and each of your opponents. In the present

example, the West player has 44000 − 26800 = +17200 more points

than I do. If you are leading against another player, the point diﬀer-

ence will be negative. For example, the Sout h player has 11800 so the

point diﬀerence is 11800 − 26800 = −15000.

22 CHAPTER 1. INTRODUCTION TO TENHOU ( )

It is important to pay a close attention to these point diﬀerences,

especially in the South round or when one of your opponents is at

the risk of bankruptcy. In the current example, if the East player wins

a 12000 hand from the South player, South will go bankrupt and the

game is terminated. Notice that East is currently ranked second, hav-

ing 17200 points fewer than West. In this case, winning a 12000 from

South is not ideal for East because East will still be ranked second

and the game is over.

You can also see the type of game you are currently playing on

the board. Just below the wall opposite to you is an indicator that

looks like this: .

• The ﬁrst letter indicates t he room: for (ippan), for

(joukyu), for (tokujou), for (houou). See

Chapter ?? for explanations of these.

• The second letter indicates if it is an East-only game ( ) or an

East–South ( ) game.

• The third letter indicates if open tanyao is allowed: (with

open tanyao) or (without open tanyao)

• A fourth letter ( ) is added if there are red ﬁves.

• A ﬁfth letter ( ) is added if it is a fast game.

As we will see later, avoiding the fourth place is very important in Tenhou rules

than it is in other rules. However, this does not mean that it is your only priority;

you would still want to improve your placement in a game when doing so is a

realistic possibility.

1.4. PLAYING A GAME 23

1.4.5 Ending of a game

A game can end in several diﬀerent ways.

• One or more player goes bankrupt (below 0 points).

• South-4 (East-4 in East-only games) ends and at least one player

has 30000 or more points.

• West-4 (South-4 in East-only games) ends.

• At any point in the West round (South round in East-only games),

at least one player has 30000 or more points.

When a game ends, ﬁnal scores are calculated as follows.

• In cases of a tie, the player sitting closer to the ﬁrst dealer wins.

• Oka (winning premium) is 20000. That is, although every player

is allocated 25000 points at the beginning of a game, they have

to return 30000 at the end of the game, meaning that 30000 will

be subtracted from the ﬁnal raw scores. The residual points of

20000 = (30000 − 25000) × 4 are awarded to the winner of the

game.

• Uma (placement bonus) is 10-20. That is, 1st player gets +20000,

2nd player gets +10000, 3rd player gets −10000, and 4th player

gets −20000.

• Each score is then scaled by dividing it by 1000 and rounding it

oﬀ.

It appears that European players are not very familiar with the

oka system (possibly because there is no oka in EMA rules), so let

me explain this with an example. Suppose that players A, B, C, and

24 CHAPTER 1. INTRODUCTION TO TENHOU ( )

Table 1.1: Final score calculation on Tenhou

Player Raw score Before uma After uma After oka

A 39000 9000 29000 49000

B 25100 −4900 5100 5100

C 22900 −7100 −17100 −17100

D 13000 −17000 −37000 −37000

D hold the following raw points at the end of a game; 39000, 25100,

22900, and 13000, as shown in Table 1.1 below.

The ﬁrst numerical column shows the raw scores. Then, 30000 is

subtracted from each of the raw scores (second column). Then, we

add uma to each score based on placements (third column). Finally,

we add oka to the winner’s score to obtain the ﬁnal scores (fourth

column).

The ﬁnal scores after adding

uma and oka and scaling will be

displayed along with the place-

ments and raw scores. In the ex-

ample to the right of this text, I (

means “me”) came in 1st, earning

50100 points (60.0 with uma and

oka), 2nd player (C , which

reads Mr. C) earns 46000 points

(+ 26.0 with uma), 3rd player earns

6700 points (−33.0 with uma), and

4th player went bankrupt (−2800 points, −53.0 with uma).

1.4. PLAYING A GAME 25

bababababababababababababababab

Notes on placement

It is important to keep in mind that your rank and rating on Tenhou

depend solely on the placement in a game and not on how many

points you earn in a game, before or after adding uma and oka.

In other words, there is absolutely no diﬀerence between getting

1st place with 30000 points and getting 1st place with, say, 80000

points in terms of their contributions to your rank and rating.

This feature adds an interesting strategic element to the game.

That is, it makes it clearer that the goal of mahjong is not to win

a hand per se but to have a better placement at the end of a game.

Winning a hand is just one of several means to securing a good

placement. On occasion, you may ﬁnd it beneﬁcial to assist one

of your opponents instead of trying to win a hand yourself. Inten-

tionally dealing into an opponent’s hand can sometimes be a good

tactic when it serves the purpose of securing your own placement.

In my impression, many European players are lacking the appreci-

ation of this aspect of mahjong. I hope you will learn to appreciate

it through playing lots of games on Tenhou.

You might wonder why they still calculate the ﬁnal scores with uma

and oka in Tenhou if they are irrelevant; I honestly have no idea.

26 CHAPTER 1. INTRODUCTION TO TENHOU ( )

1.5 Troubleshooting

At times, you may get disconnected from the Tenhou server (pos-

sibly because of poor internet connection on your end or problems

on the server).

When a player gets disconnected from the server

during a game, the game still continues. The “auto discard draw”

will be turned on for the disconnected player, so they will be simply

discarding anything they draw until they return. The player name

will turn into dark red once a player is disconnected.

When you get disconnected,

you may get a warning mes-

sage shown above, asking you

if you would like to get con-

nected again. Click Yes if you

want to. However, a warning

message does not always show up when you get disconnected. When

a screen freezes during a game for more than 15 seconds, you should

suspect that you are disconnected. You may want to hit the refresh

button on your browser to get connected to the server again.

You can create more than one accounts on Tenhou, but you will

have to wait for 7 days unless your IP address changes. If you at-

tempt to create a second account from the same IP address within

7 days, you will get an error message shown below, telling you that

you cannot create a new account from your IP address in 7 days.

You will notice that players sometimes get disconnected on purpose to quit

playing, especially when they are losing badly.

1.6. RULES 27

1.6 Rules

Here is a summary of the rules on Tenhou.

• Three red ﬁves (one in each suit) in games with red ﬁves.

• No kuikae (swap-calling). That is, you cannot discard an iden-

tical tile after pon or chii. You cannot discard the tile from other

end of the run, either.

• “Sudden death” rule when no player has 30000 or more points

after South-4 (East-4 in East-only games).

• A game is terminated when a player goes bankrupt.

• Automatic agariyame rule (i.e., the game is automatically ter-

minated if the dealer is leading after the end of Sout h-4, even if

he won a hand in South-4).

• One-han minimum all the time (i.e., no two-han minimum even

after ﬁve counters).

• Abortive draw in the following situations

– (nine terminals / honors)

– (four riichi’s)

– (three players call ron on a discard)

– (four players discard the same wind)

– (four kan by diﬀerent players)

• (nagashi mangan) is allowed. You can declare it even

when you have called pon / chii. You cannot declare it if one

or more of your discards has been called by others.

28 CHAPTER 1. INTRODUCTION TO TENHOU ( )

• Up to two players can win on a discard. riichi bets and counter

bonus go to the player sitting closer to the player who discarded

the winning tile. The dealership remains if the dealer is one of

the winners.

• The following are recognized as yakuman: (Tenhou; Bless-

ing of Heaven) / (chihou; Blessing of Earth) / (daisangen;

Big Three Dragons) / (su anko; Four Concealed Pungs)

/ (su anko tanki; Single-Wait Four Concealed Pungs)

/ (tsuiisou; All Honors) / (ryuiisou; All Green)

/ (chinroutou; All Terminals) / (kokushi muso;

Thirteen Orphans) / 13 (Thirteen-wait Thirteen Or-

phans) / (daisushi; Big Four Winds) / (shosushi;

Little Four Winds) / (su kantsu; Four Kongs) /

(churenpoutou; Nine Gates) / (junsei churenpoutou;

Nine-wait Nine Gates).

• Yakuman can be combined. For example, (Big Three Drag-

ons) can be combined wit h (All Honors), (Four

Concealed Pungs), and either of (Four Kongs), (Bless-

ing of Heaven) or (Blessing of Earth), producing a quadru-

ple yakuman (128000 points).

• There is no double yakuman unless diﬀerent yakuman are com-

bined. For example, (Thirteen Orphans) and

13 (Thirteen-wait Thirteen Orphans) are both single yakuman .

• You cannot call pon / chii / kan on the last discard in a hand.

• Sekinin barai: a player who feeds the third Dragon pon / kan

to an opponent with two melded Dragon pon / kan must pay

the full value of the hand in case Big Three Dragons is made on

a self-draw. In case another player deals into it, the two share

1.6. RULES 29

the payment equally. The same rule applies to Big Four Winds,

but not to rinshan kaihou (After a Kong).

Table 1.2: List of yaku names

Yaku Pronunciation EMA name Han (open)

(menzen-) tsumo Fully Concealed Hand 1 (NA)

riichi Riichi 1 (NA)

ippatsu Ippatsu 1 (NA)

chankan Robbing the Kong 1

rinshan kaiho After a Kong 1

haitei (-moyue) Under the Sea 1

houtei (-raoyui) Under the River 1

jikaze Seat Wind 1

bakaze Prevailing Wind 1

yakuhai / fanpai Dragon Pung 1

tanyao All Simples 1

iipeiko Pure Double Chow 1 (NA)

pinfu Pinfu 1 (NA)

chanta Outside Hand 2 (1)

ittsu Pure Straight 2 (1)

sanshoku (-doujun) Mixed Triple Chow 2 (1)

sanshoku doukou Mixed Triple Pungs 2

double riichi

Double Riichi

2 (NA)

san kantsu Three Kongs 2

toitoi All Pungs 2

san anko Three Concealed Pungs 2

shousangen Little Three Dragons 2

honroutou All Terminals and Honors 2

chiitoitsu Seven Pairs 2 (NA)

junchan Ter minals in All Sets 3 (2)

honitsu Half Flush 3 (2)

ryanpeiko Twice Pure Double Chow 3 (NA)

chinitsu Full Flush 6 (5)

nagashi mangan – mangan

dora Dora

aka dora Red ﬁve

ura dora Ura dora

Chapter 2

Advanced features of Tenhou

2.1 Rank and rating

Tenhou has two diﬀerent player

Rank N Rank N

15 7780

130 5849

592 6481

1830 6383

3140 6971

5968 9964

9957 16606

14436 14509

18174 28283

15046 132411

Table 2.1: Player distribution

rating systems — rank (kyu / dan)

and R (rate). The kyu / dan rank-

ing system is similar to the one com-

monly used in Japanese arts, games,

and martial arts. The kyu ( ) ranks

are shown in arabic numbers, go-

ing from 9 to 1 in descend-

ing order. After passing 1 , you

enter the dan ( ) ranks, shown in

kanji numbers, going from

( ; ﬁrst dan) to (tenth dan)

in ascending order. Everyone starts

with (newbie; no rank), and if you pass the rank, you are

awarded the highest rank called (Tenhoui). Since t he incep-

tion of Tenhou in 2006, there have been only nine players who have

achieved at the time of writing this book. Table 2.1 shows the

distribution of active players holding each rank as of 20 December,

2015.

2.1.1 kyu / dan rank

To advance your kyu / dan rank, you need to earn points (called

“pt” or “ pt” on Tenhou). For example, to proceed from the

(newbie) status to the 9 (kyu) rank, you need to earn 30 points.

Required amount of points for promotion gets greater and greater as

2.1. RANK AND RATING 31

you move further up. For example, to proceed from (sixth dan)

to (seventh dan), you need to earn as many as 1200 points.

To ﬁnd out how many more points you need to earn to advance

to the next rank from the current rank, see the top right part of the

main page.

Your current

rank (7 )

pt you have /

pt you need

Your R

In this example, the player currently holds the rank of 7 . The part

that reads “30 / 60 pt” means that he has earned 30 points since he

became 7 and t hat he needs 60 points in total to be promoted to

6 .

When you rise or fall in rank, your points will be reset to a default

value. For kyu rank players, the default value is 0 points. For dan

rank players, the default value is diﬀerent depending on ranks. For

example, the default points for players are 1200 points. When

they get 1200 more points and reach 2400 points, they get promoted

to . When they lose all the initial 1200 points and reach 0 points,

they get demoted to .

The amount of points you earn or lose in each game depends on

your placement (but not scores with uma and oka), the type of game

(East-only or East–South), the room in which the game is played (

, , , or ), and your current rank.

You gain positive

points only if you come in ﬁrst or second place. If you come in ﬁrst

place, you will gain the following points regardless of your rank.

Points you earn or lose in East-only games are two-thirds of those in East–South

games.

32 CHAPTER 2. ADVANCED FEATURES OF TENHOU

• 45 points in the (ippan) room

• 60 points in the (joukyu) room

• 75 points in the (tokujou) room

• 90 points in the (houou) room

If you come in second place, you will gain the following points re-

gardless of your rank.

• 0 points in the room

• 15 points in the room

• 30 points in the room

• 45 points in the room

You don’t gain or lose points if you come in third place. The points

you lose when coming in fourth place depend on your rank but not

on the room. When your rank is 3 or below, you lose 0 point. How-

ever, each time your rank rises above 3 , the points you lose get big-

ger by 15 points. That is, 2 players lose 15 points if they come in

fourth place; 1 players lose 15 × 2 = 30 points; players lose

15 × 3 = 45 points, ... , and players lose as many as 180 points if

they come in fourth place.

Notice how severe the punishment is for coming in fourth, and

it gets severer and severer as your rank goes up. This is one of the

distinctive features of Tenhou. Avoiding t he fourth place tends to be

players’ top priority in Tenhou games. This is in contrast to standard

mahjong games, where the reward for coming in ﬁrst usually out-

weighs the cost of coming in fourth, thanks to the oka system.

Recall that, although Tenhou does adopt the oka system, it is the placement,

not the scores, that determines the points you earn or lose. In this sense, EMA

games are actually more similar to Tenhou games than to standard games. Since

there is no oka in EMA games, the reward for coming in ﬁrst is much smaller

than that in standard games.

2.1. RANK AND RATING 33

To easily ﬁnd out how many

points you earn / lose for each place

in a given type of game for your rank,

mouseover the button in each

cell on the left-hand side of the main

page. Then, you will see something

like the picture above on the right-hand side of the main page. Un-

der the second bullet point, we see that, for this player’s rank ( ),

the point reward is: +75 for ﬁrst place, +30 for second place, 0 points

for third place, and −105 for fourth place.

When you earn enough

points for promotion in a game,

a new rank is awarded after

the game. A certiﬁcate mes-

sage like the picture to the

right of this text will pop up af-

ter the game.

Since you never get negative points in games until you reach 2

and there is no demotion until you reach (ﬁrst dan), it should

be relatively easy to reach . In fact, even without studying the

contents of this book, you can perhaps reach as high as (fourth

dan) if you play a few hundred games or so. However, moving fur-

ther up will probably require that you study basic strategies and tile

eﬃciency theories.

2.1.2 Rate (R)

In addition to the kyu / dan rank, Tenhou gives each player an-

other rating called R. The initial value of R is 1500, and higher-rank

players tend to have a higher R. For example, t he average R among

34 CHAPTER 2. ADVANCED FEATURES OF TENHOU

the players is 2248.

While kyu / dan rank remains relatively stable, R can change af-

ter each game. R is calculated based on your placement in a game,

but it also depends on the average R of the players you play with. A

change in R after a game, ∆R , is calculated with the following for-

mula:

∆R = (P +

R) × G

where

• P is based on your placement in the game: + 30 for ﬁrst, +10 for

second, −10 for third, and −30 for fourth;

•

R is an adjustment that reﬂects how strong your opponents are,

calculated as (Average R in the game − your R) / 40; and

• G is an adjustment based on n, the number of games you have

played before. If n ≤ 400, G is equal to 1 − 0.002 × n. If n > 400,

G is set equal to 0.2.

R initially ﬂuctuates a lot, as the scaling factor G is very close to

1 until you play many games. R may go up or down by 30 or so for

each of the ﬁrst 100 games or so. As you play more games, however,

the ﬂuctuation gets smaller and smaller as G approaches to 0.2.

Notice what the adjustment

R does. This factor is positive when

you play against players who are “stronger” than you (i.e., have a

higher R than you) while it is negative when you play against play-

ers who are “weaker” than you. Therefore, when you win against

stronger players, your reward will be bigger than when winning against

weaker players. Likewise, when you lose against weaker players,

your punishment will be severer than when losing against stronger

2.2. FOUR ROOMS 35

players. Because of these features, one might say that your R better

reﬂects your skill levels than your kyu / dan rank.

2.2 Four rooms

As we have seen, there are four diﬀerent rooms where ranking

matches are played. Qualiﬁcations to play in each room are based

on your rank and R.

1. (ippan; lower-level room)

This is the only room where you can play initially. Players with

an R higher than 1800 and a rank higher than are not allowed to

play here, however. Games in this room can sometimes be a bit ran-

dom, even chaotic at times. Some of the players in this room probably

do not understand the rules very well. You very rarely come across

strong players here.

2. (joukyu; upper-level room)

You can play here if (1) your rank is 1 or higher or (2) you buy

a two-month membership (¥ 1080 = e 8 = £ 6).

Players with an R

higher than 2000 and a rank higher than are not allowed to play

in this room, however.

Games in the joukyu room are more reasonable than those in the

lower-level room, but you still see many players who do not defend

at all, do meaningless dama / unreasonable riichi, and make serious

mistakes in maximizing tile eﬃciency. In my impression, games at

EMA tournaments most resemble games in the ippan and joukyu

rooms.

If you want to pay for the membership, click on the link that appears when you

click the sub-tab. Keep in mind that you need to buy 60 days’ worth of

membership. Choose “60 (1080 )” in the payment page.

36 CHAPTER 2. ADVANCED FEATURES OF TENHOU

3. (tokujou; advanced room)

Requirements to play in this room are pretty demanding. You

have to have a or higher rank and a 1800 or higher R. The latter

requirement is particularly diﬃcult to satisfy for intermediate play-

ers. As I wrote above, achieving the rank of is not that diﬃcult,

but satisfying the R ≥ 1800 condition requires t hat you take mahjong

rather seriously. Since weak players are shut out from the tokujou

room, games in tokujou are qualitatively diﬀerent from those in the

joukyu and ippan rooms. Games in this room feel similar to those

you’d experience at regular (furii ) mahjong parlors in Japan.

4. (houou; phoenix room)

This is the highest-level room in Tenhou. In order to play in this

room, you have to have all of the following: (1) a or higher rank,

(2) a 2000 or higher R, and (3) a paid membership (¥ 540 yen = e 4

= £ 3 per month). Satisfying the ﬁrst two conditions can be really,

really challenging.

This is arguably one of the highest-level mahjong locales in the

whole world. It is not uncommon for you to come across a houou-

level player at a regular mahjong parlor in Japan. However, you usu-

ally play against at most one houou-level player at a table, and t he two

other players at t he table are either tokujou- or joukyu-level players.

What is remarkable about games in the houou room is that you will

be surrounded by three other houou-level players. It would be safe

to say that no other public mahjong locale in the world — whether it

is online or oﬄine — could oﬀer a comparable experience.

Perhaps the highest-level leagues in professional mahjong associations in Japan

have players who are of comparable quality, but you have to become a profes-

sional player to play at such leagues. Even after becoming a professional, you

will need at least a few years to reach the highest league.

2.3. READING THE STATISTICS 37

2.3 Reading the statistics

After you play 30 games or so, you may want to start paying atten-

tion to the statistics shown on t he right-hand side of the main page.

The upper half of the player statistics shows your statistics for the

entire period, whereas the bottom half shows your statistics in the

present month for a given type of game in a given room.

2.3.1 Overall statistics

The picture below show my old player statistics (upper half) back

from when I had a rank. Let me explain how to read these statis-

tics.

Entire period (4-player games)

ﬁrst place

second place

third place

fourth place

bankruptcy

win rate

deal-in rate

call rate

riichi rate

Below a player name is the expiration date of my premium mem-

bership (17 November, 2015). When I started playing Tenhou on 17

September, 2015, I bought a 60-day membership so I can play in the

joukyu room. If you have just created a Tenhou account, the expira-

tion date will be shown as today’s or tomorrow’s date, since we are

given a 1-day premium membership when we open an account.

Af-

ter a day or two, it will turn into “—-/–/–” meaning t hat you do not

have a premium membership.

There is really no point in reading too much into the statistics when you have

played only a few games; the sample size is too small to be meaningful.

You need to have a premium membership to use Tenhou’s Windows client.

38 CHAPTER 2. ADVANCED FEATURES OF TENHOU

The box below the expiration date that reads / 4

indicates that the statistics below are for the entire period (not

just this month) and for 4-player games (not 3-player games). Below

that, we see that I had a rank, 565 points (the initial 400 points

plus 165 points earned after I became ) out of the 800 points I

need for promotion, and an R of 1987.

Three columns below these display my statistics. The ﬁrst col-

umn shows my placement rates. I had come in ﬁrst place 50% of the

games, second place 32.5%, third place 7.5%, fourth place 10 %, and

gone bankrupt 7.5% of the games. Ideally, you’d want your ﬁrst place

rate to be greater than your second place rate, your second place rate

greater than your third place rate, etc.

The middle column provides the following information. First,

shows t he number of games you have played. At this point, I had

played 40 games. Second, shows the average score (with

oka and uma) from all the games I have played. As I said before, this

does not inﬂuence your R nor rank. Third, shows the aver-

age placement. If you have obtained each of the four places equally,

the average placement would be 2.5

(

1n+2n+3n+4n

= 2.5

)

. Therefore,

any values below 2.5 indicate that you are, on average, winning more

than losing. The two rows that follow (shown in light gray) are rel-

evant only if you play games in private rooms. Since I have only

played ranking matches, they are left blank.

The third column shows my statistics based on hand-level perfor-

mance. First, (houra rate; agari rate; win rate) is the number

of hands you have won divided by the total number of hands you

have played in all games.

The denominator includes hands where no one won.

2.3. READING THE STATISTICS 39

Second, (houju rate; deal-in rate) is the number of times

you have fed the winning tile to an opponent’s hand divided by the

total number of hands you have played. You would want this rate to

be lower, but keep in mind that (1) sometimes you would be better

oﬀ dealing into an opponent’s hand to secure your placement, and

(2) sometimes you need to discard dangerous tiles (which would in-

crease your deal-in rate, on average) in order to increase the chance

of winning your hand (which would increase your win rate, on aver-

age). The rule-of-thumb is that the diﬀerence between your win rate

and deal-in rate (win rate - deal-in rate) should be at least 10 percent-

age points. That is, if you have a high deal-in rate, you need your win

rate to be higher. Likewise, if you have a low deal-in rate, it is OK to

have a lower win rate as well.

Third, (fuuro rate; call rate) is the number of hands where

you have called chii / pon / kan divided by the total number of

hands you have played. Finally, (riichi rate) is the number

of riichi calls you made divided by the number of hands you have

played.

The ranking page on Tenhou

has a table that summarizes the av-

erage values of these statistics among players with diﬀerent ranks

(under the heading that reads ). You may

want to compare your statistics with the average values among play-

ers who share your rank or those who have higher ranks than you do.

Figure 2.1 summarizes the average values of hand-level performance

statistics for players in diﬀerent ranks.

We can see some interesting patterns here. The left-hand side

panel compares average win rates ( ) and deal-in rates (

http://goo.gl/suyQ5

40 CHAPTER 2. ADVANCED FEATURES OF TENHOU

Figure 2.1: Average hand-performance statistics

Note: These graphs show the hand-performance statistics reported in a table on the

ranking page (http://goo.gl/suyQ5) as of 20 December, 2015.

) for diﬀerent ranks. Notice that the average win rate is relatively

constant across diﬀerent ranks; once you pass the rank, it stays

around 20-22 %.

On the other hand, the average deal-in rate is steadily decreasing

after players move from the kyu ranks into the dan ranks. It is around

15% for almost all kyu rank players (except for and 1 ), but it

keeps going lower and lower as players rise in the dan rank. The

fact that high-dan players have lower deal-in rates on average is re-

markable, considering that they are facing stronger opponents than

low-dan players do. This pattern signiﬁes the importance of defen-

sive skills.

Another interesting thing to notice on the left-hand side panel

is that the average scores deteriorate once you move from 2 to

1 (i.e., average win rate gets lower, and average deal-in rate gets

higher).

I can think of two reasons for why this happens. First, 1 is

where most players start playing in the joukyu (upper-level) room,

2.3. READING THE STATISTICS 41

where average player skills are much higher t han those in the ippan

(lower-level) room. If a player who belongs to the lower-level room

plays in the upper-level room, their performance will necessarily go

down, making it look that 1 players are worse than 2 players

even if they are not. Second, if you keep losing as a (ﬁrst dan)

player, you get demoted to 1 but you will never be demoted to 2

. This means that 1 players might actually be worse than 2

players, on average.

The right-hand side panel shows the average call rates ( )

and riichi rates ( ) for diﬀerent ranks. The former is increasing

as rank goes up, while the latter is decreasing, but the changes are

rather gradual for both rates.

42 CHAPTER 2. ADVANCED FEATURES OF TENHOU

2.3.2 Monthly statistics

The bottom-right part of the main page shows monthly statistics

from games you have played in a given room. The box is a pull-

down menu that lets you choose the room ( , , , )

and game type (East-only, East–South, with or without open tanyao ,

red ﬁves, etc.). In the example below, the box reads /

, which means the following: means monthly, is short

for (joukyu)

, means with open tanyao and red ﬁves.

Cumulative Average

Scores

Placement

ﬁrst place

fourth place

Below the box, you see the raw placement scores. In this example,

17+11+3+4 = 35 means that I have played 35 games this month,

and I came in ﬁrst place in 17 games, second place in 11 games, third

place in 3 games, and fourth place in 4 games. R shown here (1987)

should be the same as the R you see in the top part. 3382 means

that R=1987 puts me in 3382th place among all the active players on

Tenhou.

Two columns follow, where the left column shows the monthly

cumulative values and the right column shows the monthly average

values. In the ﬁrst row that reads shows the monthly cumula-

tive or average scores from games (after adding oka

Likewise, is short for (ippan), is for (tokujou), is for

(houou).

2.4. VIEWING GAMES 43

and uma). In this example, my cumulative score is 727 from the 35

games I played, which puts me in 106th place among players who

have played 30 or more games this mont h. Similarly,

my average score is 20.7 (= 727/35), which puts me in 5th place.

Your placement for average scores will not be shown unless you have

played 30 or more games of a given type in a given room in a given

month.

In the second row that reads shows cumulative or average

placement from games. The cumulative placement is based on place-

ment values (+30, +10, −10, or −30), whereas the average placement

is based on raw placement (1, 2, 3, or 4). The (total) score is the

sum of four placements: cumulative , cumulative , average

, and average . In this example, I earn 106th, 80th, 5th, and

3th places for these scores, so my total score is 106+80+5+3 = 194 (the

lower, the better), which puts me in 12th place among all the players

who have played 30 or more games this month. At t he

bottom, you see (ﬁrst place rate), (fourth place rate),

and (ﬁrst or second place rate) for games this

month.

2.4 Viewing games

2.4.1 Game replay ( )

Tenhou keeps t he record of all the games played there, giving each

game a unique URL. You can easily take a look at any of the last 40

games you have played on the (haifu; game record) tab on the

main page. Click on any of the link shown in the tab to start

a replay of the game. You can choose to view the game from any of

the four players’ viewpoint, not to show t he hands of the other three

players, or to go back and forth between turns / hands, etc. When

44 CHAPTER 2. ADVANCED FEATURES OF TENHOU

we play mahjong, we often wonder what the opponents are doing

(e.g., what are their waits? are they doing honitsu?, etc.). You can

ﬁnd out t he answers to these questions after the game by taking a

look at the game record.

If you would like to have someone take a look at a particular game

you played to ask for their opinions, you need to ﬁnd the unique URL

assigned to the game you want to show. You can ﬁnd out the URLs of

the last 40 games by going to the menu from the

pull-down on the main page. Clicking on will open a new

pop-up screen.

You can choose to open a game replay in t he current window (

), in a new pop-up window (

), or in a new window ( ) from the pull-

down menu above. Once you are happy with your choice, click on

one of the (replay) link next to the game you want replayed. You

will be taken to a page that looks like the one you saw after clicking

on the Play button on the top page of Tenhou. You can now ﬁnd out

the URL assigned to the game in the URL ﬁeld of your browser.

To start a replay, click on a link that reads » Flash

shown at the bottom of the page. Clicking on the HTML+JS

link will also work, but this one is the low-quality

picture version with limited options.

2.4. VIEWING GAMES 45

2.4.2 Spectating games ( )

You can watch games played in the (tokujou; advanced) and

the (houou; phoenix) rooms quasi-real time (with a ﬁve-minute

delay). Click on the (kansen; spectating) tab from the main page

and you will see the list of games you can watch. Click on one of

the player name links to start spectating the game from the chosen

player’s viewpoint.

Part II

Basic Tile Efficiency

Chapter 3

Riichi mahjong basics

3.1 Learning strategies

Mahjong is a game of skill and luck. There is a set of strategy

principles you can learn to improve your skills, but acquiring skills is

neither necessary nor suﬃcient to win a game. On the contrary, with

luck, an unskilled player can easily defeat strong players in mahjong.

At least in the short run, game outcomes are governed more by luck

than by skills.

However, learning strategy principles is crucial to

improve your performance in the long run. Moreover, you will be

able to enjoy the game in greater depth once you understand these

principles.

Because of the probabilistic nature of the game, making the best

choice does not always lead to the best outcome. The best choices

are those t hat lead to the best outcome, on average. An evaluation of

our choices thus requires a probabilistic (i.e., statistical) assessment of

diﬀerent options. For example, consider the following hand.

‰‰““”´ˆ˜˚ˇ˘ $$L

What would you discard?

This hand becomes ready to win if you discard “ or ”. Let’s com-

pare the two choices.

An interesting question would be: how short is t he “short” run here. That is,

how many games do we need in order to discern a strong player from weak

players? Studies show that we would need at least 100 games or so to have a

reliable estimate of our skill levels. Given that EMA tournaments usually have

only 8 games, winning at these tournaments requires quite a bit of luck.

3.1. LEARNING STRATEGIES 49

• Discard “ ⇒ you wait for › „ (2 kinds–8 tiles)

• Discard ” ⇒ you wait for ‰ “ (2 kinds–4 tiles)

Which discard choice is better? Although both of the two choices

yield a 2-way wait, waiting for › „ is much better than waiting

for ‰ “, probabilistically speaking. With the › „ wait, there are

four tiles of › and another four tiles of „ to win on, leaving at most

eight winning tiles.

With the ‰ “ wait, on the other hand, you

have already used up two tiles of ‰ and two tiles of “ yourself,

leaving at most four winning tiles. It is clearly better to choose t he

› „ wait over the ‰ “ wait, because that will give you a higher

probability of winning this hand.

It is possible that, after you decided on the › „ wait, your op-

ponents end up not discarding › or „ at all, while discarding lots

of ‰ “. This is the kind of things that will happen often in mahjong

(or in any game of luck, for that matter). When things like this hap-

pen, do not t hink that you made a bad call; you didn’t. You made

the right choice, but you were just unlucky. When we experience

this kind of bad luck, we just need to keep calm and carry on.

Before discussing a practical method of maximizing tile eﬃciency

in the next chapter, I will discuss some basic principles of tile eﬃ-

ciency in this chapter. In doing so, I introduce several key terms we

use in later chapters. I will also provide the original Japanese term

for each (shown in this font). I do so because you may ﬁnd these

Japanese terms used in some online strategy discussions in English.

Of course, the number of winning tiles could be smaller than eight if some of

them have already been discarded.

50 CHAPTER 3. RIICHI MAHJONG BASICS

3.2 Basic building blocks

3.2.1 Tiles

Mahjong tiles can be classiﬁed into two categories — number tiles

and honor tiles.

Number tiles

Cracks (characters) Manzu –—‰ıȷﬀﬁﬂ

Dots (circles) Pinzu ¸˛‚‹›“”„«

Bamboos Souzu ´ˆ˜¨˝˚ˇ˘¯

We further classify number tiles into simples (tanyao hai; tiles be-

tween 2 and 8) and terminals (1 and 9). They are diﬀerentiated be-

cause they serve diﬀerent yaku and generate diﬀerent minipoints

(fu).

It has become quite common to include some red ﬁve tiles. For

example, most games on Tenhou have one red ﬁve tile in each suit,

» ˙ `. These tiles are included in place of regular ﬁves; we have

three regular ﬁves and one red ﬁve in each suit. Red ﬁves are treated

as dora regardless of the dora indicator. When a 4 in a given suit is

the dora indicator, the red ﬁve in that suit will be a double dora tile.

3.2. BASIC BUILDING BLOCKS 51

Honor tiles

Dragon tiles # $ %

White (haku) Green (hatsu) Red (chun)

Wind tiles ﬄ ! "

East (ton) South (nan) West (sha) North (pei)

Some honor tiles are value tiles (fanpai / yakuhai); we get one han

if we collect three identical value tiles. All dragon tiles are value tiles

regardless of the round and seating. On the other hand, the value

status of wind tiles depends on the round and the seating. East tiles

are value tiles for everyone during the East round, and South tiles are

value tiles for everyone during the South round. In addition, each

player gets their own seating wind as a value tile. For example, West

tiles are value tiles only for the West player, but they are valueless

wind tiles (otakaze) for other players.

52 CHAPTER 3. RIICHI MAHJONG BASICS

3.2.2 Group (mentsu)

One of the major goals in playing mahjong is to win a hand.

win a standard hand, we need to complete four groups ( mentsu) and

one head (atama; ﬁnal pair).

Groups can be classiﬁed into two kinds

— run and set.

• Run (shuntsu; chow / sequence) is a set of three consecutive

number tiles: e.g., ﬀﬁﬂ, ˜¨˝.

• Set (kotsu; pung / triplet) is a set of three identical tiles: e.g.,

———, $$$.

3.2.3 Ready and n-away

We say a hand is ready (tenpai) when the hand can be complete

with one more tile. For example, the following hand is ready.

Ready hand

‚‹›„„ˆ˜˚ˇ˘%%%

This hand becomes complete wit h either ´ or ¨. We say that this

hand waits for ´ ¨.

Another important goal is not to deal into an opponent’s hand. See Chapter 8

for discussions of defense strategies. However, the most important goal of all

is to win a game. Winning a hand and playing defense are merely two means

to this end. See Chapter 10 for more discussions of t his.

There are three exceptions to this; chiitoitsu (Seven Pairs), kokushi musou

(Thirteen Orphans), and nagashi mangan (All Terminals and Honors Discard)

do not require four groups and one head.

EMA rules refer to run as “chow” and set as “pung.” I realize that my use of

diﬀerent terminology here might be confusing at ﬁrst, but I hope you will get

used to it soon.

Technically speaking, there is a third type of groups, namely quad (kantsu;

kong), a set of four identical tiles. We treat quads as a variant of sets. See

Section 9.3 for discussions on this.

3.2. BASIC BUILDING BLOCKS 53

We say a hand is 1-away from ready (1-shanten) when the hand

can become ready with one more tile. For example, the following

hand is 1-away from ready.

1-away hand

‚››„„ˆ˜˚ˇ˘%%%

This hand becomes ready if you draw any of ‹›„´¨. We say

this hand accepts ‹›„´¨ (5 kinds–16 tiles) as any of them can

make this hand advance from 1-away to ready. Tile acceptance (ukeire)

refers to the kinds and the number of tiles a hand can accept. Other

things being equal, having a 1-away hand with greater tile accep-

tance is better than having one with smaller tile acceptance.

More generally, we say a hand is n-away from ready (n-shanten)

when the hand can be ready with n more steps. For example, the

following hand is 2-away from ready.

2-away hand

‚››„„ˆ˜˚ˇˇ%%%

This hand accepts all the tiles that the 1-away hand above accepts

(‹›„´¨), plus seven additional kinds of tiles ‚ˆ˜˝˚ˇ˘.

The hand will become 1-away if any of these tiles gets drawn.

A hand can also be 3-away, 4-away, 5-away, or 6-away from ready.

In practice, however, there is not much point in distinguishing 3-

away hands from 4-away (or worse) hands. You thus need to be able

‚ˆ˜˚ will make this hand 1-away for chiitoitsu (Seven Pairs).

6-away happens when a hand has no pair, in which case it takes 6 more tiles to

make it ready for chiitoitsu.

54 CHAPTER 3. RIICHI MAHJONG BASICS

to distinguish between four kinds of hands — ready hands, 1-away

hands, 2-away hands, and 3-away or worse hands.

Tile acceptance shrinkage

As n gets smaller and the hand gets closer to completion, the

kinds and the number of tiles it can accept will necessarily get smaller.

Consider the three stages of a hand we have seen above.

• When 2-away, it accepts: ‚‹›„´ˆ˜¨˝˚ˇ˘.

• When 1-away, it accepts: ‹›„´¨.

• When ready, it waits for: ´¨.

Tile acceptance is minimized when the hand is ready. Note also that

it is virtually minimized when it is 1-away. This is because with a

ready hand you can utilize not only the tiles you draw but also the

tiles discarded by others to complete the hand. With n-away hands,

however, you have to rely (almost) solely on the tiles you draw your-

self to advance your hand.

Therefore, in choosing a discard from a

2-away hand, we should try not to make for a 1-away hand with too

small tile acceptance.

Advancing your hand

To win a hand, we need to advance our hand by reducing the n of

an n-away hand until it is ready. When a hand is 2-away, we should

aim to make the hand 1-away. When a hand is 1-away, we should aim

to make the hand ready. For example, consider the following hand.

Melding (calling pon / chii) is not always possible. For example, the 2-away

hand above can accept ˜ if you draw one, but you can neither pon nor chii ˜.

3.2. BASIC BUILDING BLOCKS 55

2-away vs. 1-away

ıȷﬀﬀ‚‹¨˝˝ˇ¯$$$

What would you discard?

Discarding ¯ makes the hand 2-away, whereas discarding either ﬀ

or ¨ makes the hand 1-away. You should thus discard ﬀ or ¨ to

make the hand 1-away. Reverting a 1-away hand to 2-away makes

sense only in some exceptional cases where tile acceptance at 1-away

becomes unbearably small (i.e., fewer than 2 kinds). With this hand,

the hand will be able to accept ˛›˘ (3 kinds–12 tiles) when it be-

comes 1-away.

3.2.4 Protoruns (taatsu)

Of the two kinds of groups, it is easier to complete a run than

to complete a set. There are only four identical tiles, and complet-

ing a set requires that you collect three out of the four identical tiles.

Therefore, we usually prioritize runs over sets in advancing a hand.

A pair of tiles that can become a run with one more tile is called a

protorun (taatsu). There are three types of protoruns, summarized

in Table 3.1.

Table 3.1: Types of protoruns

Name Japanese Example Wait Acceptance

side wait ryanmen ‰ —-ı 2 kinds–8 tiles

closed wait kanchan ˛‹ ‚ 1 kind–4 tiles

edge wait penchan ˘¯ ˇ 1 kind–4 tiles

As we can see in the table, a side-wait (ryanmen) protorun can

56 CHAPTER 3. RIICHI MAHJONG BASICS

accept twice as many tiles as a closed-wait (kanchan) protorun or

an edge-wait (penchan) protorun can. Therefore, building side-wait

protoruns is the key to advancing a hand. Winning tiles of side-wait

protoruns are often denoted with a hyphen in the middle, such as

´-¨ or ›-„.

Closed wait vs. edge wait

There is no diﬀerence in the kinds and the number of tiles ac-

cepted by closed-wait and edge-wait protoruns; they both accept 1

kind–4 tiles. However, closed-wait protoruns are superior to edge-

wait ones because they can more easily evolve into a side-wait pro-

torun.

A closed-wait protorun can evolve into a side-wait protorun in

just one step. For example, a protorun ¸‚ can become a side-wait

one if you draw ‹ and discard ¸.

¸‚ ⇒ ‚‹

draw ‹

On the other hand, it requires two steps for an edge-wait protorun to

evolve into a side-wait protorun. For example, a protorun „« can

become a side-wait one if you draw “ ﬁrst and then ›.

„« ⇒ “„ ⇒ ›“

draw “ draw ›

Value ranking of protoruns

side wait > closed wait > edge wait

3.2. BASIC BUILDING BLOCKS 57

Tile versatility

Some tiles are more versatile than others. For example, number

tiles are more versatile t han honor tiles because honor tiles can never

form a run. Moreover, we can rank order the versatility of number

tiles based on the types of protoruns they can form.

Number tiles between 3 and 7 are the most versatile. This is be-

cause each of them can form a protorun with four kinds of number

tiles. For example, ‚ can form a protorun with ¸, ˛, ‹, and ›.

Two out of the four resulting protoruns will be side wait.

2 and 8 are less versatile. They can form a protorun with only

three kinds of number tiles. For example, ˛ can form a protorun

with ¸, ‚, and ‹. Only one out of the three resulting protoruns is

side wait.

Terminals (1 and 9) are the least versatile. They can form a pro-

torun with only two kinds of tiles. For example, ¸ can form a pro-

torun only with ˛ and ‚. Neither of the two resulting protoruns is

side wait.

Versatility ranking of tiles

3–7 tiles > 2, 8 tiles > 1, 9 tiles > honor tiles

Applying the same logic, we can also rank order the versatility of

closed-wait protoruns. For example, a closed-wait protorun ¸‚

can become a side-wait one only if we draw ‹. Likewise, a closed-

wait protorun ˛‹ can become a side-wait one only if we draw ›.

However, a closed-wait protorun ‚› can become a side-wait one if

we draw ˛ or “. Clearly, ‚› is more versatile than ¸‚ or ˛‹.

58 CHAPTER 3. RIICHI MAHJONG BASICS

Versatility ranking of closed-wait protoruns

35, 46, 57 > 13, 24, 68, 79

3.2.5 Pairs (toitsu)

A set of two identical tiles is called a pair (toitsu). Pairs can per-

form several diﬀerent roles. A pair can be the head (ﬁnal pair) of a

hand, a protoset (a candidate for a set), or a component of chiitoitsu

(Seven Pairs).

Building the head

Any hand — including Thirteen Orphans and Seven Pairs — re-

quires the head to be complete. Since building the head is much eas-

ier than building a group, we usually don’t worry too much about

the head. For example, consider the following hand.

Hand with no head

ﬀﬁﬂ˛‚‹›“”˜¨˝ˇ

This hand currently lacks the head and the wait is not very good.

The hand is complete only with ˇ (1 kind–3 tiles). However, if we

draw any of ȷﬂ¸˛‹›”„ˆ˜˝˚ (12 kinds–41 tiles), the

wait will be signiﬁcantly improved. For example, if we draw › and

discard ˇ, the hand becomes:

ﬀﬁﬂ˛‚‹››“”˜¨˝

This hand is now waiting for ˛ ›-„ (3 kinds–9 tiles). When a

3.2. BASIC BUILDING BLOCKS 59

hand is missing the head, it is often the case that the wait gets signif-

icantly improved quite easily.

3.2.6 Pairs and sets

Another important role a pair can play is to work as a candidate

for a set. Especially when a hand has two pairs, we can count on

one of the two pairs to become the head while the other becomes a

set. In other words, the value of pairs is maximized when there are

two (and only two) pairs in a hand. Let’s see why this is the case by

comparing hands with one, two, and three pairs.

1. Hand with one pair

–ı›“”„«ˆˆ˚ˇ"

This 2-away hand has one pair: ˆˆ. This pair is not very useful as

a candidate for a set for two reasons. First, if we draw another ˆ,

we will complete a set but then we will lose the head at the same

time. The hand will still be 2-away from ready after all. Second, the

probability of drawing another ˆ is not very high because there are

only two tiles left.

What if a hand has two pairs? Suppose we drew –and discarded

", as follows.

2. Hand with two pairs

––ı›“”„«ˆˆ˚ˇ

This hand is also 2-away, but it has two pairs: –– and ˆˆ. Each

of these pairs is now functioning as an eﬀective candidate for a set.

Whenever one pair becomes a set, the other pair becomes the head.

Drawing – or ˆ will advance this hand from 2-away to 1-away.

60 CHAPTER 3. RIICHI MAHJONG BASICS

Moreover, whereas the hand with one pair was able to accept two

tiles of ˆ, the hand with two pairs can accept four tiles (two of ˆand

two of –). The probability of drawing any one of four tiles is obvi-

ously higher than the probability of drawing any one of two tiles. In

general, for each additional pair in a hand, tile acceptance increases

by two.

What if a hand has three pairs? Suppose we draw ˚, as follows.

3. Hand with three pairs

––ı›“”„«ˆˆ˚ˇ ˚

Draw

If we keep the second ˚ and discard the ˇ or the ı, the hand has

three pairs. However, keeping three pairs in a hand is ineﬃcient. Re-

call that each additional pair increases tile acceptance by two tiles. In

this case, keeping a pair of ˚ means that the hand can accept two

additional tiles of ˚. However, doing so comes with a cost. Keep-

ing three pairs by discarding the ˇ means the hand can no longer

accept ˝ ˘ (2 kinds–8 tiles). The net tile acceptance gain will be

negative (2 − 8 = −6). Similarly, keeping three pairs by discarding

the ı means the hand can no longer accept ‰ (4 tiles). Therefore,

discarding a ˚ to maintain two pairs is the most eﬃcient.

What we have seen so far is generalizable beyond the current ex-

amples. As long as we intend to keep the hand closed (i.e., not calling

pon or chii), we should avoid having three pairs in a hand. Having

three pairs makes for the weakest form, whereas having two pairs

makes for t he strongest form.

What if there are four or more pairs? Whenever a hand has four pairs, it is

2-away from ready for chiitoitsu (Seven Pairs). It may be faster to pursue

chiitoitsu than pursuing a standard hand in such cases.

3.2. BASIC BUILDING BLOCKS 61

Value of pairs: closed hand

2 pairs > 1 pair, 4 pairs > 3 pairs

Open hand

There is an important caveat to the above rule. When we intend

to call pon, having three pairs is actually better than having two pairs.

This is because the hand will become a two-pair hand after we call

pon once. For example, consider the following hand.

Two pairs vs. three pairs

––ı‹›ˆ¨¨ˇ˘¯%% %

Dora

What would you discard?

We would deﬁnitely intend to call pon on %. Anticipating that, we

should discard ı to keep t hree pairs in this case rather than dis-

carding – to have two pairs. After calling pon on %, we will have a

choice between discarding  or ˆ.

––‹›ˆ¨¨ˇ˘¯ %M%

––‹›¨¨ˇ˘¯ %M%

In either case, the hand will have two pairs after calling pon.

Value of pairs: open hand

3 pairs > 2 pairs

62 CHAPTER 3. RIICHI MAHJONG BASICS

3.2.7 Perfect n-away

Perfect 1-away

When a 1-away hand has two side-wait protoruns and two pairs,

it is called perfect 1-away.

Perfect 1-away

—¸˛‚““˚ˇ###

The hand above is an example of perfect 1-away. It is called “perfect”

because this hand can become ready either by calling chii, calling

pon, or drawing a tile to complete a run or a set, and no matter how a

hand becomes ready, you will always have the option to choose side

wait as the ﬁnal wait.

Perfect 2-away

One step prior to achieving perfect 1-away, we may get a perfect

2-away hand. Perfect 2-away is made up with three side-wait pro-

toruns and three pairs, as follows.

Perfect 2-away

—˛‚““˚ˇˇ###

When a perfect 2-away hand becomes 1-away, it can always be perfect

1-away (unless you choose not to, for some reason). However, not all

perfect 1-away hands evolve from a perfect 2-away hand.

3.2. BASIC BUILDING BLOCKS 63

3.2.8 Putting things all together: an example

Let’s see some hand examples that illustrate how we can apply

the tile eﬃciency logics we have learned so far. Consider the follow-

ing 2-away hand.

Advancing a hand 1

ııﬀˆ¨˘˘¯˛‚”„« ˆ

Draw

What would you discard?

The hand now has three pairs, and we should avoid it. In order to re-

duce the number of pairs in this hand from three to two, our discard

candidates should be ı, ˆ, or ˘. Which one should we choose?

Recall t hat a closed-wait protorun of 57 is stronger than a closed-

wait protorun of 24 or an edge-wait protorun of 89. Therefore, it is

OK to cut down the ııﬀ shape to ıﬀ by discarding ı. This is

because ıﬀ can become a side-wait protorun relatively easily. On

the other hand, the ˆˆ¨ shape and the ˘˘¯ shape are both

weak; the ﬁrst can become a side-wait protorun only if we draw ˝,

and the second one will never become a side-wait protorun in one

step. Therefore, both ˆˆ¨ and ˘˘¯ should be kept as a can-

didate for the head or a group rather than making them into weak

closed-wait protoruns.

Let’s say we discard ı, and then we draw ﬁ, resulting in the

following hand.

Advancing a hand 2

ıﬀˆˆ¨˘˘¯˛‚”„« ﬁ

Draw

What would you discard?

64 CHAPTER 3. RIICHI MAHJONG BASICS

Now that we have a side-wait protorun ﬀﬁ, we should discard ı.

Let’s say we draw ˇ, resulting in the following hand.

Advancing a hand 3

ﬀﬁˆˆ¨˘˘¯˛‚”„« ˇ

Draw

What would you discard?

This hand is now 1-away from ready, and our discard choice is be-

tween ¨ and ˘. Both tiles are equally useless from our perspective,

and so we will eventually discard them both. The question is which

one we should discard ﬁrst. Recall that a 4 is more versatile than an

8. This means that ¨ in this hand may later become dangerous for

the opponents; we should thus discard ¨ now rather than later.

Let’s say we draw ‹ after that, resulting in the following hand.

Advancing a hand 4

ﬀﬁˆˆˇ˘˘¯˛‚”„« ‹

Draw

What would you discard?

The hand is now ready. We should discard ˘ and call riichi. If we

win on ﬂ, we can claim riichi, pinfu, and sanshoku (Mixed Triple

Chow), giving us 7700 points.

We will discuss scoring and yaku more extensively in later chapters.

3.3. COMPLEX SHAPES 65

3.3 Complex shapes

The three basic types of tile blocks we have covered so far —

groups (runs and sets), protoruns (side wait, closed wait, and edge

wait), and pairs — form the basis of any standard mahjong hands.

When a hand has some tiles that do not constitute any of these three

shapes, we treat them as ﬂoating tiles. For example, ı and ˚ in

the following hand are both ﬂoating tiles.

Hand with ﬂoating tiles

ıﬁﬂ˛‚‹›“”´ˆ˜˚

In addition to these basic blocks, we often come across complex

shapes that are made up of two or more groups, protoruns, pairs,

and ﬂoating tiles combined. It is useful to comprehend such com-

plex shapes as they are rather than breaking them up into smaller

parts. We will discuss three-tile complex shapes and four-tile com-

plex shapes in turn.

3.3.1 Three-tile complex shapes

There are two kinds of three-tile complex shapes — double closed

shape and protorun plus one shape.

Standard hands are those with four groups and one head. Non-standard hands

are chiitoitsu (Seven Pairs) and kokushi musou (Thirteen Orphans).

66 CHAPTER 3. RIICHI MAHJONG BASICS

Double closed (ryankan) shape

When two closed-wait protoruns are combined, we have a dou-

ble closed (ryankan) shape. There are ﬁve diﬀerent patterns in each

suit, as follows.

´˜˝ ˆ¨˚ ˜˝ˇ

¨˚˘ ˝ˇ¯

Each shape accepts as many as 2 kinds–8 tiles. For example, ´˜˝

accepts ˆ (4 tiles) and ¨ (4 tiles). This is twice as many as the num-

ber of tiles an isolated closed-wait protorun can accept.

Sometimes a double closed shape is embedded within a tile block,

making it diﬃcult to detect it. For example, consider the following

1-away hand.

Hand with a double closed shape

Red

‰»ıﬀﬁﬂ‹›““”ˇˇ ˛

Draw

What would you discard?

Before drawing ˛, the hand was already in a very good shape. It was

perfect 1-away, accepting any of ıȷ›„ˇ (6 kinds–19 tiles).

The question is whether we should keep ˛ and discard ı instead.

Notice that, if we keep ˛, we have a double closed shape ˛‹“.

This is because the block ˛‹›““” can be split into ˛‹“

and ›“”. If we keep ˛ and discard ı, the hand is still 1-away

from ready, accepting ȷ‚›„ (5 kinds–19 tiles). The beneﬁt of

discarding ı to keep the double closed shape is that t he hand can

3.3. COMPLEX SHAPES 67

always be pinfu when it is ready. On the other hand, discarding ˛

means that the hand may become a yaku-less hand when drawing ı

Double closed shapes are particularly useful when a hand is rel-

atively far from ready (2-away or worse). As a hand advances, how-

ever, its usefulness diminishes because this block requires three (not

two) tiles even though it is not a complete group. Moreover, it will

ultimately become a single closed-wait protorun when t his block re-

mains incomplete when the hand is ready. Therefore, we should not

rely too much on a double closed shape. For example, consider the

following two hands.

––‰ı“”„ˆ¨˚%% —

Draw

‰ﬀﬁﬂ‚‹›„„ˆ¨˚ ”

Draw

Both hands are 1-away from ready and both contain a double

closed shape in souzu (bamboos) tiles. Maintaining the double closed

shape in these cases will not be ideal. It is true that, if the hand be-

comes ready by drawing ˜ or ˝ ﬁrst, each of the hands makes for

a good-wait ready hand. However, if the ﬁrst hand becomes ready

by calling pon on % or the second hand becomes ready by drawing

— or ı ﬁrst, they only make for a closed-wait ready hand.

Therefore, when we draw a tile next to the head, creating a side-

wait protorun, we should keep it and break the double closed shape

instead. In the ﬁrst example above, as we draw — t hat creates a side-

wait protorun ‰ı, we should keep it and discard t he ˆ instead.

68 CHAPTER 3. RIICHI MAHJONG BASICS

In the second example above, as we draw ” that creates a side-wait

protorun ”„, we should keep it and discard ˆ instead.

Protorun plus one shape

As we saw with the ﬁrst example in Section 3.2.8, we often come

across a tile combination such as ııﬀ that is made up with one

protorun plus one ﬂoating tile (ıﬀ + ı).

Depending on the type

of protoruns, we can classify protorun plus one shapes into three

types, as summarized in Table 3.2.

Table 3.2: Types of protorun plus one shapes

Name Example Wait Acceptance

side wait +1 ‰ —-ı  3 kinds–10 tiles

closed wait +1 ˛˛‹ ˛ ‚ 2 kinds–6 tiles

edge wait +1 ˘˘¯ ˇ ˘ 2 kinds–6 tiles

A protorun plus one can accept two additional tiles that an iso-

lated protorun cannot. This is because these blocks can now be a

candidate for a set as well as for a run.

Breaking a protorun plus one can be ineﬃcient. For example, if

we break a closed wait plus one shape into an isolated pair (i.e., dis-

card ‹ from ˛˛‹), the tile acceptance decreases from 6 to 2; it

can accept only ˛ (1 kind–2 tiles). Similarly, if we break it into an

isolated protorun (i.e., discard ˛ from ˛˛‹), the tile acceptance

decreases from 6 to 4; it can accept only ‚ (1 kind–4 tiles). With this

in mind, consider the following hand.

Alternatively, we can think of these combinations as a pair plus one ıı + ﬀ.

3.3. COMPLEX SHAPES 69

Protorun plus one

––ıﬀ›´ˆ˜¨ˇ˘¯¯ ı

Draw

What would you discard?

Discarding ı or ﬀ to break the protorun plus one ııﬀ is ineﬃ-

cient here. Discarding ı decreases tile acceptance by two, and dis-

carding ﬀ decreases tile acceptance by four. Moreover, discarding

ﬀ leaves three pairs in this hand, which should be avoided. Discard-

ing ¯ is much more eﬃcient.

Sometimes we have to make a choice between multiple protorun

plus one shapes, just like we did in examples in Section 3.2.8. Con-

sider the following hand. What would you discard?

Multiple protoruns plus one

‰‰ȷ‚‚‹´ˆ˜¨˚˘""

What would you discard?

There are two protorun plus one shapes in this hand: ‰‰ȷ and

‚‚‹. We have to break one of the two into either an isolated Pair

or an isolated protorun, because the other parts of t his hand are more

or less self-suﬃcient. Which one should we choose?

When choosing between which protoruns plus one to break, pri-

ority should be given to the weaker one. Since the side-wait pro-

torun ‚‹ is much stronger than the closed-wait protorun ‰ȷ, we

should prioritize the latter and maintain ‰‰ȷ. In other words, the

side-wait protorun ‚‹ is so strong that we do not need to provide

a cover by maintaining the “plus one” tile, ‚. On the other hand,

the closed-wait protorun ‰ȷ is weaker so we should cover it by

keeping another ‰ as a back-up. You should thus discard ‚.

70 CHAPTER 3. RIICHI MAHJONG BASICS

3.3.2 Four-tile complex shapes

Among several diﬀerent kinds of four-tile complex shapes, we

will focus on those that are made up of one group and one ﬂoating

tile. There are three variants of this kind — stretched single, bulging

ﬂoat, and skipping.

Stretched single (nobetan) shape

A set of four consecutive tiles such as ˛‚‹› is called a stretched

single (nobetan) shape. Stretched single shapes are very useful both

when a hand is ready and when a hand is 1-away or worse.

When a stretched single shape is in a ready hand, that part forms

the wait of the hand. For example, the following hand is ready, wait-

ing for – ‰.

Ready hand with a stretched single shape

–—‰˛‚‹ˆ˜¨%%%

In a ready hand, the stretched single shape can be thought of as a

candidate for the head (– or ‰) and a candidate for a run (—‰

or –—). For example, if we win this hand on –, then – becomes

the head, and —‰ becomes a run. On the other hand, if we win

this hand on ‰, then ‰ becomes the head, and –— becomes a

run.

Another important role that a stretched single shape can play is

to work as a candidate for two runs. When a hand is 1-away or worse,

we can count on a stretched single shape to produce two runs. For

example, consider a stretched single shape ‚‹›“. If we draw ‹,

we will have a side-wait protorun ‚‹ in addition to a complete run

3.3. COMPLEX SHAPES 71

‹›“. Similarly, if we draw ›, we will have a side-wait protorun

›“ in addition to a complete run ‚‹›. Moreover, if we draw ˛

”

, we will have a 3-way side-wait shape

˛‚‹›“

(waiting for

¸-‹-”) or ‚‹›“” (waiting for ˛-›-„).

There are six patterns of stretched single shapes, from 1234 through

6789. Table 3.3 summarizes the tiles each shape can accept to produce

various waits.

Table 3.3: Types of stretched single shapes

Shape 3-way 2-way 1-way Pair Acceptance

–—‰  ı — ȷ – ‰ 6 kinds–20 tiles

—‰ı ȷ –  ‰ ﬀ — ı 7 kinds–24 tiles

‰ıȷ — ﬀ ‰ ı – ﬁ  ȷ 8 kinds–28 tiles

‰ıȷﬀ  ﬁ ı ȷ — ﬂ ‰ ﬀ 8 kinds–28 tiles

ıȷﬀﬁ ‰ ȷ ﬀ ﬂ  ı ﬁ 7 kinds–24 tiles

ȷﬀﬁﬂ ı ﬀ ‰ ﬁ ȷ ﬂ 6 kinds–20 tiles

As we can see, the middle two ones — 3456 and 4567 — are the

most versatile. They can accept two diﬀerent tiles to produce a 3-way

wait (27 or 38), two diﬀerent tiles to produce a 2-way side wait (45 or

56), and two diﬀerent tiles to produce a 1-way wait (18 or 29 to pro-

duce a closed wait). The 3456 and 4567 shapes are the most valuable

of all four-tile shapes, and we should not lightly break such shapes

when a hand is far away from ready. With this in mind, consider the

following 2-away hand.

72 CHAPTER 3. RIICHI MAHJONG BASICS

2-away hand with a stretched single shape

‰ıȷ¸¸‚›´ˆ˚˚ˇ¯

What would you discard?

It is true that discarding  or ȷ would lead to the greatest tile ac-

ceptance (7 kinds–24 tiles) temporarily. However, doing so is too my-

opic. If we do that, all the remaining protoruns will be closed-wait or

edge-wait ones. We should rather discard ¯ to keep the 3456 shape,

which we can expect to produce two side-wait protoruns later. The

resulting tile acceptance (6 kinds–20 tiles) is not much smaller, either.

Four-tile complex shapes 1: nobetan

Try to keep a stretched single shape if a hand has one. In partic-

ular, 3456 and 4567 should be kept until the hand becomes ready

or 1-away from ready.

Bulging ﬂoat (nakabukure) shape

When we have a ﬂoating tile in the middle of a run (e.g., ‰‰ı),

we have a bulging ﬂoat (nakabukure) shape. Bulging ﬂoat shapes

are quite good at producing side-wait protoruns. Any bulging ﬂoat

shapes from 2334 t hrough 6778 can accept four kinds of tiles to pro-

duce a side-wait protorun and a complete run. Take ‰‰ı, for

example. It can produce a side-wait protorun and a complete run if

we draw any of —ıȷ. With this in mind, consider the following

2-away hand.

3.3. COMPLEX SHAPES 73

Hand with a bulging ﬂoat shape

‰‰ı‹“„„´˜˝˝˚˘

What would you discard?

Discarding ‰ to break the bulging ﬂoat shape is not ideal. Although

doing so increases tile acceptance temporarily, the hand will be ﬁlled

with closed-wait protoruns. Alternatively, you should discard ˘ to

maintain the bulging ﬂoat shape.

That being said, when this shape remains as is when a hand is

ready, it does not make for a good wait. For example, consider the

following ready hand.

Ready hand with a bulging ﬂoat shape

‰‰ı˛‚ˆ˜¨"%%% ‹

Draw

What would you discard?

Discarding " to keep the bulging ﬂoat shape ‰‰ı makes the

wait of this hand pretty bad. It is waiting for ‰, but we are already

using two of it in t he hand, leaving only two winning tiles. We should

rather discard ‰ to wait for ".

Four-tile complex shapes 2: nakabukure

Try to keep a bulging ﬂoat shape until a hand becomes 1-away.

74 CHAPTER 3. RIICHI MAHJONG BASICS

Skipping shape

When we have a ﬂoating tile two tiles away from a run, we have

a skipping shape. For example, in a shape ıȷﬀ,  is ﬂoating

next next to a run ıȷﬀ.  in a skipping shape is more valuable

than isolated , because it increases t he kinds of tiles the hand can

accept to produce a protorun or a 3-way side-wait shape. Table 3.4

summarizes all the skipping shapes and the tiles each shape can ac-

cept.

Table 3.4: Types of skipping shapes

Shape 3-way 2-way 1-way Pair Acceptance

–‰ı —  ȷ – 4 kinds–14 tiles

—‰ıȷ  – ‰ ﬀ — 5 kinds–18 tiles

ıȷﬀ ‰ — – ı ﬁ  6 kinds–22 tiles

‰ȷﬀﬁ ı  — ȷ ﬂ ‰ 6 kinds–22 tiles

ıﬀﬁﬂ ‰ ȷ  ﬀ ı 5 kinds–18 tiles

–—ı ‰ ȷ  ﬀ ı 5 kinds–18 tiles

—‰ȷ ı ﬀ – ﬁ ȷ 6 kinds–22 tiles

‰ıﬀ ȷ ﬁ — ı ﬂ ﬀ 6 kinds–22 tiles

‰ıȷﬁ ﬀ  ȷ ﬂ ﬁ 5 kinds–18 tiles

ıȷﬀﬂ ﬁ ‰ ﬀ ﬂ 4 kinds–14 tiles

Bearing in mind that  of ıȷﬀ is more valuable than iso-

lated , consider the following hand.

3.3. COMPLEX SHAPES 75

Hand with a skipping shape

ﬀﬁ›“”ˆ¨˚ˇ%%% ‚

Draw

What would you discard?

We should keep ‚ and discard  instead. This is because ‚ is a

part of a skipping shape ‚›“”, but  is an isolated ﬂoating tile.

As we can see in Table 3.4, skipping shapes with a terminal tile

(1345 and 5679) are also valuable. The 1 of 1345 and the 9 of 5679 can

accept more tiles than an isolated 2 or 8 (let alone than an isolated 1

or 9).

76 CHAPTER 3. RIICHI MAHJONG BASICS

3.4 Waits

There are ﬁve basic wait patterns, as summarized in Table 3.5.

More complicated wait patter ns can emerge when some of these ﬁve

basic patterns are combined.

Table 3.5: Five basic waits

Name Japanese Example Wait Acceptance

side wait ryanmen ‰ —-ı 2 kinds–8 tiles

dual pon wait shanpon ˜˜›› ˜ › 2 kinds–4 tiles

closed wait kanchan ˚˘ ˇ 1 kind–4 tiles

edge wait penchan ¸˛ ‚ 1 kind–4 tiles

single wait tanki — — 1 kind–3 tiles

As we can see in the table, side wait is the strongest of all the basic

waits in terms of the kinds and the number of tiles to win on. Single

wait appears to be much worse than others, but single-wait hands

tend to have many possibilities of improving the wait and/or scores

further. Moreover, single wait of an honor tile has a relatively high

chance of winning it by ron.

Stretched single wait and semi side wait

Table 3.6 summarizes two wait patterns, each of which can be

thought of as a combination of some basic wait patterns. As I men-

tioned before, a stretched single shape in a ready hand forms a 2-way

single wait. It is a decent wait pattern, as the number of tiles to win

on (2 kinds–6 tiles) is twice as big compared with a regular single

wait.

3.4. WAITS 77

Table 3.6: Stretched single wait and semi side wait

Name Example Waits Acceptance

stretched single wait —‰ı — ı 2 kinds–6 tiles

semi side wait ˜˜¨˝ ˜-˚ 2 kinds–6 tiles

However, stretched single wait should not be confused with side

wait for a few reasons. First, the number of tiles a 2-way stretched-

single-wait hand can win on is at most 6, whereas it is 8 for a 2-way

side-wait hand. The diﬀerence between 6 and 8 is non-trivial. Sec-

ond, stretched single wait is still a variant of single wait, which means

two t hings. On the one hand, we cannot claim pinfu when the wait is

stretched-single wait. For example, the following hand has no yaku

and hence we cannot win it by ron without calling riichi.

‰ıȷ¸˛‚‹´ˆ˜¨˝˚

On the other hand, we get 2 minipoints (fu) with a stretched single

wait. For example, if we win t he following hand by drawing ¸, we

get 40 minipoints (20 base minipoints + 8 for a concealed set of honor

tiles + 2 for self-draw + 2 for single wait = 32, rounded up to 40).

‰ıȷ¸˛‚‹´ˆ˜%%%

When we have a side-wait protorun right next to a pair (e.g., 1123,

2234, 7899, etc.), we call it semi side wait. We distinguish this from

regular side wait for two reasons. First, the number of tiles to win

on is smaller (6 rather than 8) because we are already using 2 of the

We will discuss methods of scoring and minipoints calculations extensively in

Chapter 6.

78 CHAPTER 3. RIICHI MAHJONG BASICS

8 winning tiles in our hand. Second, we can treat this wait pattern

either as single wait or as side wait, depending on which interpre-

tation gives us a greater score. For example, consider the following

hand.

‰ıȷ¸¸˛‚´ˆ˜¨˝˚

We will treat the wait in this hand as side wait because that will give

us pinfu. However, consider the following hand that has the exact

same wait pattern: ¸¸˛‚.

‰ıȷ¸¸˛‚´ˆ˜%%%

If we win this hand by drawing ¸, we will treat the wait as single

wait: ¸ + ¸˛‚, which will give us 40 minipoints. If we treated

the wait as side wait: ¸¸ + ˛‚, we would get only 30 minipoints.

Of course, if we win this hand on ‹, we cannot think of the wait as

side wait (because it is not). Similarly, if we win it by ron, it does not

make a diﬀerence if it is side wait or single wait (either way we get

40 minipoints).

3.4. WAITS 79

3-way side wait

When a side-wait protorun is combined with an adjacent run, we

get a regular 3-way side-wait pattern. There are only three of this

kind, summarized in Table 3.7.

Table 3.7: Regular 3-way side wait

Example Wait Acceptance

—‰ıȷ –-‰-ﬀ 3 kinds–11 tiles

‚‹›“” ˛-›-„ 3 kinds–11 tiles

¨˝˚ˇ˘ ˜-˚-¯ 3 kinds–11 tiles

When we have a stretched single shape or semi side-wait shape

combined with an adjacent run, we also get a 3-way wait pattern.

Table 3.8 summarizes some examples.

Table 3.8: Some irregular 3-way waits

Example Wait Acceptance

–—‰ıȷﬀ – ‰ ﬀ 3 kinds–9 tiles

˛‚‹››“” ˛ ›-„ 3 kinds–9 tiles

¨˝˚ˇ˘¯¯ ˜-˚-¯ 3 kinds–9 tiles

Notice that the number of tiles to win on in each pattern is smaller

than those for the regular 3-way side waits, although the kinds of

tiles to win on are t he same (either 1-4-7, 2-5-8, or 3-6-9). This is

because we are already using some of the winning tiles within the

hand.

80 CHAPTER 3. RIICHI MAHJONG BASICS

Notice also that not all the wait patterns qualify as side wait, so

claiming pinfu is not always possible (similarly, claiming single wait

is not always possible). For example, the ﬁrst pattern in Table 3.8 is

essentially a 3-way stretched single shape; none of the waits embed-

ded in this shape qualiﬁes as side wait. In the second pattern, if we

win on ˛, the wait must be interpreted as single wait; if we win on

„, the wait must be interpreted as side wait; and if we win on ›,

we adopt whichever interpretation that generates the higher score.

In the third pattern, winning on ¯ allows us to claim single wait if

doing so gives us a higher score.

Complex waits

When a set is combined with a ﬂoating tile nearby, we get some

complex wait patterns with multiple waits. Table 3.9 summarizes a

few examples of irregular waits that involve a set and a ﬂoating tile.

Table 3.9: Some irregular waits (set and a ﬂoating tile)

Example Combination Wait Acceptance

–——— single and edge –  2 kinds–7 tiles

¸‚‚‚ single and closed ¸ ˛ 2 kinds–7 tiles

ˆ˜˜˜ single and side ˆ ´-¨ 3 kinds–11 tiles

When a set is combined with a protorun, pair, or a four-tile shape,

we get even more complicated waits. Table 3.10 summarizes only a

few representative examples.

3.4. WAITS 81

Table 3.10: Some irregular waits (set and a protorun, pair, or a four-tile shape)

Example Wait Acceptance

––——— – —  3 kinds–5 tiles

¸¸˛˛‚‚‚ ¸ ˛ ‚ 3 kinds–5 tiles

´´ˆˆ˜˜¨¨"" ´ ¨ " 3 kinds–6 tiles

ıııȷﬀ ı-ﬁ 3 kinds–7 tiles

“”„„««« ›-„ ” 3 kinds–9 tiles

ˆˆˆ˜¨¨˝ ˜-˚ ¨ 3 kinds–9 tiles

ıȷﬀﬁ ‰ ı ﬁ 3 kinds–10 tiles

¸¸¸‚››› ˛ ‚ ‹ 3 kinds–11 tiles

82 CHAPTER 3. RIICHI MAHJONG BASICS

3.5 Glossary

Simple tiles (tanyao hai) are tiles between 2 and 8.

Terminal tiles are 1 and 9.

Honor tiles (jihai) are non-number tiles (dragon tiles and wind tiles).

Value tiles (fanpai / yakuhai) include dragon tiles, seat wind tiles,

and prevailing wind tiles. We get one han for a set of value tiles.

Valueless wind tiles (otakaze hai) are wind tiles that are neither a

prevailing wind tile nor a seat wind tile.

Run (chow / sequence; shuntsu) is a set of three consecutive num-

ber tiles.

Set (pung / triplet; kotsu) is a set of three identical tiles.

Quad (kong; kantsu) is a set of four identical tiles.

Protorun (taatsu) is a set of two tiles in the same suit that can be-

come a run when one more tile is added.

Pair (toitsu) is a set of two identical tiles.

Ready (tenpai) is when a hand is ready to win.

1-away (1-shanten) is when a hand can be ready with one more tile.

Perfect 1-away is when a 1-away hand has two side-wait protoruns

and two pairs.

Tile acceptance (ukeire) refers to the kinds and the number of tiles

a hand can accept.

Stretched single (nobetan) shape is a set of four consecutive num-

ber tiles.

3.5. GLOSSARY 83

Bulging ﬂoat (nakabukure) shape is a four-tile shape that is made

up with a run and one ﬂoating tile in the middle of the run.

Skipping shape is a four-tile shape made up with a run and one

ﬂoating tile located at two tiles away from the run.

Chapter 4

The ﬁve-block method

In introducing basic building blocks of riichi mahjong in the pre-

vious chapter, I have also touched upon a number of important tile

eﬃciency principles — e.g., superiority of side-wait protoruns, the

value of having two pairs in a hand rather than three, and the value

of stretched single or bulging ﬂoat shapes, to name a few.

These principles are all important, but trying to take all of the

important principles into consideration at once could be a daunting

task. We have to make our discard choice in a limited amount of

time,

and tile eﬃciency is not the only factor we need to consider in

making a discard choice. Moreover, some of the tile eﬃciency prin-

ciples can at times clash with one another, requiring us to make a

judgement call about which principle to follow. For example, we may

at times wonder whether to retain a bulging ﬂoat shape or to retain

two pairs in a hand, when we have to discard one of the two.

The ﬁve-block method I introduce in this chapter will help us

prioritize between competing principles and ﬁnd the most eﬃcient

discard choice quickly.

The core idea of t he ﬁve-block method is

deceptively simple; we ﬁrst identify ﬁve tile blocks in a hand — four

groups + one head, or their candidates — and try to complete each

block.

Recall that, on regular (slower) tables on Tenhou, each discard choice must be

made within 5 seconds. In oﬄine games, we should make choices even faster

so as not to irritate your fellow players.

As I mentioned in the Preface, the exposition of this chapter is based on Makoto

Fukuchi’s books. In particular, I am indebted to Makoto Fukuchi 2015 Haik-

ouritsu Nyumon Doriru 76, Yousensha. ISBN978-4-8003-0634-0.

4.1. FINDING A REDUNDANT TILE 85

4.1 Finding a redundant tile

We all understand that a standard hand must have ﬁve blocks of

tiles — four groups and one head — to win. The ﬁve-block method

encourages us to be always conscious about ﬁve tile blocks in a hand.

Consider the following hand. What would you discard and why?

ııﬀﬁ˛‹››´˜˜%%%

To ﬁgure out which tile is the least useful in this hand, let’s divide

the hand into tile blocks, as follows.

ıı

|{z}

ﬀﬁ

|{z}

˛‹››

| {z }

´˜˜

|{z}

%%%

|{z}

(4.1)

Notice that, although we do not know which block is going to be the

head and which blocks are going to be four groups at the moment,

the hand already has ﬁve tile blocks. This means that there is no need

to increase or decrease the number of blocks from here.

Looking at each of the ﬁve blocks, the pair of ı, the protorun

ﬀﬁ, and the set of % are all self-suﬃcient; we keep them as they

are. Our discard choice should thus be from the third or the fourth

blocks, ˛‹›› or ´˜˜. Let’s now compare these two closed-

wait blocks. While ´ is being useful within the block it belongs to,

enabling the hand to accept ˆ, ˛is completely redundant; the hand

can accept ‚ without having ˛. Therefore, the ideal discard here

is ˛.

There are two key points to remember in applying t he ﬁve-block

method. First, we should not make any one of the ﬁve blocks “too

86 CHAPTER 4. THE FIVE-BLOCK METHOD

weak.”

In the current example, if we discard ˜, the ´˜˜ block

becomes an isolated closed-wait protorun, which is too weak com-

pared with the other blocks. Likewise, if we discard

, this block

becomes a pair of ˜. Since this hand already has two other pairs,

having a third pair makes all the pairs in the hand too weak.

Second, each of the ﬁve tile blocks should ideally have three tiles.

In t he current example, the ´˜˜ block has exactly three tiles and

so we should not choose a discard from this block. On the other

hand, the ˛‹›› block currently has four tiles so we should dis-

card one from this block to make this a three-tile block.

Five-block method

Identify ﬁve tile blocks in a hand. Try to make sure:

(1) there is no block that is too weak; and

(2) each block has at most three tiles.

Basically, any block that is weaker than a side-wait protorun is a weak block.

4.1. FINDING A REDUNDANT TILE 87

Let’s see another example.

ıﬀ‹›““”¨˚˚˘## #

Dora

We can easily see that there is one block in manzu (cracks), two blocks

in pinzu (dots), and a pair of white dragons, giving us four blocks.

This means that we need to have only one more block in souzu (bam-

boos). Therefore, we divide the hand as follows.

ıﬀ

|{z}

‹›“

|{z}

“”

|{z}

¨˚˚˘

| {z }

|{z}

(4.2)

Since we should not create a block that is too weak, discarding  or

ﬀis not an option. Notice that the block in souzu (bamboos) has four

tiles. We should thus discard one from this block. In case the pair

of white dragon later becomes a set, we should keep the pair of ˚,

leaving ¨ or ˘ as a discard candidate. Given that ¨ has a higher

chance of creating a side-wait protorun, we should discard ˘. Then,

none of the ﬁve blocks is too weak, and each block has at most three

tiles.

In the two examples we saw above, you might have been able

to identify the redundant tiles without really thinking too hard. If

so, that was probably because you have implicitly and unconsciously

applied t he ﬁve-block method in your mind. The goal of this chapter

is to train our mind further, so that it becomes our second nature to

identify ﬁve tile blocks in a hand.

88 CHAPTER 4. THE FIVE-BLOCK METHOD

4.2 Alternative conﬁgurations

Consider the following hand. What would you discard and why?

‰ȷ˛˛‹›““”%%%

Let’s ﬁrst divide the hand into ﬁve tile blocks.

‰ȷ

| {z }

˛˛

|{z}

‹›“

|{z}

“”

|{z}

%%%

|{z}

(4.3)

This makes it clearer that, just like the previous example, ȷ is

creating a redundant closed-wait protorun, so we should discard it.

Also, discarding ȷ makes this a three-tile block.

However, there is an alternative way to divide this hand into ﬁve

blocks, and situational changes may call for such an alternative con-

ﬁguration. Suppose that your opponents have already discarded all

four tiles of —. Suppose also that ‚ seems live in the wall. Or, sup-

pose -ȷ tiles seem too dangerous to discard against an opponent.

Then, we might want to divide the hand in the following way instead.



|{z}

‰ȷ

|{z}

˛˛‹›““”

| {z }

%%%

|{z}

(4.4)

That is, we aim to make the pair of  the head of this hand, and

we seek to have two runs in pinzu (dots). If we discard ˛, this block

becomes ˛‹›““”. Recall that a block like this can be split into

4.2. ALTERNATIVE CONFIGURATIONS 89

˛‹“ + ›“” (recall the discussion of double closed shape in

Section 3.3.1). Therefore, this block can accept ‚ as well as ›-„ to

make two runs in

pinzu

(dots). The block in

pinzu

(dots) will have

six tiles, but this is OK because this block is worth two.

To master the ﬁve-block method, we need to be able to instanta-

neously envision the ﬁrst block conﬁguration (4.3) the moment we

see this hand. However, that is not enough. We should also be able

to imagine an alternative conﬁguration (4.4) at the same time. In the

game of mahjong, situations change very quickly each time a new

tile gets drawn or a new tile gets discarded. Therefore, the ideal

ﬁve-block conﬁguration would also change accordingly as situations

evolve. We thus need to develop our skills to picture many possi-

ble ﬁve-block conﬁgurations and to prepare for possible situational

changes t hat would call for a change in the conﬁguration.

I provide several exercises in the following pages. The answer

key to each exercise is provided on the next page. Try not to look at

the answers before you actually derive your own answer.

Exercises: ﬁnding a redundant tile

Exercise 1

What would you discard?

How do you divide the hand into ﬁve tile blocks?

––‰‚‹›˜˜˝˚ˇˇ ¨

Draw

90 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 1

––

|{z}

‰

|{z}

‚‹›

|{z}

˜˜¨˝˚ˇˇ

| {z }

With the draw of ¨, we now have a 3-way side-wait block in

souzu (bamboos). ˜ or ˇ could be our back-up candidate for

the head, in case we draw another –. Since there is sanshoku

(Mixed Triple Chow) of 345, we discard ˜.

Exercise 2

What would you discard?

How do you divide the hand into ﬁve tile blocks?

ıﬀ¸¸˛‚‹‹›”˜˝˝ ‹

Draw

4.2. ALTERNATIVE CONFIGURATIONS 91

Answer 2

ıﬀ

|{z}

¸ ¸˛‚

|{z}

‹‹‹

|{z}

›”

|{z}

˜˝˝

|{z}

Before we drew the third ‹, the pinzu (dots) tiles were ¸¸ +

˛‚‹ + ‹›, so the ” was simply a ﬂoating tile. Now that

we have another ‹, the ﬁve-block conﬁguration changes accord-

ingly. The ideal discard is ¸, as this has become redundant.

Exercise 3

What would you discard?

How do you divide the hand into ﬁve tile blocks?

‰‰ȷȷﬀﬁ˛‚˜¨˚˚ˇ 

Draw

92 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 3

‰‰

|{z}

ȷ ȷﬀﬁ

|{z}

˛‚

|{z}

˜¨

|{z}

˚˚ˇ

|{z}

There are two “side wait plus one” shapes, ‰‰ and ˚˚ˇ,

that might later become the head or a run. At this point, how-

ever, we cannot determine which one will be which, so we should

keep them as they are. One of the two ȷ has become an obvious

redundancy so we should discard one.

Exercise 4

What would you discard?

How do you divide the hand into ﬁve tile blocks?

—ıȷ‹‹“„«˜¨˝" ‰

Draw

4.2. ALTERNATIVE CONFIGURATIONS 93

Answer 4

—‰ıȷ

| {z }

‹‹

|{z}

“„

|{z}

« ˜¨˝

|{z}

The " is obviously redundant, but « is also useless. Without

«, the hand can accept ”. Since honor tiles can be used as a

safety tile (see Chapter 8), we discard « ﬁrst.

Exercise 5

What would you discard?

How do you divide the hand into ﬁve tile blocks?

—ıﬀﬀ˛‹“„¨˝ˇ˘ ˘ ‹

Draw Dora

94 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 5

—

|{z}

ıﬀﬀ

|{z}

˛‹“„

| {z }

¨˝

|{z}

ˇ˘˘

|{z}

This is a bit diﬃcult, as there are so many closed-wait protoruns.

Recall that each tile block should have at most three tiles and that

pairs are most valuable when there are two of them in a hand.

The block in pinzu (dots) has four tiles, so we discard one from

this block. Since ‹ is dora, we discard „, leaving the double

closed shape around dora: ˛‹“.

Exercise 6

What would you discard?

How do you divide the hand into ﬁve tile blocks?

—‰‰‚‹›“¨˝˝˚˚ ‹

Draw

4.2. ALTERNATIVE CONFIGURATIONS 95

Answer 6

—‰

|{z}

‰ ‚‹‹›“

| {z }

¨˝˝˚˚

| {z }

Finding the best discard by actually comparing tile acceptance

counts for each possible discard candidate is super tedious. The

ﬁve-block met hod simpliﬁes the process quite a bit. Since we

have two blocks in pinzu (dots) and two blocks in souzu (bam-

boos), we only need one block in manzu (cracks), hence one ‰ is

redundant. If we discard ‰, the hand can be made ready with

11 kinds–34 tiles. If we discard ‚ ‹ or ˝, the hand can be

ready only with 6 kinds–19 tiles.

Exercise 7

What would you discard?

‰ﬀ˛‚‹””„¨˝ˇ¯ ˚

Draw

96 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 7

‰

|{z}

ﬀ ˛‚‹

|{z}

””„

|{z}

¨˝˚ˇ¯

| {z }

Do not discard the ¯ just because it forms a closed wait or be-

cause discarding it gets us tanyao (All Simples). Avoiding closed

wait too much and being hung up on tanyao are two pathologies

common among intermediate players.

The block in souzu (bamboos) is actually not too bad; this is a

stretched single plus one, which can become either two runs im-

mediately (if we draw ˘), one run plus one side-wait protorun

(if we draw any of ˜˝˚), or one run plus the head (if we draw

¨ or ˇ). Note also that we need both ” and „, because t his

part may become the head if we get two runs in souzu (bamboos);

when we get t he head in souzu (bamboos), we will treat this part

as a side-wait protorun. We thus discard ﬀ.

4.3. SELECTING TILE BLOCKS 97

4.3 Selecting tile blocks

All the hands we have seen so far in this chapter already have

ﬁve tile blocks. In practice, however, this is not always the case. A

hand can sometimes have fewer or more tile blocks. Since we need

to have exactly ﬁve blocks to win a hand, we will need to bump up

tile block s by using a ﬂoating tile when we have fewer of them or to

discard some blocks entirely when we have a plethora of them.

In selecting which tile blocks to keep and which ones to discard,

we focus on a combination of the following three criteria:

1. tile eﬃciency;

2. hand value;

3. the safety of tiles to be discarded.

As we will see below, we can sometimes ﬁnd a block to discard based

on all the three criteria. Consider the following hand. How do we

divide the hand into tile blocks, and what would you discard?

Red

–—‚‹››”„¨`ˇ˘ ˛

Draw

We can see that the hand currently has six tile blocks, as follows.

–—

|{z}

˛‚‹

|{z}

››

|{z}

”„

|{z}

¨`

|{z}

ˇ˘

|{z}

Since the ﬁrst two tile blocks are already complete and the third block

is the head, our discard choice should be from the last three tile

blocks, ”„, ¨`, or ˇ˘.

98 CHAPTER 4. THE FIVE-BLOCK METHOD

From a perspective of tile eﬃciency, discarding the ”„ block

means that we lose the ability to accept both “ and «. On the other

hand, if we discard the

ˇ˘

block, we only lose the ability to accept

¯; because of the ¨` block, we can still accept ˚. We should

thus choose between the two blocks in souzu (bamboos). Keeping

the ` is desirable from a perspective of hand value (it is a red ﬁve) as

well as safety (discarding ˇ˘ is much safer than discarding ¨`,

generally speaking). Therefore, the three criteria collectively suggest

that we should discard ˇ˘.

In practice, however, satisfying all three criteria may not be fea-

sible. A common tradeoﬀ we face is between speed and hand value.

That is, maximizing tile eﬃciency to gain speed often entails giving

up possibilities of pursuing an expensive hand. Consider t he follow-

ing hand.

–ıﬀﬀ‚”„˜¨˘˘¯¯

Let’s divide the hand into tile blocks. There are several ways to do

this. One way to do this is to split it into the following blocks.

–ı

|{z}

ﬀﬀ

|{z}

‚ ”„

|{z}

˜¨

|{z}

˘˘

|{z}

¯¯

|{z}

If we simply maximize tile eﬃciency, we discard ‚, as we already

have six tile blocks and we won’t need any more ﬂoating tile.

However, as it stands, the hand has no yaku and it is likely to be

a very cheap riichi-only hand. Moreover, the hand has three pairs,

which is not ideal as we saw in the previous chapter. Therefore, we

4.3. SELECTING TILE BLOCKS 99

might want to split the hand into the following ﬁve blocks.

–ıﬀﬀ

| {z }

‚

|{z}

”„

|{z}

˜¨

|{z}

˘˘

|{z}

¯¯

We count the ﬂoating ‚ as an independent block, hoping it to grow

into a run. We also treat the tiles in manzu (cracks) as a single block,

hoping to get at least one group or the head out of it. We thus discard

one ¯ now, then another ¯ in the next turn. Depending on what

tile gets drawn, our ﬁve-block conﬁguration will be diﬀerent.

For example, suppose we draw ‹ and then ˝. We will then have

the following.

–ıﬀﬀ

| {z }

‚‹

|{z}

”„

|{z}

˜¨˝

|{z}

˘˘

|{z}

We will discard the – as a ﬁrst step toward reducing the number of

tiles in the manzu (cracks) block to three. We can now see that this

hand has a potential of getting sanshoku of 345 as well as pinfu and

tanyao.

On the other hand, if we draw ȷ and then ﬁ, we can expect to

have two groups in manzu (cracks) so we will discard ‚.

–ıȷﬀﬀﬁ

| {z }

‚ ”„

|{z}

˜¨

|{z}

˘˘

|{z}

In selecting tile blocks, we should try to achieve the best balance

between speed and hand value. Don’t fantasize too much about get-

ting an expensive hand. At the same time, don’t ﬁxate too much

100 CHAPTER 4. THE FIVE-BLOCK METHOD

about tile eﬃciency at the cost of hand value. This is of course easier

said than done; it is quite diﬃcult even for advanced players.

Exercises: selecting tile blocks

Exercise 8

What would you discard?

How do you divide the hand into ﬁve tile blocks?

—ﬀﬁ›“´´ˆ¨¯¯ ¯

Draw

4.3. SELECTING TILE BLOCKS 101

Answer 8

—

|{z}

ﬀﬁ

|{z}

›“

|{z}

´´

|{z}

ˆ¨

|{z}

¯¯¯

|{z}

The hand currently has six blocks so we need to get rid of one.

The ˆ¨ block is the weakest – it is the only closed-wait pro-

torun – so we should get rid of this one. We should discard ¨

ﬁrst; if we draw ˜ we will discard ´ to leave the possibility of

pinfu. If not, we discard ˆ next, and then .

Exercise 9

What would you discard?

How do you divide the hand into ﬁve tile blocks?

—‰”„«««ˆˆ˚ˇ## ‰

Draw

102 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 9

—‰‰

|{z}

”„«««

| {z }

ˆˆ

|{z}

˚ˇ

|{z}

We were planning to discard the —‰ block because this was the

weakest block among the six blocks in this hand. However, now

that we drew another ‰, the ˆˆ block is now the weakest. We

thus discard ˆ.

Exercise 10

What would you discard?

How do you divide the hand into ﬁve tile blocks?

–ıȷﬀ¸‚˜¨ˇ˘"" ˝

Draw

4.3. SELECTING TILE BLOCKS 103

Answer 10

–

|{z}

ıȷﬀ

|{z}

¸‚

|{z}

˜¨˝

|{z}

ˇ˘

|{z}

The hand currently has six blocks so we need to get rid of one.

Comparing the two closed-wait blocks – and ¸‚, the –

block is more valuable because it is adjacent to a run. If we draw

‰, we will get a 3-way side-wait block. On the other hand, the

¸‚ block will only become a 2-way side-wait block when we

draw ‹. We should discard ¸ ﬁrst, not ‚, because if we draw

‹ next, we will discard the – block.

Exercise 11

What would you discard?

How do you divide the hand into ﬁve tile blocks?

ıﬀ¸¸‚””´´¨˝˚ˇ ˜

Draw

104 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 11

ıﬀ

|{z}

¸¸‚

|{z}

””

|{z}

´´

|{z}

˜¨˝˚ˇ

| {z }

Now that we have a 3-way side-wait block in souzu (bamboos),

we should get rid of one block. Comparing a closed-wait block

ıﬀ and two pairs ”” and ´´, we should value the closed-

wait block. This is because the hand has three pairs already so we

should get rid of one of them. Since we see a (remote) possibility

of sanshoku of 567, we should discard ´.

4.4 Building a block

When a hand has fewer than ﬁve blocks, we need to build a new

block, possibly from a ﬂoating tile we already have in the hand. In

doing so, we should envision the kind of yaku that the hand is going

to have eventually. Consider the following hand. Suppose you are

the dealer and t his is East-1. What would you discard?

ȷﬀﬀﬁ‚‚‹«¨ˇ˘¯%"

As usual, we will split the hand into blocks. Notice that the hand

has at most four blocks only.

ȷﬀﬀﬁ

| {z }

‚‚‹

|{z}

« ¨ ˇ˘¯

|{z}

% "

We should thus compare the four ﬂoating tiles « ¨ % " in terms

4.4. BUILDING A BLOCK 105

of their relative capabilities to grow into an independent block. Of

these four tiles, ¨ is the strongest candidate, because it can form a

side-wait protorun with two kinds of tiles, ˜ and ˝. Any simple tiles

between 3 and 7 are a strong ﬂoating tile because of their ability to

form a side-wait protorun. Terminals (1 and 9) will never become

a side-wait protorun, and 2 and 8 can become a side-wait protorun

when paired with only one kind of tiles (3 or 7). However, number

tiles are still stronger than honor tiles because honor tiles can never

form a run.

We should thus choose between the two honor tiles, % and ".

Which one should we discard? Notice that this hand is clearly a

pinfu hand and that it is currently lacking the head. Since value tiles

can never be the head of a pinfu hand, we should discard % rather

than ".

We may want to choose a discard from an existing block rather

than discarding a ﬂoating tile in order to enhance the hand value.

Consider the following hand.

ıȷȷﬁ¸˛˛“´´¨˝˚ˇ

From a pure perspective of tile eﬃciency, the discard choice should

be either ¨ ˇ or “, for discarding either of the three will maximize

tile acceptance. The block conﬁguration behind such a decision is as

follows.

ıȷ

|{z}

ȷﬁ

|{z}

¸˛˛

|{z}

“ ´´

|{z}

¨˝˚ˇ

| {z }

However, doing so makes it almost inevitable that the hand ends up

having a low score and/or a bad wait. Alternatively, we can expect

106 CHAPTER 4. THE FIVE-BLOCK METHOD

the stretched single shape ¨˝˚ˇ to produce two runs, “ to form

a run, and the tiles in manzu (cracks) to produce one run, as follows.

ıȷȷﬁ

| {z }

¸˛˛

|{z}

“

|{z}

´´ ¨˝˚ˇ

| {z }

We should thus discard the ¸ for now, anticipating to discard the

pair of ´ eventually. That way, we can expect to have tanyao, pinfu,

and possibly sanshoku.

Exercises: building a block

Exercise 12

What would you discard?

How do you divide the hand into ﬁve tile blocks?

Red

–»ﬁﬂ‚‹‹ˆˆ˜˝˚ˇ

4.4. BUILDING A BLOCK 107

Answer 12

–»

|{z}

ﬁﬂ

|{z}

‚‹‹

|{z}

ˆˆ˜

|{z}

˝˚ˇ

|{z}

If we were to simply maximize tile acceptance, the discard choice

should be either – or ». However, that would make the block

in manzu (cracks) too week. Breaking the ‚‹‹ or ˆˆ˜ is

not ideal, as these blocks are very strong. We should therefore

discard the ﬂ to get rid of this edge-wait block. This will tem-

porarily reduce the number of blocks from ﬁve to four, but we

can expect to get back to ﬁve soon with this hand.

Exercise 13

What would you discard?

How do you divide the hand into ﬁve tile blocks?

Red

–»ı‚‹‹›ˆˆ˜˝˚ˇ

108 CHAPTER 4. THE FIVE-BLOCK METHOD

Answer 13

–

|{z}

»ı

|{z}

‚‹‹›

| {z }

ˆˆ˜

|{z}

˝˚ˇ

|{z}

Discarding ‹ or ˆ will make this hand 1-Away, so our choice is

between these two options. Notice that the – block is weaker

than the other four. As a back up, we should keep two ‹ to

maintain the bulging ﬂoat block in pinzu (dots) for now, hoping

to get two runs out of it. If we draw ˛-› or ‚-“ ﬁrst, we will

get rid of the – block. We should thus discard ˆ. If we draw

any of ´¨—, we should do insta-riichi.

Exercise 14

What would you discard?

How do you divide the hand into ﬁve tile blocks?

Red

––»ı‚‹‹›ˆ˜˝˚ˇ

4.4. BUILDING A BLOCK 109

Answer 14

––»ı

| {z }

‚‹‹›

| {z }

ˆ˜

|{z}

˝˚ˇ

|{z}

As we drew another –, the block in manzu (cracks) is now a de-

cent shape. This can become one group and the head with a

draw of – — ‰ ı. Therefore, we should discard ‹ to break

the bulging ﬂoat shape.

Chapter 5

Pursuing yaku

As we saw in the previous chapter, we often face a tradeoﬀ be-

tween speed (tile eﬃciency) and hand value. In modern riichi mahjong,

the value of pursuing expensive yaku is much diminished because

of red ﬁves. For example, ryanpeiko (Twice Pure Double Chow) is

a beautiful three-han yaku, but it is extremely diﬃcult to make this

yaku. We can achieve the same hand value more easily with riichi +

dora + one red ﬁve. We thus tend to think of expensive yaku as some-

thing that emerges in a hand (almost) by chance, not something we

actively pursue. Given that we can get high scores also from ippatsu,

ura dora, and tsumo, getting the hand ready for riichi is generally

more important than pursuing expensive yaku.

That being said, always trying to maximize tile eﬃciency with-

out regard for yaku is not the best strategy, either. We should thus

design a ﬁve-block conﬁguration with an eye to possible yaku we can

reasonably get. Moreover, sometimes situations call for an expensive

hand. For example, when you are ranked fourth in South-4, and the

player who is currently ranked third has 10000 more points than you

do, you should aim for mangan tsumo or haneman ron to improve the

placement (more on this in Chapter 10 ), which will require that your

hand has some yaku other than just riichi and dora.

In this chapter, I will discuss some tips to get the following ﬁve

set of yaku.

5.1 sanshoku (Mixed Triple Chow) 5.2 ittsu (Pure Straight)

5.3 pinfu (Pinfu) 5.4 honitsu (Half Flush)

5.5 toitoi (All Pungs) and chiitoitsu (Seven Pairs)

5.1. HOW TO GET SANSHOKU 111

5.1 How to get sanshoku

Sanshoku (Mixed Triple Chow) is an elusive yaku. Even when we

make our hand ready for sanshoku, we may lose it at the very last

minute. For example, suppose we manage to get the hand ready for

sanshoku of 345, and we have a side-wait protorun ‰ı as the ﬁnal

wait. We will get sanshoku only if we win the hand on ; if we win

on ȷ, we will lose sanshoku.

On t he other hand, it is possible to have a conﬁrmed sanshoku,

but doing so often entails a signiﬁcant loss in tile eﬃciency. For ex-

ample, if our wait were a closed-wait protorun ıinstead, sanshoku

is conﬁrmed; but, a closed wait of ‰ is not very good. As long as

we seek to utilize side waits to maximize tile eﬃciency, sanshoku be-

comes diﬃcult to achieve. I will discuss the following seven methods

to capture this elusive yaku.

5.1.1 Floating 5.1.2 Switching

5.1.3 Double closed shape 5.1.4 Stretched single

5.1.5 Lining pairs 5.1.6 Golden

5.1.7 Crashing a meld

112 CHAPTER 5. PURSUING YAKU

5.1.1 Floating

Sanshoku 1

ıȷ‚‹›´´˜¨ˇ˘¯"

What would you discard?

 in this hand is essentially a redundant ﬂoating tile from a pure

tile eﬃciency perspective; we do not need it to accept ‰. However,

if we keep it and discard " instead, we can hope to get sanshoku of

345. The best case scenario is to draw ‰ ﬁrst, after which we discard

ȷ.

That being said, keeping a ﬂoating tile this way is risky. What if

an opponent calls riichi after we discard "? We will have to discard

either  or ȷ against the riichi’ed player when this hand becomes

ready. To make things worse, if we draw ˆ ﬁrst, we will have to

discard the potentially dangerous  (instead of ") with no beneﬁt

of getting sanshoku.

Moreover, when we draw ˝ ﬁrst, a diﬃcult question arises. Con-

sider the following hand.

ıȷ‚‹›´´˜¨ˇ˘¯ ˝

Draw

Should we discard ȷ and have a closed-wait hand in hopes of get-

ting sanshoku, or should we discard  and give up on sanshoku in

pursuit of tile eﬃciency? If a hand has at least one dora or tanyao

(All Simples), we should discard  to choose a side wait pinfu hand.

Only when there is no other yaku or dora in a pinfu hand, it is OK

to choose closed-wait sanshoku over side-wait pinfu.

We will talk more about a tradeoﬀ of this kind in Chapter 7.

5.1. HOW TO GET SANSHOKU 113

5.1.2 Switching

Sanshoku 2

—‰ȷȷ”„„ˆ˜¨¨˝ ‚

Draw

What would you discard?

The hand already has ﬁve tile blocks and all the blocks are strong;

in fact, this is a perfect 1-away hand.

It is thus OK to discard ‚

we just drew. After all, that is the best discard from a tile eﬃciency

perspective.

However, if we need an expensive hand, we can keep ‚ and dis-

card „ instead. The resulting loss in tile eﬃciency is not very big,

as we would still have a strong 1-away hand with two side-wait pro-

toruns: ”„ and ¨˝. In addition, ‚ serves as a ﬂoating tile to

approach sanshoku of 234. Keep in mind that we should give up on

sanshoku and do insta-riichi if we draw any of “«˜˚ ﬁrst (un-

less you absolutely need mangan or above to improve the placement

in South-4).

On the other hand, if we draw ˛ or ‹ ﬁrst, t he hand will be 1-

away from ready for sanshoku. For example, with a draw of ‹, the

hand becomes the following.

—‰ȷȷ‚”„ˆ˜¨¨˝ ‹

Draw

We should discard ” and then „ to aim for sanshoku of 234. We

are switching from one protorun ”„ to another protorun ‚‹ to

approach sanshoku.

For the deﬁnition of perfect 1-away, see Section 3.2.7.

114 CHAPTER 5. PURSUING YAKU

The key here is that we are keeping the hand 1-away through-

out the entire process of switching from a perfect 1-away hand to a

sanshoku 1-away hand. You should not pursue sanshoku if switch-

ing requires reverting a 1-away hand to 2-away.

5.1.3 Double closed (ryankan) shape

Sanshoku 3

——ıȷ˛‚›“”ˆ˜˝ˇ ¸

Draw

What would you discard?

As we have two blocks in manzu (cracks) and another two blocks in

pinzu (dots), we only need one block in souzu (bamboos). Our choice

is thus between (a) keeping a side-wait protorun ˆ˜ to maximize

tile eﬃciency and (b) keeping a double closed shape ˜˝ˇin hopes

of getting sanshoku of 567.

If the hand has at least one dora or some yaku (such as tanyao),

we should give up sanshoku and discard ˝ˇ. Only if the hand has

no other yaku or dora, it is OK to discard ˆ to aim for sanshoku.

Keep in mind, though, that pursuing sanshoku with a hand like

this is risky, even compared with the ﬂoating method we discussed

in 5.1.1. We will end up with a bad-wait yaku-less hand if we draw

‰ ﬁrst. If we give up on sanshoku sooner and choose the side-wait

protorun ˆ˜, we can at least get pinfu.

5.1. HOW TO GET SANSHOKU 115

5.1.4 Stretched single (nobetan) shape

Sanshoku 4

‰ıȷﬀﬁ‚‚“”˜¨˝˚ ˇ

Draw

What would you discard?

Notice that this hand has two possibilities of sanshoku , 567 or 678,

and we do not know at this point which one we can get. An excellent

way to aim for sanshoku with a hand like this is to discard ‰ to have

a stretched single shape ıȷﬀﬁ that contains both 567 and 678.

If we draw any of ›ˆ˝ ﬁrst, we discard ﬁ to aim for sanshoku of

567.

ıȷﬀﬁ‚‚“”˜¨˝˚ˇ ˆ

Draw

If we draw „ or ˘ ﬁrst, we discard ı to aim for sanshoku of 678.

ıȷﬀﬁ‚‚“”˜¨˝˚ˇ „

Draw

Either way, you get sanshoku without any loss of tile eﬃciency.

116 CHAPTER 5. PURSUING YAKU

5.1.5 Lining pairs

Sanshoku 5

‰ı‚‚‹‹““¨˝˚ˇ¯ ˘

Draw

What would you discard?

As we have one block in manzu (cracks) and two blocks in souzu (bam-

boos), we need to have two blocks in pinzu (dots). More speciﬁcally,

we need the head and a group (preferably a run) in pinzu. We there-

fore view the tiles in pinzu not as a collection of three pairs but as

a collection of one pair ““ and two side-wait protoruns ‚‹ +

‚‹.

From a pure tile eﬃciency perspective, discarding ‚ and dis-

carding ‹ are equally good, and they are better than any other dis-

cards. However, there is a clear diﬀerence between the two from a

perspective of hand value. Suppose we discard ‚ ﬁrst. If we then

draw ›, we will get the following hand.

‰ı‚‹‹““¨˝˚ˇ˘¯ ›

Draw

Discarding ‹ makes this hand ready, but it is just a pinfu-only hand.

On the other hand, suppose we had discarded ‹ before drawing ›.

‰ı‚‚‹““¨˝˚ˇ˘¯ ›

Draw

We can make this hand ready for pinfu and sanshoku of 456 by dis-

carding “.

5.1. HOW TO GET SANSHOKU 117

Note that pre-committing to sanshoku of 456 by discarding “

before drawing › is massively ineﬃcient. If we do that, the hand

becomes a very weak hand as follows.

‰ı‚‚‹‹“¨˝˚ˇ˘¯

This hand relies too much on the possibility of drawing › ﬁrst. If

we draw any of ȷ‚‹ ﬁrst, the hand will be a yaku-less and/or

bad-wait hand.

118 CHAPTER 5. PURSUING YAKU

5.1.6 Golden

Sanshoku 6

˛‚‹›”„«˝˝˚ˇ˘¯ ﬁ

Draw

What would you discard?

If we keep ﬁ and discard ˚, the hand becomes what is known as

golden 1-away, as follows.

ﬁ˛‚‹›”„«˝˝ˇ˘¯

It is called “golden” because the hand is 1-away from ready for sanshoku

and 1-away from ready for ittsu (Pure Straight), two of t he most pop-

ular two-han yaku in riichi mahjong. Drawing ¸ or “ makes the

hand ready for ittsu, whereas drawing ﬀ or ﬂ makes the hand

ready for sanshoku of 789. The following are examples of golden

1-away.

–‰ıﬀﬁﬂ‹˜¨˝$$

ıȷﬀ¸˛‚‹›“”˜˜˚

5.1. HOW TO GET SANSHOKU 119

5.1.7 Crashing a group

Sanshoku 7

‰ıȷ¸˛‚‹›¨˝˝˚˚ ‚

Draw

What would you discard?

Notice that the hand can be made ready for pinfu if we discard ˝.

However, that gives us a pinfu-only hand. If we need an expensive

hand, we could take a rather high-handed approach and crash an

already complete run by discarding ¸. This might sound crazy, but

look how good a 1-away hand it becomes.

‰ıȷ˛‚‚‹›¨˝˝˚˚

If we draw “ or ¨, the hand becomes ready for tanyao + pinfu +

sanshoku + iipeiko (Pure Double Chow). Drawing ˇ also gets us

tanyao + pinfu + sanshoku, and drawing any of ˛ ‚ ‹ gets us

at least tanyao + pinfu, and possibly iipeiko as well.

120 CHAPTER 5. PURSUING YAKU

5.2 How to get ittsu

Ittsu (Pure Straight) is another popular two-han yaku. As we

will see below, even when ittsu appears to be a possibility, it is not al-

ways worthwhile to pursue this yaku at the cost of tile eﬃciency. We

will see several instances where pursuing ittsu is and is not worth

the cost.

5.2.1 Two non-overlapping runs

The key to getting ittsu is to pay attention to two non-overlapping

runs in a given suit. For example, suppose a hand has two non-

overlapping runs such as the following.

–—+ıȷﬀ

˛‚‹+“”„

˜¨˝+ˇ˘¯

Then, as soon as we draw another non-overlapping tile in the same

suit, ittsu is almost around the corner. Consider the following hand.

Ittsu 1

–—ıȷﬀ“„„ˆ¨˚˚ ﬂ

Draw

What would you discard?

From a pure tile eﬃciency perspective, the best discard choice is ﬂ.

However, doing so means giving up on ittsu and poking our way

toward a bad-wait yaku-less hand. That is not a very good path to

take even if we are ahead of the game in South-4 and don’t need an

expensive hand.

We would still want to have at least one yaku in a hand so that we can win it

5.2. HOW TO GET ITTSU 121

We should rather treat ﬂ as a treasure; we now have a realis-

tic possibility of getting ittsu. Let’s apply the ﬁve-block method to

ﬁgure out an alternative discard.

–—

|{z }

ıȷ

|{z}

ﬀﬂ

|{z}

“„„

|{z }

ˆ¨˚˚

| {z }

We are hoping to get three blocks in manzu (cracks) to have ittsu,

so we need one block in pinzu (dots) and another block in souzu

(bamboos). Recall the principle that each block should have at most

three tiles, which suggests we discard one tile from the block in souzu

(bamboos). The choice now boils down to discarding ˚ or ˆ. Recall

also that the value of pairs is maximized when there are two pairs in

a hand. We should thus discard ˆ.

5.2.2 Six-tile block with intervals

Consider diﬀerent six-tile conﬁgurations where we have a chunk

of six tiles with a few intervals among them. For example, consider

the following six-tile blocks.

–—+‰ȷ+ﬀﬂ

¸‚+›“+„«

´˜+¨˚+ˇ¯

We do see ittsu on the horizon with each of these tile chunks, but

aiming for ittsu with these blocks is not very realistic. Take the ﬁrst

six-tile block in manzu (cracks), for example. Even when we draw ﬁ,

we would want to discard – to have a double closed shape —‰ȷ

without calling riichi.

122 CHAPTER 5. PURSUING YAKU

and a run ﬀﬁﬂ rather than trying too hard to pursue ittsu. With

this in mind, consider the following hand.

Ittsu 2

–ıı‹““´˜¨˚ˇ¯ ”

Draw

What would you discard?

Although we see a remote possibility of ittsu in souzu (bamboos),

pursuing it requires we ﬁll in three closed-wait protoruns in souzu:

´˜, ¨˚, and ˇ¯.

It would be more practical to discard ´ and then ¯; we would

consider the tile blocks in souzu as a collection of two side-wait pro-

toruns: ˜¨ and ˚ˇ + two redundant terminal tiles: ´ ¯, rather

than considering it as a collection of three closed-wait protoruns.

5.2.3 Run + side-wait protorun

At one step prior to getting two non-overlapping runs, we may

have one run and a non-overlapping side-wait protorun in a given

suit. The following blocks are examples of such run + side-wait pro-

torun combinations.

‰ı+ﬀﬁ

‚‹+“”„

¨˝+ˇ˘¯

When we have a combination like these, a draw of – or — (left ex-

ample), ¸ or « (middle example), ´ or ˆ (right example) gener-

ates a realistic probability of getting ittsu. Below are the resulting

5.2. HOW TO GET ITTSU 123

tile blocks in each instance. You can pursue ittsu with any of these

blocks.

–+‰ı+ﬀﬁ ¸+‚‹+“”„ ´+¨˝+ˇ˘¯

—+‰ı+ﬀﬁ ‚‹+“”„+« ˆ+¨˝+ˇ˘¯

However, if the protorun is instead a closed-wait or an edge-wait

one, the chance of getting ittsu is much diminished. The follow-

ing blocks are examples of such run + closed- or edge-wait protorun

combinations. You may not want to pursue ittsu with these blocks.

‰ı+ﬁﬂ

¸˛+“”„

´˜+ˇ˘¯

If the run becomes a bulging ﬂoat block, you may want to give up on

ittsu and discard the closed- or edge-wait protorun part. With this

in mind, consider the following 2-away hand.

Ittsu 3

–‰ıﬁﬂ››˜˜˚## ‰

Draw

What would you discard?

Now that we drew a tile that creates a bulging ﬂoat block in manzu

(cracks), it is about time to give up on ittsu. Discarding – allows

124 CHAPTER 5. PURSUING YAKU

the hand to accept 9 kinds–25 tiles; if we stick with ittsu and discard

‰, the hand can accept only 4 kinds–10 tiles.

Moreover, although the hand is 2-away from ready, it is 3-away

from ittsu. Pursuing ittsu with a hand like this is not very practi-

cal.

5.2.4 Ittsu vs. side wait

As we saw with sanshoku hands, we often face a choice between

pursuing yaku and keeping a side-wait protorun. Consider the fol-

lowing 1-away hand.

Ittsu 4

‰ııȷﬀﬁﬂ‹››´´ – ›

Draw Dora

What would you discard?

If we discard – we drew, the hand is a perfect 1-away hand; the ﬁnal

wait can always be side wait. On the other hand, if we discard ı,

we have a conﬁrmed ittsu hand. Which option should we choose?

If we compare tile acceptance counts for t he two scenarios, the

option of conﬁrming side wait is slightly better (6 kinds–18 tiles vs.

5 kinds–16 tiles). However, doing so means we give up on ittsu.

Moreover, giving up on ittsu means that we can never call pon or

chii with this hand because there is no yaku in the hand. On the

other hand, the second option allows us to call pon on ›´ or call

chii on —. Even though the kinds and the number of acceptable

tiles are smaller, the second option would be more eﬃcient if we take

melding into account.

5.3. HOW TO GET PINFU 125

5.3 How to get pinfu

Although pinfu is only worth one han, the requirements to claim

pinfu are rather demanding. The key to getting pinfu is to build

side-wait protoruns even at the cost of tile eﬃciency. Consider the

following hand.

Pinfu 1

—ııﬁﬂ‹”„ˆ˜¨˚ –

Draw

What would you discard?

We already have ﬁve tile blocks in t his hand. From a pure tile eﬃ-

ciency perspective, discarding one of the two ﬂoating tiles ‹ or ˚

is the best. However, doing so signiﬁcantly reduces our chance of

getting pinfu. If we aim for pinfu we should discard t he edge-wait

protorun ﬁﬂ and keep the two ﬂoating tiles, which we hope may

grow into a side-wait protorun.

Suppose we discarded ﬁ, then we drew ›, after which we dis-

carded ﬂ. Now the hand is 1-away again, this time with two side-

wait protoruns. Suppose further that we drew !.

Pinfu 2

–—ıı‹›”„ˆ˜¨˚ !

Draw

What would you discard?

We should keep ! as a safe tile and discard ˚. It is true that keeping

˚ has an advantage; if we draw ˝, we will get a 3-way side-wait

block in souzu (bamboos). Even a draw of ˇ improves this hand

slightly. This is because having ‹› and ”„ is not very eﬃcient

due to the overlap of the waiting tiles; both blocks wait for “. It

126 CHAPTER 5. PURSUING YAKU

would be better to have ˚ˇ and ‹›, rather than having ‹›

and ”„.

However, keeping ˚ comes at a cost. Even when we draw ˝,

we will then have to discard ‹› or ”„, possibly against an op-

ponent’s riichi. Therefore, once we get a 1-away with two side-wait

protoruns (side ’n’ side 1-away; ryanmen-ryanmen 1-shanten), we

should try to keep a safety tile in the hand.

Even when we draw a tile that makes a hand perfect 1-away, we

may still want to have a safety tile. For example, drawing any of

‹›”„ makes the hand above perfect 1-away. Although perfect

1-away is better than side ’n’ side 1-away in terms of tile acceptance,

a perfect 1-away hand can end up not having pinfu because a set can

emerge in the hand.

There is one exception to this, however. If the ﬂoating tile leaves

a possibility of enhancing the hand value by at least three han, it is

OK to keep it instead of a safety tile. Consider the following hand.

Pinfu 3

–—ıﬀﬁﬂ‚‹””ˆ˜ ! ‰

Draw Dora

What would you discard?

Keeping ! is safer, but keeping ı leaves t he possibilities of getting

ittsu and having dora, possibly at the same time. In this case, we

would rather discard !.

5.3. HOW TO GET PINFU 127

Building the head

To claim pinfu, the head must be a pair of number tiles or value-

less wind tiles. Therefore, try not to discard terminals or valueless

wind tiles lightly when having a pinfu hand. Keep this in mind espe-

cially when a hand is lacking any pair. Assuming you are the South

player in the 1st turn in East-1, consider the following hand.

Pinfu 4

–‰ȷ˛‚“”«˝˚˚ˇ!$

What would you discard?

The hand has a potential to have pinfu, so we should not discard any

of –«! at this point. All of these three tiles may appear useless,

but they can be t he head of a pinfu hand when any of them grows

into a pair. On the other hand, the value tile $ cannot be the head

of pinfu. We should discard $ in this case.

128 CHAPTER 5. PURSUING YAKU

5.4 How to get honitsu

Going for honitsu (Half Flush) can be a good way to achieve high

hand values. As we can combine honitsu with many other yaku, in-

cluding fanpai, toitoi, chanta (Outside Hand), ittsu, among oth-

ers,

we can aim for mangan relatively easily. The fact that honitsu

is worth two han even when we open our hand means we can also

enhance the speed by melding without making our hand too cheap.

5.4.1 Conditions to go for honitsu

When judging whether to go for honitsu or not, we should con-

sider two factors — ﬁve-block potential and hand values with and

without honitsu.

1. Five-block potential

The most important factor to consider is whether or not your hand

has ﬁve tile blocks or block candidates (i.e., ﬂoating tiles) necessary

for honitsu. Assuming you are the South player in the 6th turn in

East-1, consider the following hand.

Honitsu 1

–ııﬀﬂ„ˆ¨ﬄﬄ!%%

Would you go for honitsu?

In order to ﬁgure out if it is practical to pursue honitsu with this

hand, let’s apply the ﬁve-block method.

ıı

|{z}

ﬀﬂ

|{z}

ﬄﬄ

|{z}

–!

|{z}

„ˆ¨

Technically speaking, honitsu can be combined with chiitoitsu, shousangen

(Little Three Dragons), honroutou (All Terminals and Honors), pinfu, iipeiko,

ryanpeiko, san ankou (Three Concealed Pungs), and san kantsu (Three

Kongs) as well.

5.4. HOW TO GET HONITSU 129

We can count on the two pairs of fanpai, ﬄﬄ and %%, to be two

tile blocks, a pair of ıı and a protorun ﬀﬂ to be another two

blocks, yielding four blocks in total. In addition, we can reasonably

expect either of the two ﬂoating tiles !– to produce the ﬁfth block.

Therefore, you can go for honitsu with this hand.

In addition, the tiles that are made redundant in t he hand if we

choose honitsu are an isolated „ and a closed-wait protorun ˆ¨.

Keeping these tiles would not make this hand particularly more eﬃ-

cient anyway, so we can go for honitsu without much hesitation.

Even when we have a side-wait protorun or a pair to discard, we

may still want to go for honitsu. For example, with the following

two hands, you should go for honitsu even though doing so means

you have to discard a side-wait protorun or a pair.

—¸‚‹››“„„!###

‹‹´ˆ¨¨˝˝¯ﬄ "%

2. Hand value

Another factor to consider is hand value comparison with and

without honitsu. If your hand does not have any yaku potential (e.g.,

pair or set of fanpai) other than honitsu, you may end up getting a

honitsu-only hand, which is very cheap (2000 or 2600 points). In

such situations, you should not aim for honitsu; you should try to

make the hand ready wit hout melding and go for riichi. Assuming

you are the South player in East-1, consider the following hand.

130 CHAPTER 5. PURSUING YAKU

Honitsu 2

–—‰ȷﬁﬁ‹›ˇˇ!!""

Would you go for honitsu?

Since West and North are both valueless wind tiles, this hand is likely

to become honitsu-only if you decide to go for honitsu. Although

this hand has ﬁve tile blocks necessary for honitsu, you should not

go for honitsu.

At the same time, when the hand value is suﬃciently high (≥

5200) without honitsu, you should not go for honitsu at the cost of

tile eﬃciency. Assuming you are the South player in East-1, consider

the following hand.

Honitsu 3

Red

––‰»ﬀ‚‹$$$! H

What would you discard?

This hand is worth 5200 points without honitsu (Seat Wind + Green

Dragon + red ﬁve), so you should discard ﬀ to maintain a side ’n’

side 1-away status. If were not your seat wind, you should go for

honitsu.

5.4. HOW TO GET HONITSU 131

5.4.2 Discard

When pursuing honitsu, pay attention to the order of your dis-

cards. Consider the following hand. You called pon on # just now,

deciding what to discard.

Honitsu 4

––—ȷﬁ”˜˝""% ##K

What would you discard?

You should pursue honitsu, so ”˜˝ are your discard candidates.

You will discard all three of them eventually, but you should discard

them in a way that looks less obvious that you are collecting tiles in

manzu (cracks). If you discard ” ﬁrst then ˝ next, the opponents

might (correctly) guess that you are doing honitsu with manzu tiles.

In particular, the Left player may stop discarding tiles in manzu that

you could call chii on. You should thus discard ˝ ﬁrst then ˜

next, so that the opponents cannot know if you are collecting manzu

or pinzu (dots). They will eventually ﬁnd out that you are collecting

manzu, but you should delay that as much as possible.

132 CHAPTER 5. PURSUING YAKU

5.4.3 Melding

When you start melding with a honitsu hand, try to leave the

possibility of achieving the maximum hand value. Assuming you

are the South player in the 6th turn in East-1, consider the following

hand.

Honitsu 5

ﬂﬂ˛‚›”„«!""$$

Which tile would you call?

With this hand, do not start melding with a chii of ‹ or a pon of

" if you are playing wit hout red ﬁves; you may end up with a very

cheap (2000 points) honitsu-only hand. Suppose you managed to

call pon on $, resulting in the following hand.

ﬂ˛‚›”„«!"" $L$

Calling chii on ‹ or pon on " is still not ideal. The only melding

you should do is to call chii on ¸ to have the following hand.

›”„«!"" 3˛‚ $L$

Notice that the two ﬂoating tiles › ! allow us to envision two

possibilities of getting a 7700 hand. On the one hand, if you draw !

or call pon on ", you get honitsu + Green Dragon + chanta (Outside

Hand), as follows.

”„«!!"" 3˛‚ $L$

5.4. HOW TO GET HONITSU 133

”„«! ""J 3˛‚ $L$

On the other hand, if you draw ‹ or “, the hand will be ready

for honitsu + Green Dragon + ittsu.

‹›”„«"" 3˛‚ $L$

›“”„«"" 3˛‚ $L$

134 CHAPTER 5. PURSUING YAKU

5.5 How to get toitoi / chiitoitsu

5.5.1 Toitoi vs. chiitoitsu

When pursuing chiitoitsu (Seven Pairs), you may ﬁnd yourself

standing at a crossroad between chiitoitsu and toitoi (All Pungs).

Speciﬁcally, what should we do when one of the pairs in a 1-away

chiitoitsu hand becomes a set? Assuming you are the South player

in the 6th turn in East-1, consider the following hand.

Toitoi vs. chiitoitsu

ﬀﬀﬁﬁ¸‚‚˝˝ˇ$$" ‚

Draw

What would you discard?

If we discard ‚ that we drew, the hand is 1-away from ready for

chiitoitsu, accepting ¸ˇ" (3 kinds–9 tiles). On the other hand, if

we keep it and discard ˇinstead, the hand is still a 1-away chiitoitsu

hand, albeit with smaller tile acceptance. However, doing so makes

the hand also 2-away from ready for toitoi and possibly su anko

(Four Concealed Pungs).

Judgement criteria for a choice of this kind are summarized as fol-

lows.

Toitoi vs. chiitoitsu

Choose chiitoitsu in the following situations.

1. There is a futile pair in your hand.

2. There is no pair of value tiles in your hand.

3. There are three or more pairs of simple tiles between 3 and

7 in your hand.

5.5. HOW TO GET TOITOI / CHIITOITSU 135

The ﬁrst condition is by far the most important one. With the current

hand example, if the opponents have already discarded two tiles of

ﬀ, the pair of ﬀ in the hand is a futile pair (dead pair) that will

never become a set. If there is one or more futile pair in your hand,

you must stick with chiitoitsu. If not, you can go for toitoi.

In addition, you may also want to take into account the second

and t he third conditions. Speciﬁcally, without having a pair of value

tiles (fanpai), you may end up with a toitoi-only hand (2600 or even

2000 points). With one pair of value tiles, you can aim for 5200 with

toitoi; with two pairs of them, you can aim for mangan.

The third factor to consider is whether there are not three or more

pairs of simple tiles between 3 and 7. Consider the following hand.

ȷﬀﬀ››ˆˆ˚˚##$%

Suppose you start melding by calling pon on #,

then get another

pon on ˆ, resulting in the following hand.

ﬀﬀ››˚˚% ˆˆ* #K#

Since the remaining three pairs are all simple tiles between 3 and 7,

the hand advancement often stops here. Because of their high versa-

tility, simple tiles between 3 and 7 are very likely to be used by the

opponents.

Calling pon on # makes this 1-away chiitoitsu hand 2-away from ready for

toitoi. Doing so would be acceptable if the remaining pairs were not simple

tiles between 3 and 7.

136 CHAPTER 5. PURSUING YAKU

5.5.2 Standard hand vs. chiitoitsu

Another kind of crossroad is between chiitoitsu and standard

hand. Assuming you are the South player in the 6th turn in East-1,

consider the following hand.

Pinfu vs. chiitoitsu

‰ıȷȷ‹›„„˚˚ˇ"" ı

Draw

What would you discard?

As we draw ı, we now have ﬁve pairs in the hand, making it 1-away

from ready for chiitoitsu. However, the hand is also 2-away from

ready if we interpret this hand as a standard hand.

When a hand has this many side-wait protoruns, it makes more

sense to view it as a standard hand rather than as a chiitoitsu hand.

To ﬁgure out what tile to discard, let’s apply the ﬁve-block met hod.

‰ııȷȷ

| {z }

‹›

|{z}

„„

|{z}

˚˚ˇ

|{z}

Since we already have ﬁve tile blocks in simple tiles, the pair of " is

redundant. Discarding " means we are giving up on chiitoitsu,

but we are maximizing tile acceptance to make the hand ready for

tanyao as soon as possible. Doing so leaves a decent chance of getting

pinfu and iipeiko as well. The expected hand value will actually be

higher if we give up on chiitoitsu.

On the other hand, when a hand has few side-wait protoruns but

has several pairs, you should pursue a pair-based (set-based) hand

5.5. HOW TO GET TOITOI / CHIITOITSU 137

rather than a run-based hand. Assuming you are the South player in

the 6th turn in East-1, consider the following hand.

Run-based hand vs. set-based hand

––——‹›ˆ¨¯¯ﬄ # #

Draw

What would you discard?

Although the hand has one side-wait protorun, we would not be very

happy if it were to evolve into a complete run; we may end up with a

very cheap hand with a bad wait. Alternatively, you should pursue

chiitoitsu or toitoi wit h a hand like this. Let’s apply the ﬁve-block

method to ﬁgure out what tile to discard.

––

|{z}

——

|{z}

¯¯

|{z}

‹›ˆ¨ﬄ

Since we intend to build four blocks using the four pairs in the hand,

we only need one more block from the rest of t he tiles: ‹›ˆ¨ﬄ .

Of these six tiles, ﬄ ˆ are clearly more valuable than others be-

cause of their low versatilities. On the other hand, ‹›¨ are less

useful for you because these tiles have high versatility for the op-

ponents. We should discard ‹ or ¨ ﬁrst, keeping › just in case

we draw a red ˙ (in which case the hand is 1-away from ready for

chiitoitsu).

Part III

Strategy Principles

138

Chapter 6

Scoring

The scoring system in mahjong is quite complex. Getting proﬁ-

ciency in score calculation requires a lot of practice. The good news

is that scoring is automatically done once you win a hand when you

play online. Even when you play oﬄine, you can usually count on

your fellow players to help you get t he correct score once you win a

hand.

However, you often need to calculate the (potential) scores of your

hand before you win the hand. This is because a lot of important

judgements you make during the game — riichi judgement, defense

judgement, and melding judgement, among others — depend on the

potential scores of your hand. Therefore, developing ability to calcu-

late t he scores correctly and quickly without any help of others is of

utmost importance. I introduce some eﬃcient methods of score cal-

culation in this chapter before we discuss riichi, defense, and meld-

ing judgements in the subsequent chapters.

6.1 Three steps in score calculation

Every rule book of mahjong has comprehensive scoring tables

(similar to Tables 6.10 and 6.11 at the end of this chapter) that show

all possible scores for all possible minipoints (fu). Although such

tables are a good reference to have, it is not very eﬃcient to try to

memorize everything in such tables.

A more practical approach would be to focus on a small number

of frequently observed patterns of scoring and memorize them cor-

rectly, while ignoring other, less important ones. Before introducing

6.1. THREE STEPS IN SCORE CALCULATION 141

some shortcuts to do eﬃcient scoring, let’s ﬁrst review the three re-

quired steps in score calculation, summarized in a box on the next

page.

Three steps in score calculation

Step 1: Count the number of han.

First, you need to ﬁgure out how many han a hand has. If

a hand has ﬁve or more han, skip Step 2 and go directly to

Step 3. If not, proceed to Step 2.

Step 2: Figure out the minipoints.

When a hand has four or less han, you then need to know

the hand’s minipoints. This does not mean, however, that

you always need to do some maths to get the correct mini-

points. We will discuss some practical shortcuts below.

Step 3: Get the score.

Based on t he number of han (and possibly minipoints), you

get the score. You will have to memorize some score pat-

terns.

In the remainder of this chapter, I will ﬁrst introduce basic meth-

ods of score calculation in Section 6.2. The basic methods involve

using some shortcuts in Step 2 above. Once you master the basic

methods, you will be able to calculate scores correctly most of the

time.

When you master the contents of Section 6.2, you may skip

Section 6.3 and proceed to the next chapter. Section 6.3 covers more

advanced methods of score calculation, which would be necessary

In my impression, roughly 75 % of the hands we encounter can be covered by

the basic methods.

142 CHAPTER 6. SCORING

only in exceptional cases. This involves an exact calculation of mini-

points in Step 2 above.

6.2 Basic scoring

1. Counting the number of han

Step 1 in score calculation is counting the number of han in a

hand. This is the most important part in score calculation, and there

is no useful shortcut here. You need to be able to identify all the yaku

in a hand as well as the associated han counts for each.

A good way to practice this is to try to beat the automatic score

counting on Tenhou. Whenever someone wins a hand, Tenhou dis-

plays all the yaku and the associated han counts one after another in

a few seconds. Try to identify all the yaku of your opponent’s hand

before they get displayed automatically.

Scores for limit hands

When a hand has ﬁve or more han, the hand is a limit hand.

Scores of limit hands do not depend on minipoints so we can go di-

rectly to Step 3. To get the score for a given han count, we utilize Table

6.1. This is something you need to memorize.

As you can see, there are some regularities and redundancies that

make it relatively easy to memorize this table. The most important

score of all is 8000 (mangan ron for non-dealer). This is the basis of

all the other scores in this table. For example, haneman scores are

1.5 times mangan scores, baiman scores are two times mangan scores,

sanbaiman scores are three times mangan scores,

and yakuman scores

are four times mangan scores. In addition, scores for dealer are exactly

Since bai means “twice” or “double” in Japanese, baiman literally means dou-

ble mangan in Japanese. Similarly, sanbai means “triple,” so sanbaiman literally

means triple mangan.

6.2. BASIC SCORING 143

Table 6.1: Scores for limit hands

Han Name Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

5 mangan 8000 12000 2000-4000 4000-all

6–7 haneman 12000 18000 3000-6000 6000-all

8–10 baiman 16000 24000 4000-8000 8000-all

11–12 sanbaiman 24000 36000 6000-12000 12000-all

13+

∗

yakuman 32000 48000 8000-16000 16000-all

∗

A hand with 13+ han is scored as a sanbaiman wit h the revised EMA rules.

1.5 times the corresponding scores for non-dealer in all limit hands.

Finally, scores for tsumo (self draw) cases are simple and straight-

forward; the dealer pays one half of the total, and each of the two

non-dealers pays one fourth of the total. For example, in the case of

mangan tsumo (8000), the dealer pays 4000 and non-dealers pay 2000

each.

2. Figuring out the minipoints

When a hand has four or less han, you have to know the mini-

points. As I pointed out earlier, this does not mean that you always

have to count all the minipoint contributions from all the melds and

wait in a hand. Such a calculation is required only in special cases.

Instead, you can use the chart in Figure 6.1 that summarizes the six

most frequently observed patter ns you need to memorize.

You can use this Figure as a ﬂowchart. You ﬁrst check if the hand

is a chiitoitsu (Seven Pairs) hand. If it is, it is always 25 minipoints.

If it is not, you t hen check if the hand has one or more quads (kongs).

If it does, the hand is out of the scope of the basic methods. Ask for

help from more experienced players after winning the hand. Ad-

vanced methods we discuss in Section 6.3 will cover this exceptional

144 CHAPTER 6. SCORING

Shortcut for minipoint calculation

1. chiitoitsu ⇒ always 25 minipoints

2. A hand has one or more quads ⇒ Don’t bother.

3. toitoi ⇒ almost always 40 minipoints

4. Pinfu

• ron ⇒ always 30 minipoints

• tsumo ⇒ always 20 minipoints

5. Closed hand without pinfu

• ron ⇒ almost always 40 minipoints

• tsumo ⇒ almost always 30 minipoints

6. Open hand ⇒ almost always 30 minipoints

Figure 6.1: Six most observed patterns

case. Third, you check if the hand is toitoi (All Pungs).

If it is, it’s

almost always 40 minipoints unless it is a tanyao toitoi hand that

is likely to have 30 minipoints.

Once you rule out the ﬁrst three cases (chiitoitsu, quads, and

toitoi), the last three are the most important ones; the great ma-

jority of hands you see will be one of these three. Here, you check

two things. First, check if it is a pinfu hand. If it is, it’s always 30

minipoints (ron) or 20 minipoints ( tsumo). If it is not pinfu, you then

check if it is a closed hand or an open hand. If it is a closed hand, it

Note that we are only talking about open toitoi here. A closed toitoi does

not require a minipoint calculation under any circumstance. If you win it by

tsumo, it’s yakuman (su anko; Four Concealed Pungs); if you win it by ron, it’s

at least mangan (toitoi and san anko).

6.2. BASIC SCORING 145

is almost always 40 minipoints (ron) or 30 minipoints (tsumo). If it is

an open hand, it is almost always 30 minipoints whet her you win it

by ron or tsumo.

Because of the importance of t he cases 4, 5, and 6 ( pinfu, closed,

and open hand), we will ﬁrst discuss these three cases. We will then

discuss cases 1 and 3 ( chiitoitsu and toitoi), which are way more

exceptional.

What do we mean by “almost always”?

Before getting to score calculation, let me explain what we mean

when we say something is almost always X minipoints in cases 5 and

6 in Figure 6.1. With a non-toitoi hand without quads, check if the

hand has one or more concealed set of terminal/honor tiles or its

equivalent. Recall that one concealed set of terminal/honor tiles is

equivalent to two open sets of terminal/honor tiles or two concealed

sets of simple tiles in minipoints. Therefore, what we check is whether

a hand has any of the following:

• at least one concealed set of terminal/honor tiles;

• at least two open sets of terminal/honor tiles;

• one open set of terminal/honor tiles and one concealed set of

simples;

• at least two concealed sets of simples.

When none of the four above exists in a non-toitoi hand wit hout

quads, a hand is always 40 minipoints if closed and always 30 mini-

points if open. Therefore, there is really no need of an actual calcu-

lation of minipoints with hands like the following.

––‹›ˆ˜¨ˇ˘¯ %M%

146 CHAPTER 6. SCORING

‰ı´´˜¨˝ Gﬄﬄ ?

———ﬀﬁˆ˜¨˚˚ˇ˘¯

With these hands, we can simply check if it is a closed hand or not to

determine if each hand has 40 minipoints (closed) or 30 minipoints

(open).

On the other hand, if a hand satisﬁes any of the four conditions

above, we need to calculate the minipoints by actually counting and

summing all the minipoint contributions from all the melds, the head,

and the wait to determine the minipoints of a hand. We will discuss

this in Section 6.3.

3. Getting the scores

As Figure 6.1 makes clear, t he case of 30 minipoints is the most

important pattern. We thus start with this pattern. We will then

proceed to the cases of 40, 20, and 25 minipoints.

30 minipoints

You get 30 minipoints when you get:

• pinfu ron (always);

• closed hand tsumo (almost always); or

• open hand ron / tsumo (almost always).

Scores for 30 minipoints are summarized in Table 6.2.

Compared with the limit hands table (Table 6.1), the regularities

in the 30 minipoints table are less precise. For example, scores for

dealer are only roughly 1.5 times those for non-dealer; tsumo scores

6.2. BASIC SCORING 147

Table 6.2: Scores for 30 minipoints

Han Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

1 1000 1500 300-500 500-all

2 2000 2900 500-1000 1000-all

3 3900 5800 1000-2000 2000-all

4 7700 11600 2000-3900 3900-all

5+ limit hand

are sometimes slightly bigger than the corresponding ron scores. For

example, one-han tsumo (300-500) gives you 300 + 300 + 500 = 1100,

which is slightly bigger than one-han ron (1000).

Therefore, it would be more eﬃcient if we just memorize these

patterns as they are, rather than trying to simplify them. Japanese

players tend to memorize Table 6.2 column-wise, as follows:

Scores for 30 minipoints

• ron (non-dealer): 10, 20, 39, 77

• ron (dealer): 15, 29, 58, 116

• tsumo (non-dealer): 3-5, 5-10, 10-20, 20-39

The beneﬁt of memorizing this score table column-wise is that scores

get (roughly) twice as big for an additional han. Moreover, if we

memorize the tsumo scores for non-dealer, we can easily derive those

for dealer. When I was trying to memorize these, I used to recite these

sequences a number of times so that they get beaten into my head.

148 CHAPTER 6. SCORING

40 minipoints

You get 40 minipoints when you get:

• non-pinfu closed hand ron (almost always); or

• toitoi ron / tsumo (almost always).

Scores for 40 minipoints are summarized in Table 6.3.

Table 6.3: Scores for 40 minipoints

Han Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

1 1300 2000 400-700 700-all

2 2600 3900 700-1300 1300-all

3 5200 7700 1300-2600 2600-all

4+ limit hand

Just like we did with scores for 30 minipoints, I recommend you

memorize this column-wise.

Scores for 40 minipoints

• ron (non-dealer): 13, 26, 52, mangan

• ron (dealer): 20, 39, 77, mangan

• tsumo (non-dealer): 4-7, 7-13, 13-26, mangan

The good news is that the ron score sequence for non-dealer —13,

26, 52— is a geometric progression; 13 × 2 = 26, and 26 × 2 = 52.

Also, the ron scores for dealer —20, 39, 77— should look familiar to

you if you have already memorized the 30 minipoints ron scores for

non-dealer —10, 20, 39, 77.

6.2. BASIC SCORING 149

20 minipoints (pinfu tsumo)

Scores for 20 minipoints are summarized in Table 6.4. This table is

special in the sense that it does not have the ron score component

nor the one-han row. This is because you get 20 minipoints only

when you get pinfu + tsumo (hence we have at least two han). Even

though this is a special case, pinfu + tsumo is far from a rare occur-

rence. It is thus important to know how to get the correct scores for

this case.

Table 6.4: Scores for 20 minipoints

Han Tsumo

Non-dealer Dealer

2 400-700 700-all

3 700-1300 1300-all

4 1300-2600 2600-all

5+ limit hand

Notice that there is an interesting similarity between the 20 mini-

points table (Table 6.4) and the 40 minipoints table (Table 6.3). The

scoring patterns are almost identical, except that the required num-

ber of han to get a certain score is one han smaller for 40 minipoints

than for 20 minipoints; that is, to get 400-700, we need 2 han with 20

minipoints, whereas we only need 1 han with 40 minipoints.

This is not a coincidence. In general, when we double the mini-

points, we need one less han to get the same score. For example, a

3 han–30 minipoints hand and a 2 han–60 minipoints hand have the

same score (3900 for non-dealer; 5800 for dealer); a 2 han–25 mini-

points hand and a 1 han–50 minipoints hand have the same score

(1600 for non-dealer; 2400 for dealer). Therefore, once you memo-

150 CHAPTER 6. SCORING

rize scores for 30 minipoints, we can easily deduce scores for hands

with 60 minipoints. Similarly, we can deduce scores for hands with

50 or 80 minipoints once we memorize scores for 25 or 40 minipoints,

respectively.

25 minipoints (chiitoitsu)

The ﬁnal case we cover as part of the basic scoring is scores for a

chiitoitsu hand. As chiitoitsu is a rather exceptional yaku, it is

given exceptional minipoints — 25 minipoints. Scores for 25 mini-

points are summarized in Table 6.5.

Table 6.5: Scores for 25 minipoints

Han Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

2 1600 2400

3 3200 4800 800-1600 1600-all

4 6400 9600 1600-3200 3200-all

5+ limit hand

Since chiitoitsu is itself a two-han yaku, the scoring table only

starts with two han for ron and three han for tsumo. Again, we mem-

orize this table column-wise.

Scores for 25 minipoints

• ron (non-dealer): 16, 32, 64, mangan

• ron (dealer): 24, 48, 96, mangan

• tsumo (non-dealer): 8-16, 16-32, mangan

All the sequences are a geometric progression, making it rela-

6.2. BASIC SCORING 151

tively easy to memorize.

Practice, practice, practice

This completes the basic methods of score calculation. Memoriz-

ing all the scores for limit hands as well as the cases of 30, 40, 20, and

25 minipoints should be more than enough. Of course, no one will

be able to master this method just by reading and understanding the

materials in this chapter. You would need to actually practice what

you have learned, and you will have to do so repeatedly.

152 CHAPTER 6. SCORING

bababababababababababababababab

Notes on pinfu tsumo

One of the common mistakes that beginners tend to make is to

claim 1000-2000 for riichi + pinfu + tsumo (mistaking 3 han–20

minipoints for 3 han–30 minipoints) or claim mangan for riichi +

pinfu + tsumo + dora (mistaking 4 han–20 minipoints for 4 han–

40 minipoints). If you have trouble wrapping your head around

why pinfu + tsumo hands are given lower minipoints, knowing

the origin of this rule might be helpful.

The yaku pinfu is realized when a hand has no component that

generates an additional minipoint. A pinfu hand cannot have a

set, edge wait, closed wait, single wait, or a pair of value tiles be-

cause they all generate an additional minipoint. Recall that tsumo

also generates 2 minipoints. Therefore, logically speaking, pinfu

cannot be claimed when you win it by tsumo. In fact, some tradi-

tional mahjong rule sets do not allow a combination of pinfu and

tsumo. Under such rule sets, when you win a pinfu-only hand

by tsumo, you are allowed to claim tsumo only, giving you 1 han–

30 minipoints (20 base minipoints + 2 minipoints for tsumo = 22,

rounded up to 30 minipoints), which gives you 300-500.

However, some people thought that this is a bit unfair, claiming

that a pinfu tsumo hand should be given a higher score than a

tsumo-only hand. At the same time, they recognized that the score

for pinfu + tsumo should be lower than that for a “proper” 2-han

tsumo hand (e.g., tanyao + tsumo). Therefore, they decided that

pinfu + tsumo should be placed in between these two — 1 han–30

minipoints (tsumo only; 300-500) and 2 han–30 minipoints (tanyao

+ tsumo; 500-1000) — giving it the score of 400-700. Therefore, the

score for riichi + pinfu + tsumo (3 han–20 minipoints; 700-1300)

is higher than riichi + tsumo (2 han–30 minipoints; 500-1000) but

lower than riichi + tanyao + tsumo (3 han–30 minipoints; 1000-

2000).

6.3. ADVANCED SCORING 153

6.3 Advanced scoring

As I mentioned in the previous section, a hand with one or more

concealed set of terminal/honor tiles (or its equivalent) or quads may

have unusually high minipoints, calling for an actual calculation of

minipoints.

6.3.1 Minipoint calculation

Let’s ﬁrst review the basics of minipoint calculations. All stan-

dard hands (i.e., hands wit h melds) have the base 20 minipoints.

Then, we add the following minipoints depending on how we win

the hand:

• tsumo (open or closed, except for pinfu): 2

• ron (closed): 10

• ron (open): 0

We add further minipoints for each set and quad in a hand depend-

ing on whether it is a concealed one or an open one. Table 6.6 sum-

marizes minipoint contributions from a set and a quad.

Table 6.6: Minipoint contributions from a set and a quad

Tile Minipoint

Open Concealed

set simple 2 4

terminal/honor 4 8

quad simple 8 16

terminal/honor 16 32

154 CHAPTER 6. SCORING

Finally, we add 2 minipoints for each of the following, if any.

• Pair of dragon tiles

• Pair of seat wind tiles

• Pair of prevailing wind tiles

• Closed, edge, or single wait

When the head of a hand is of seat wind and prevailing wind (e.g.,

ﬄ for the East player in the East round), we get 2 + 2 = 4 minipoints.

If the wait is eit her side wait or dual pon wait, we don’t get any mini-

point for it. As we saw when we discussed wait patterns in 3.4, we

may get diﬀerent minipoints depending on which of the multiple

winning tiles to win on. For example, consider the following hand.

‰‰ıȷ‹‹%%% ˜˜+

The hand is waiting for —-ı. If we win by ron on —, we get no

minipoint for the wait and so this hand has 30 minipoints (base 20

+ 8 for a concealed set of honors + 2 for an open set of simple = 30).

However, if we win this hand on ı, we get additional 2 minipoints

for closed wait. This is because ‰‰ıȷ can be thought of as

‰ı + ‰ȷ. Therefore, the hand has 40 minipoints in that case

(30 + 2 = 32, rounded up to 40).

6.3.2 Scores for 50 minipoints or above

When a hand has one or more concealed set of honor tiles, the

hand may have 50 minipoints or above. You may want to memorize

This is the case with EMA rules and Tenhou rules, but this is not a universally

adopted rule.

6.3. ADVANCED SCORING 155

the case of 50 minipoints, summarized below. If you are a perfec-

tionist, you may also want to memorize the cases of 70 and 110 mini-

points as well, but I can assure you that it would not be worth t he

eﬀort.

50 minipoints

Scores for 50 minipoints are quite easy to memorize if you have al-

ready memorized scores for 25 minipoints (chiitoitsu), summa-

rized in 6.5. Recall that the score for a 1 han–50 minipoints hand

should be the same as that for a 2 han–25 minipoints hand.

Table 6.7: Scores for 50 minipoints

Han Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

1 1600 2400 400-800 800-all

2 3200 4800 800-1600 1600-all

3 6400 9600 1600-3200 3200-all

4+ limit hand

70 minipoints

Hands with 70 minipoints do not appear very often (probably once

in 20 games or so). Table 6.8 summarizes scores for 70 minipoints.

If you would like to memorize the table, notice that it is sequential

(until the end): 23 (non-delaer) → 34 (dealer) → 45 (non-dealer) →

68 (dealer).

110 minipoints

For the sake of completeness, Table 6.9 summarizes scores for 110

minipoints.

156 CHAPTER 6. SCORING

Table 6.8: Scores for 70 minipoints

Han Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

1 2300 3400 600-1200 1200-all

2 4500 6800 1200-2300 2300-all

3+ limit hand

Table 6.9: Scores for 110 minipoints

Han Ron Tsumo

Non-dealer Dealer Non-dealer Dealer

1 3600 5300 — —

2 7100 10600 1800-3600 3600-all

3+ limit hand

110 minipoints occur only in extremely rare occasions. Consider

the following hand.

“”„ﬄﬄ%% ﬃ´´ﬃ ﬃ¯¯ﬃ

Suppose you are in the East round. If the dealer wins this hand by

ron on %, he gets 20 (base) + 10 (closed hand ron) + 4 (pair of seat &

prevailing wind tiles) + 4 (open set of honors) + 32 (concealed quad

of terminals) + 32 (concealed quad of terminals) = 102, rounded up

to 110 minipoints. The score for 1 han–110 minipoints is 5300.

If he wins by ron on ﬄ, on the other hand, he gets more han (set

of seat & prevailing wind) but lower minipoints. This is because the

minipoint contribution of the pair of % (2) is smaller than that of the

pair of ﬄ (4). Since t he score of 2 han–100 minipoints hand is the

same as that of 3 han–50 minipoints hand, he obtains 9600 points.

6.3. ADVANCED SCORING 157

6.3.3 Examples

Let’s see how scores change as we advance a hand. For each of

the examples below, try calculating the scores for diﬀerent winning

tiles and for tsumo and ron.

Scoring 1

——‰ı‹››“”%%%

What are the scores?

If you win this hand by ron, the hand has 1 han (Red Dragon) and

40 minipoints: 20 (base) + 10 (closed hand ron) + 8 (concealed set of

honors) = 38, rounded up to 40, so you get 1300 points.

If you win it by drawing ‚, the hand has an additional yaku,

menzen tsumo (Fully Concealed Hand), with 30 minipoints: 20 + 8 +

2 (tsumo) = 30. You thus get 500-1000 tsumo = 2000 points.

However, if you win it by drawing “, you get 40 minipoints be-

cause of the additional 2 minipoints for

closed wait: ›”. You thus get 700-1300 tsumo = 2700 points.

Let’s say you draw ›. What would you discard?

——‰ı‹››“”%%% ›

Draw

If you discard ›, the wait is ‚-“ (2 kinds–7 tiles). If you discard

‹, however, you get a 3-way wait: ›-„ — (3 kinds–7 tiles). Let’s

say you choose the latter, resulting in the following hand. Now, let’s

think about the scores for each winning tile.

158 CHAPTER 6. SCORING

Scoring 2

——‰ı›››“”%%%

What are the scores?

If you win this hand on —, the hand is still 1 han–40 minipoints =

1300. However, if you win on › or „, the three tiles of › within

the hand are treated as a concealed set, giving you 4 additional mini-

points: 20 (base) + 10 (closed ron) + 8 (set of %) + 4 (set of ›) = 42,

rounded up to 50 minipoints. You thus get 1 han–50 minipoints =

1600 points. If you win by tsumo, you get 40 minipoints so you will

get 700-1300 = 2700 points.

Let’s say you draw . What would you discard?

——‰ı›››“”%%% 

Draw

If you discard ı, t he wait is –-‰. Let’s think about the scores for

each winning tile.

Scoring 3

——‰›››“”%%%

What are the scores?

If you win the hand by ron on ‰, you get an additional yaku, iipeiko

(Pure Double Chow), giving you 2 han–40 minipoints = 2600 points.

If you win the hand by tsumo, the minipoints are now lower than

before because you have side wait and only one concealed set; we

6.3. ADVANCED SCORING 159

cannot think of the three tiles of › as a set any more. You get 2 han–

30 minipoints if you draw – (500-1000 tsumo = 2000), whereas you

get 3

han

–30 minipoints if you draw

‰

(1000-2000

tsumo

= 4000).

Let’s say you draw “. What would you discard?

——‰›››“”%%% “

Draw

Discarding ”is t he best option. To understand why, let’s think about

the scores.

Scoring 4

——‰›››““%%%

What are the scores?

Notice that the wait and the potential han counts did not change at

all. However, you get increased minipoints because you now have

the concealed set of › back again. You will get 2 han–50 minipoints

= 3200 points if you win by ron on ‰. If you draw ‰, you will get 3

han–40 minipoints, giving you 1300-2600 tsumo = 5200 points.

Let’s say you draw another “. What would you discard?

——‰›››““%%% “

Draw

The best discard is ‰, which makes this hand a toitoi (All Pungs)

hand, as follows.

160 CHAPTER 6. SCORING

Scoring 5

——›››“““%%%

What are the scores?

The hand has three concealed sets already, giving you at least san

anko (Three Concealed Pungs) in addition to toitoi and Red Dragon.

Now you no longer need any tedious minipoints calculation. If you

win this hand by tsumo, it is yakuman (su anko; Four Concealed Pungs).

If you win it by ron, you get ﬁve han (toitoi, san anko, and Red

Dragon); it is mangan regardless of minipoints.

6.4. SCORING TABLES 161

6.4 Scoring tables

Table 6.10: Scores for non-dealer

Minipoints 1 han 2 han 3 han 4 han

20 — — — —

— (400-700) (700-1300) (1300-2600)

25 — 1600 3200 6400

— — (800-1600) (1600-3200)

30 1000 2000 3900 7700

(300-500) (500-1000) (1000-2000) (2000-3900)

40 1300 2600 5200 8000

(400-700) (700-1300) (1300-2600) (2000-4000)

50 1600 3200 6400 8000

(400-800) (800-1600) (1600-3200) (2000-4000)

60 2000 3900 7700 8000

(500-1000) (1000-2000) (2000-3900) (2000-4000)

70 2300 4500 8000 8000

(600-1200) (1200-2300) (2000-4000) (2000-4000)

80 2600 5200 8000 8000

(700-1300) (1300-2600) (2000-4000) (2000-4000)

90 2900 5800 8000 8000

(800-1500) (1500-2900) (2000-4000) (2000-4000)

100 3200 6400 8000 8000

(800-1600) (1600-3200) (2000-4000) (2000-4000)

110 3600 7100 8000 8000

— (1800-3600) (2000-4000) (2000-4000)

Note: Numbers in parentheses are tsumo scores.

162 CHAPTER 6. SCORING

Table 6.11: Scores for dealer

Minipoints 1 han 2 han 3 han 4 han

20 — — — —

— (700) (1300) (2600)

25 — 2400 4800 9600

— — (1600) (3200)

30 1500 2900 5800 11600

(500) (1000) (2000) (3900)

40 2000 3900 7700 12000

(700) (1300) (2600) (4000)

50 2400 4800 9600 12000

(800) (1600) (3200) (4000)

60 2900 5800 11600 12000

(1000) (2000) (3900) (4000)

70 3400 6800 12000 12000

(1200) (2300) (4000) (4000)

80 3900 7700 12000 12000

(1200) (2300) (4000) (4000)

90 4400 8700 12000 12000

(1500) (2900) (4000) (4000)

100 4800 9600 12000 12000

(1600) (3200) (4000) (4000)

110 5300 10600 12000 12000

— (3600) (4000) (4000)

Note: Numbers in parentheses are tsumo scores.

Chapter 7

Riichi judgement

7.1 To riichi or not to riichi?

Riichi is a really powerful tool in riichi mahjong. Once you riichi,

the opponents would have to slow down t heir attacks or even com-

pletely fold to avoid dealing into your hand. Therefore, one of our

top priorities in playing riichi mahjong is to try to make the hand

ready as fast as possible and call riichi before anyone else does.

At the same time, however, there are situations where you should

keep dama (i.e., not calling riichi when having a closed ready hand).

If you have played mahjong long enough, you must have come across

many instances where you wondered if you should call riichi or keep

dama. Knowing when to call riichi is one of the most fundamental

elements of mahjong strategies, yet it appears this is not very well

understood among European players. Let’s ﬁrst review the pros and

cons of calling riichi.

Obvious downsides of calling riichi can be summarized as fol-

lows.

Demerits of riichi

• You have to pay 1000 points as a riichi bet.

• The opponents may play defense and may not discard your

winning tiles that they would otherwise discard.

• You cannot change your hand any more; you cannot play

defense nor improve the wait / scores.

However, there are many more upsides as well.

164 CHAPTER 7. RIICHI JUDGEMENT

Merits of riichi

• You get one yaku to win.

• You can expect to increase the score with ippatsu and ura

dora.

• The opponents may completely fold or play more defen-

sively than otherwise. As a result, the opponents may fail

to make their hand ready, in which case:

– you will have more opportunities to draw tiles;

– you are less likely to deal into the opponents.

• When your previous discards make it look that your win-

ning tile is safe, riichi may actually increase t he chance that

the opponents discard your winning tile.

Comparing these pros and cons, it should be evident that calling ri-

ichi is a low-risk & high-return oﬀense tactic. Moreover, riichi could

also work as a defense tactic. If you riichi before the opponents do,

it may prevent them from building a ready hand, which obviously

reduces your chance of dealing into their hand. Although calling ri-

ichi means that you can no longer play defensively by choosing safe

tiles to discard, it poses less of a problem if your defensive skills are

not very good.

Riichi judgement criteria I recommend are summarized as fol-

lows.

7.1. TO RIICHI OR NOT TO RIICHI? 165

Riichi judgement

Choose riichi over dama if at least one of the following three con-

ditions is met.

1. Your hand has at least one han other than riichi.

2. Your hand has a good wait.

3. You are the dealer.

This means that you should call riichi if

• your hand has a bad wait but has one han or more (including

dora) other than riichi;

• your hand is riichi-only but with a good wait, or;

• your hand is riichi-only with a bad wait, but you are the dealer.

In other words, the only type of riichi that these criteria prohibit is a

bad-wait riichi-only hand by a non-dealer.

166 CHAPTER 7. RIICHI JUDGEMENT

7.2 Insta-riichi

Keep in mind that, when you call riichi you should do so immedi-

ately when your hand becomes ready (insta-riichi). There is usually

no point in waiting for a few turns to have the “right” moment to

riichi.

Let me describe a few frequently seen examples where you should

do insta-riichi. In all the examples that follow, we assume that you

are the South player in the 6th turn in East-1.

7.2.1 Examples of insta-riichi

Pinfu-only hand

ﬀﬁﬂ‚‹›„„ˆ˜˜˚ˇ˘ "

Dora

You should do insta-riichi with a pinfu-only hand. It is true that this

hand will have tanyao if you draw ȷ and discard ﬂ, but waiting for

that to happen is simply ineﬃcient. Even after you replace ﬂ with

ȷ, you will lose tanyao anyway if you win the hand on ´. Getting

either ippatsu or one ura dora has a much higher probability than

drawing ȷ ﬁrst and then winning on ¨.

⇒ Insta-riichi, discarding ˜ !

Bad wait, one dora

‰ıȷﬁ”„«´´"""! «

Dora

Since this hand has one dora, you should do insta-riichi. Do not shy

away from riichi even with closed-wait or edge-wait hands. It is true

7.2. INSTA-RIICHI 167

that the wait can be improved with as many as four kinds of tiles

(—‰ı´), but drawing one of those would take about eight more

turns, on average.

Since this is a yaku-less hand, you cannot win

it by ron while waiting in dama. Moreover, even when the wait gets

improved, this hand will never become pinfu anyway, so the score

will not be improved.

⇒ Insta-riichi, discarding ! !

1-away from sanshoku

Red

‰»ȷﬁ“”„ˇ˘¯""ﬄ «

Dora

It is true that there are some tiles that can improve the scores and/or

the wait of this hand. For example, if you draw ˚, the hand will

have sanshoku (Mixed Triple Chow). If you draw any of —‰ı, the

hand will have pinfu. However, since the hand already has one han

(red ﬁve), you should do insta-riichi.

⇒ Insta-riichi, discarding ﬄ !

Unconﬁrmed sanshoku

‰ﬂﬂ¸˛‚‹›“”ˆ˜¨ "

Dora

You would want to win this hand on — (rather than on ı) so that

you can claim sanshoku. However, waiting for — without riichi is

absurd. The worst case scenario is to draw ı while waiting in dama,

This rough calculation is based on an assumption that the probability of draw-

ing an arbitrary tile is

= about 3%.

168 CHAPTER 7. RIICHI JUDGEMENT

in which case you only get 400-700. If you riichi and draw ı, you

will get at least 700-1300. With one ura dora or ippatsu you will get

1300-2600.

⇒ Insta-riichi, discarding ¸ !

Riichi-only hand

ﬀﬁﬂ‚‹›ˆ˜˚ˇ˘¯## "

Dora

This hand has no yaku or dora, but the wait is good. You can do

insta-riichi with a riichi-only hand as long as the hand has a good

wait.

⇒ Insta-riichi, discarding ¯ !

Waiting for dora

—‰›“”„„„ˆ˜¨˚ˇ˘ 

Dora

You should do insta-riichi even when waiting for dora.

⇒ Insta-riichi, discarding „ !

When you wonder whether or not you should riichi in a given situa-

tion, choose riichi. You will be correct most of the time.

7.2.2 Good wait vs. high scores

We have discussed in previous chapters the diﬃcult tradeoﬀ we

face between speed (tile eﬃciency) and high scores. In riichi judge-

ment, this tradeoﬀ manifests itself as a choice between (a) having a

7.2. INSTA-RIICHI 169

good wait with lower scores and (b) pursuing higher scores with a

bad wait.

In the following examples, there are more than one discard candi-

dates to make the hand ready. I will discuss how to take a balance of

tile eﬃciency and hand value in calling riichi. Again, we will assume

that you are the South player in the 6th turn in East-1.

Cheap hand

ﬀﬁﬂ‚‹›ˆ˜˜˚ˇ˘## "

Dora

Calling riichi by discarding ˜makes for a good wait (´-¨; 2 kinds–

8 tiles), but the hand becomes riichi-only. On the other hand, calling

riichi by discarding ˆ leaves the possibility that the hand has an ad-

ditional yaku (White Dragon), although the wait is not as good (˜

#; 2 kinds–4 tiles). Which one should we choose?

In cases like this, you should aim for high scores by discarding

ˆ. Discarding ˆ and winning on # would double the hand value

(riichi + White Dragon = 2600 versus riichi-only = 1300), but the prob-

ability of winning on # would not shrink below half of the proba-

bility of winning on ´-¨.

⇒ Insta-riichi, discarding ˆ !

Here is a simple rule of thumb: when the minimum (guaranteed)

hand value is below 5200 (when won by ron), you should value scores

over wait. When the minimum hand value is 5200 or above, you

should value wait over scores. We use 5200 as a cut-point because

an additional han (roughly) doubles the hand value until it reaches

5200.

170 CHAPTER 7. RIICHI JUDGEMENT

More expensive hand

ﬀﬁﬂ‚‹›ˆ˜˜˚ˇ˘## ﬁ

Dora

This hand has one dora, but the minimum hand value is still below

5200 (riichi + dora = 2600). Therefore, again, you should value scores

over wait.

⇒ Insta-riichi, discarding ˆ !

Expensive hand

ﬀﬁﬂ‚‹›¨˝˝###$$ ﬂ

Dora

With this hand, calling riichi guarantees 5200 (riichi + White Dragon

+ one dora). Therefore, you should value wait over scores this time.

It goes without saying that riichi is better than going dama.

⇒ Insta-riichi, discarding ˝ !

Good wait vs. sanshoku

—‰ıˆ˜¨ˇˇ˛‚‹««« "

Dora

Calling riichi by discarding — gives you riichi-only with a good wait,

whereas calling riichi by discarding ı gives you riichi + sanshoku

with a bad wait. Since riichi-only is short of 5200, you should value

scores over wait.

⇒ Insta-riichi, discarding ı !

7.2. INSTA-RIICHI 171

Good wait vs. ittsu

–‰ııȷﬀﬁﬂ‚‹›«« "

Dora

Calling riichi by discarding – gives you only riichi + pinfu with a

good wait, whereas calling riichi by discarding ı gives you riichi

+ ittsu with a bad wait. Since riichi + pinfu is short of 5200, you

should value scores over wait.

⇒ Insta-riichi, discarding ı !

Expensive pinfu hand

Red Red

–‰»ȷﬀﬀﬁﬂ‚‹˙«« "

Dora

Discarding ﬀ would give you a dama mangan hand with a bad wait.

That is not too bad, but it is much better to riichi by discarding –

(riichi + pinfu + two red ﬁves = 7700 with a very good wait).

⇒ Insta-riichi, discarding – !

Let’s summarize what we have learned so far.

Good wait or high scores?

• Scores are more important than wait

when the minimum hand value is < 5200.

• Wait is more important than scores

when the minimum hand value is ≥ 5200.

172 CHAPTER 7. RIICHI JUDGEMENT

7.2.3 Chiitoitsu waiting for dora

We will discuss riichi criteria for a chiitoitsu (Seven Pairs) hand

in this and t he next sections.

Chiitoitsu 1

——¸¸‹««ˆˆ¨%%$$ ‹

Dora

You should not shy away from riichi with a chiitoitsu hand, es-

pecially when waiting for dora. Since dora of ‹ is something your

opponents wouldn’t lightly discard even if you keep dama, the chance

of winning this hand by ron is not very high anyway. Therefore, you

would rather riichi and aim to improve the score.

A two-dora chiitoitsu hand can be a game-deciding hand that

secures you the ﬁrst place in a game. If you tsumo after riichi, it is at

least haneman and it can easily be baiman with ura dora (ura dora

always come in pairs with a chiitoitsu hand). Even when you win

by ron, it will be haneman with either ura dora or ippatsu.

⇒ Insta-riichi, discarding ¨ !

Chiitoitsu 2

——¸¸««!ˆˆ¨%%$$ !

Dora

Dora in this example is a value-less wind tile, which may be easily

discarded by your opponents if you keep dama. Nevertheless, you

should still do insta-riichi by discarding ¨. Aim for haneman or

baiman rather than being content with 6400.

⇒ Insta-riichi, discarding ¨ !

7.2. INSTA-RIICHI 173

Riichi criteria for chiitoitsu hands waiting for dora are really sim-

ple.

Riichi judgement for chiitoitsu

Riichi any chiitoitsu hand if waiting for dora!

7.2.4 Chiitoitsu not waiting for dora

Riichi criteria for chiitoitsu hands get a bit more complicated

when you are not waiting for dora, summarized as follows.

Riichi judgement for chiitoitsu

Do insta-riichi with chiitoitsu (not waiting for dora) if one or

more of the following holds:

• You are the dealer;

• You have tanyao;

• You have one red ﬁve, waiting for a regular ﬁve;

• The wait is a suji-trap wait;

• The wait is any tile other than 4, 5, 6;

• The score without riichi is below mangan.

An example of a suji-trap wait is: you are waiting for a 3, and a 6 in the

same suit is among your discards. See Chapter 8.

This means that the only two cases where you should go dama are

(a) when you are a non-dealer and you are waiting for 4,5,6, and (b)

you have honitsu (Half Flush) or chinitsu (Full Flush) chiitoitsu.

It is also (theoretically) possible to have tsu iso (All Honors) chiitoitsu. Do

whatever you want with such a once-in-lifetime hand. I would riichi.

174 CHAPTER 7. RIICHI JUDGEMENT

The criteria do not change when your hand already has two dora and

waits for a non-dora tile.

Chiitoitsu 3

‰‰¸¸““ˆˆˆ!!$$ "

Dora

If you make the hand ready for chiitoitsu, you will be waiting for

, a non-4,5,6 tile. riichi is better than dama in this case.

⇒ Insta-riichi, discarding ˆ !

Chiitoitsu 4

‰‰ȷ““„„ˆˆˆ˝˝ˇˇ "

Dora

Waiting for a 6 is not ideal, but having tanyao justiﬁes riichi. Aim for

6400 ron or mangan tsumo.

⇒ Insta-riichi, discarding ˆ !

Chiitoitsu 5

Red

——¸¸‹˙««ˆˆ%%$$ "

Dora

Likewise, waiting for a 5 is not ideal, but having a red ﬁve justiﬁes

riichi.

⇒ Insta-riichi, discarding ‹ !

7.2. INSTA-RIICHI 175

Chiitoitsu 6

´´ˆˆ¨¨ˇ˘˘¯$$## "

Dora

Since you can get mangan ron or haneman tsumo without riichi, you

should keep dama with this hand (discard ˇ).

Chiitoitsu 7

‰‰ȷ““„„ˆˆ¨˝˝$$ "

Dora

With this hand, you should keep dama unless you are the dealer.

However, if you have already discarded  & ﬂ or ´ & ˇ, mak-

ing for a suji-trap wait, you can call riichi. You can also call riichi if

your wait is “cheap” in the board (more on this in the next section).

Why should we refrain from calling riichi with a single wait of

4, 5, 6? Because of the diﬀerences in versatility,

some tiles make for

a better single wait than others. Speciﬁcally, valueless wind tiles are

the best candidate for a single wait, followed by value tiles, terminals

(1s and 9s), and then simple tiles. Among simple tiles, 2s and 8s are

better than 3s and 7s, and 4,5,6 tiles make for the worst kind of single

wait.

Since having 4,5,6 tiles is crucial to utilize red ﬁves, your oppo-

nents are not very likely to discard them. Moreover, single waits of 4,

5, 6 are less likely to become a suji-trap wait before or after calling

riichi, compared with single waits of 1,2,3,7,8,9. A single wait of 4

requires both 1 and 7 to be discarded to become a suji-trap wait. A

single wait of 1, on the other hand, only requires 4 to be discarded.

Recall our discussions of tile versatility on in Section 3.2.4.

176 CHAPTER 7. RIICHI JUDGEMENT

7.3 When not to riichi

Keeping a hand dama for no reason is one of the two biggest sins

in riichi mahjong (the other will be introduced in Chapter 9). Do not

ever do meaningless dama. To put it the other way around, it is OK

to keep dama if there is a reason to do so.

That said, there are not many instances where going dama is better

than calling riichi. It is thus useful to memorize all these exceptional

cases; then you should call riichi in all other instances. Here is a list

of ﬁve situations where dama is justiﬁable.

Reasons to keep dama

7.3.1 Bad wait and no dora 7.3.2 In t he lead

7.3.3 Genbutsu wait 7.3.4 Expensive hand

7.3.5 Many possibilities of improving the hand

7.3.1 Bad wait

It is OK to go dama if the wait of your hand is really bad, especially

when your hand has at least one yaku without riichi so you can win

it by ron. The question t hen is, what is a really bad wait? The answer

depends on three things:

1. the kinds and the number of winning tiles lef t;

2. whether your wait is “expensive” according to your reading of

the board; and

3. whether your wait is likely to appear safe in t he eyes of your

opponents.

7.3. WHEN NOT TO RIICHI 177

1. The number of winning tiles left

The ﬁrst factor to consider is the pure number of winning tiles

of your hand. The more tiles you can win on, the better the wait is.

Table 7.1 provides a list of representative waits (roughly) in the order

of desirability.

Table 7.1: Typical wait patterns

Name Example Wait Kinds & Number

side wait ‰ ˆ-˝ 2 kinds–8 tiles

semi side wait ˛˛‚‹ ˛-› 2 kinds–6 tiles

stretched single ˆ˜¨˝ ˆ-˝ 2 kinds–6 tiles

dual pon wait ——‹‹ —‹ 2 kinds–4 tiles

closed wait ˆ¨ ˜ 1 kind–4 tiles

edge wait –—  1 kind–4 tiles

single wait ˛ ˛ 1 kind–3 tiles

In general, wait is said to be good if there are at least two kinds

and more t han four tiles left to win on. Therefore, dual pon wait,

closed wait, edge wait, and single wait are generally considered to

be a bad wait.

In counting the kinds and the number of winning tiles for your

hand, keep in mind that you have to count the kinds and the number

of live tiles to win on. For example, if your opponents have already

discarded all the four tiles of ˆ somehow, a side wait of ˆ-˝ es-

sentially becomes an edge wait of ˝, leaving only 1 kind–4 tiles to

win on.

178 CHAPTER 7. RIICHI JUDGEMENT

2. Cheap / expensive waits

However, this is only a part of the picture. When judging whether

your wait is good enough, you should also take into account the sec-

ond factor; namely, whether or not your winning tiles are likely to

be used by other players in their hands. As we learned in discussing

chiitoitsu hands, middle simple tiles 4, 5, and 6 generally have a

high chance of being used by the opponents.

Judging whether certain tiles are likely to be used by the oppo-

nents also involves a bit of board reading. If your opponents have

already discarded a lot of tiles in souzu (bamboos), for example, we

say that souzu are “cheap” in the board. Cheap waits are good waits.

Suppose three or more of ¨ have already been discarded by your op-

ponents. In such situations, an edge wait of ˜ is not bad at all. This

is because the paucity of ¨ makes it rather diﬃcult for anyone to

utilize ˜ in their hand. There is also a good chance that ˜ tiles are

live in the wall. Even when an opponent draws ˜ after you riichi,

they will have diﬃculty utilizing it in their hand; they have to either

discard ˜ or completely fold.

Applying the same logic, we can see why honor tiles make for a

good wait. For example, suppose % had already been discarded and

you now have a pair of % in your hand. Then, the opponents will

have diﬃculty utilizing the last % when they draw it. In a situation

like this, a dual pon wait of % (and another tile) is almost as good as

side wait, despite the fact that the number of tiles to win on is smaller.

On the other hand, when tiles in one suit are not being discarded

as much as those in the other two suits, that suit is being “expensive”

in the board. For example, suppose one or more opponents are pur-

suing a ﬂush hand (i.e., honitsu / chinitsu) in souzu. Then, even a

7.3. WHEN NOT TO RIICHI 179

side wait of ˆ-˝ can be bad.

3. Trap waits

The third factor you may want to consider is whether your wait

would appear safe in the eyes of your opponents. For example, when

you have a closed wait of ˆ and you have already discarded ˝,

there is a good chance that your opponent is tricked into thinking

that ˆ is safe. This is called a suji-trap wait (see Chapter 8). For

another example, suppose someone has a concealed quad of ﬁwhen

you happen to have a dual pon wait of ﬂ and something else. Then,

the opponents may think that ﬂ may be safe to discard even when

it is not.

That being said, reading the board requires some experience, and

reading the opponents’ thought is even more diﬃcult. You may want

to concentrate more on advancing your own hand rather than spend-

ing too much time trying to read the board. Just keep in mind that

having a pair/closed/edge/single wait does not automatically mean

that the wait is bad. Here is a rule of thumb to simplify your decision

making.

A reason to keep dama: bad wait

• Call riichi if there are three or more winning tiles left in the

board.

• Go dama if there are only one or two simple tiles left to win

on.

When waiting for an honor/terminal tile, you can call riichi even when

only one tile is left in the board.

180 CHAPTER 7. RIICHI JUDGEMENT

7.3.2 In the lead

The second case where going dama may be preferred to riichi is

when you are ahead of the game by much, and you just want to pro-

ceed to the next hand or ﬁnish the game while keeping your leading

position. This is especially the case towards the end of the South

round. For example, let’s say you are in South-4 and the scores are

as follows:

East (you) 39000 South 22900

West 13000 North 25100

You are currently in a safe lead because even the second ranked

player (North) cannot defeat you with a mangan tsumo. You can se-

cure your position even if you deal into a mangan hand by the South

or the West players. However, if you riichi, t he North player can now

get the ﬁrst place with a mangan tsumo, and the South player can get

the ﬁrst place with a direct hit mangan ron from you. Do not run such

risks by calling riichi. Even when you get a ready hand with a really

good wait, you must go dama. For more discussions of what to do in

South-4, see Chapter 10.

7.3.3 Genbutsu wait

The third case where going dama is OK is when another player is

already getting a lot of attention from others (e.g., riichi, dora pon, or

honitsu) and one of your winning tiles is among his genbutsu tiles.

One player’s genbutsu tiles are all the tiles discarded by that player

and the tiles that are passed up by that player.

For example, suppose the dealer has the following hand and dis-

cards in East-1.

See Section 8.2.1 for a more detailed explanation.

7.3. WHEN NOT TO RIICHI 181

Discards

¨ˆ˛!‹˘

‹“˚ﬁ

ﬃﬃﬃﬃ K## ﬂEﬂ ﬄﬄG ´

Dora

This is a conﬁrmed 7700 hand (seat & prevailing wind + White Dragon),

and the hand value can easily go up to haneman (18000) or baiman

(24000).

In such a situation, everybody will be paying attention

to the dealer (as they should). Suppose further that you manage

to make your hand ready for pinfu, waiting for ˜-˚. Then, you

should keep the hand dama, as ˚ is one of the dealer’s genbutsu

tiles. There is a good chance that the other two players fold against

the dealer completely and try to discard nothing but his genbutsu

tiles.

Keep in mind, though, that there is a bad kind of attention as well.

For example, suppose someone is doing the following.

ﬃ @ı %%M ¯1¯ 6˛‚

Suppose further he has already discarded dora. Then, he is getting

a lot of attention, but no one really cares about him, let alone folds

against him. In such a situation, you should call riichi even when

your winning tiles are among his genbutsu. Punish a player who

makes bad calls like this.

The hand can have (a combination of) the following yaku in addition to

what’s already visible: toitoi (All Pungs), Green Dragon, Red Dragon, dora,

honitsu, honroutou (All Terminals and Honors), and shousangen (Little Three

Dragons). With this hand, the maximum possible hand value is sanbaiman

(36000).

182 CHAPTER 7. RIICHI JUDGEMENT

7.3.4 High scoring hand

When your hand is already expensive without riichi (minimum

of 7700 if playing wit h red ﬁves; 5200 if playing without red ﬁves), it

is OK to go dama. Let’s see a few examples.

Expensive hand 1

‰ı‚‚‚‹“„ˆ˜¨˚˚ ‚

Dora

You should keep this hand dama because the hand is already expen-

sive (tanyao + three dora = mangan) and the wait is not very good.

⇒ Keep dama, discarding „ !

Why should we discard „, not ‹ (which will make for a suji-trap

wait)? This is to leave the possibility of improving the wait. If you

draw or call pon on ˚ or the fourth ‚ after discarding „, the wait

gets signiﬁcantly better, as follows:

‰ı‚‚‚‹“ˆ˜¨˚˚ ‚

Dora

↓

‰ı‚‚‚‹ˆ˜¨ .˚˚

‰ı‚‹ˆ˜¨˚˚ ‚5‚

Note that, to justify dama your hand has to have at least 7700 when

won by ron. This means that (1) your hand has to have at least one

yaku (without it you cannot win by ron) and (2) its value is at least

7700 when winning on a tile that gives you the lowest possible score.

7.3. WHEN NOT TO RIICHI 183

Expensive hand 2

‰ı‚‚‚‹“„´ˆ˜˚˚ ‚

Dora

This hand does not have yaku and so you cannot win by ron without

riichi. You should do insta-riichi with t his hand by discarding ‹.

Expensive hand 3

‰‰ȷȷ‚‹›˜¨˝˚ˇ˘ "

Dora

This hand is worth 7700 when winning on ı (tanyao + pinfu +

sanshoku), but the hand is worth only 2000 if winning on — (tanyao

+ pinfu). You should therefore do insta-riichi by discarding ‰.

When you have an expensive hand, going dama is acceptable but

calling riichi is also an option, especially when you have a good wait.

Consider the following hand. Should we riichi?

Expensive hand 4

Red

‰ȷȷ˛‚‹¨`˚˚ˇ˘¯ ‚

Dora

Discarding ¯, we get a conﬁrmed 7700 hand without riichi. I al-

ready said above that it is acceptable to keep dama when the mini-

mum hand value is 7700. However, would calling riichi be even bet-

ter?

Riichi criteria for hands like this are as follows.

184 CHAPTER 7. RIICHI JUDGEMENT

Riichi judgement for an expensive hand

• Riichi if you are far behind from other players.

• Riichi if it is the 6th turn or before and the wait is 2-way

side wait or better.

• Riichi if it is the 10th turn or before and t he wait is 3-way

side wait or better.

• Don’t riichi if the minimum hand value is haneman or better.

7.3.5 Many possibilities of improving the hand

It is OK to keep dama when there are many possibilities of further

advancing your hand. Keep in mind, however, that it is rather rare

that waiting in dama is worthwhile; doing insta-riichi is still better

in most instances even when there are some possibilities of advanc-

ing your hand. It makes sense to wait in dama only when both of the

following two conditions are met.

A reason to keep dama: improving the hand

• It is still an early turn (8th turn or before);

• There are at least six kinds of tiles that can improve the

scores and/or the wait,

there is at least one kind of tiles that can improve the score

by at least three han in one step.

Keep in mind that waiting in dama becomes less and less desirable

towards the end of a hand. After passing the 9th turn (the midpoint

7.3. WHEN NOT TO RIICHI 185

of the middle row of your discards), you’d better call riichi even if the

second condition is met. Remember that the probability of drawing

a particular tile is very small (roughly 3%).

If you decide not to riichi, it often makes more sense to revert the

hand to 1-away rather than maintaining a ready hand. As we learned

in Chapter 3, a ready hand can accept fewer tiles than a 1-away hand

can. For example, consider the following hand.

–—ıﬀﬁﬂ”„«««˝ˇ

Since calling riichi by discarding ı gives you riichi-only with a bad

wait, you may want to refrain from riichi. However, if you discard

ı, the hand can be improved only if you draw ¨ or ˘. Moreover,

even when you luckily draw ¨, the hand is still only riichi + pinfu,

albeit with an improved wait.

A more sensible choice here is to discard ˝ and revert the hand

to 1-away, as follows.

–—ıﬀﬁﬂ”„«««ˇ

This is another example of golden 1-away. This 1-away hand is so

much better than the ready hand you’d get by discarding ı. Specif-

ically, there are four kinds of tiles that can improve the score by at

least three han.

• If you draw ¯, the hand is ready for junchan (Terminals in All

Sets) + sanshoku — mangan without riichi and haneman with ri-

As I mentioned before, when there are four kinds of tiles that can improve your

hand, it will take (on average) eight turns to draw one of them.

186 CHAPTER 7. RIICHI JUDGEMENT

ichi. You may want to go riichi in this case because your pre-

vious discard of ˝ makes for a suji trap wait, although going

dama is also OK.

• If you draw ˘, the hand is ready for pinfu + junchan + sanshoku

— haneman without riichi. You may still want to riichi. It would

be a shame to win this monster hand on ˚, but doing so with-

out riichi is the worst.

• If you draw ‰ or ȷ, the hand is ready for pinfu + ittsu —

7700 with riichi. You should deﬁnitely riichi. Going dama with

this hand is absurd.

For another (less exciting) example, consider the following hand.

‰ıȷ´˜ˇ˘¯¸¸›“” "

Dora

Suppose this is the 5th turn in a hand. The choice is between (1)

discarding  to make this hand ready or (2) discarding ´ to revert

the hand to 1-away.

It is OK to choose either of the two in this case, but there is one

thing you should not do. That is discarding  without calling riichi.

If, according to your reading of the board, a closed wait of ˆ is good

enough (e.g., none of ˆ has been discarded yet, but lots of other tiles

in souzu (bamboos) have been discarded), do insta-riichi. Waiting in

dama with a yaku-less hand is generally a bad move. If you discard

 just to keep the hand ready, the hand cannot be won by ron, and

it can be improved only if you draw ¨. It makes no sense to have

7.3. WHEN NOT TO RIICHI 187

such a hand.

If you want to wait and improve the hand, you should discard

´ to revert the hand to 1-away. If you draw any of —‰ıﬀ¨ (5

kinds–15 tiles), both the wait and the scores get improved. If you

draw any of ȷ˜¸ (4 kinds–11 tiles), at least the wait gets im-

proved. A basic rule of thumb with a yaku-less hand is as follows.

What to do with a yaku-less hand

• Riichi if you make the hand ready.

• Don’t make it ready if you don’t riichi.

Yaku-less dama

In some exceptional situations, it may make sense to wait in dama

while keeping a yaku-less ready hand.

‰ıﬁ‚‹›«´ˆ˜˚ˇ˘ ˜

Dora

This hand has one dora with a bad wait (single wait of « or ﬁ).

According to the riichi criteria given at the beginning of this chapter,

you could call riichi with this hand. In fact, if it is already the 9th turn

or later, you should deﬁnitely do insta-riichi. However, if it is the 6th

turn or before, waiting in dama by discarding « or ﬁ is an option.

If you draw any of ı‚›´˚˘ (7 kinds–21 tiles), the hand be-

comes pinfu with a semi side wait. Moreover, with —ȷ˛“¨˝¯

Of course, if this were towards the very end of a hand (15th–18th turn), it

would make a lot of sense to have a yaku-less dama hand in order to avoid noten

penalty.

188 CHAPTER 7. RIICHI JUDGEMENT

(7 kinds–28 tiles) the wait will become a stretched single wait. When

there are this many possibilities of improving a hand (14 kinds–49

tiles), it is OK to go dama with a yaku-less hand.

7.4 Glossary

Insta-riichi is to riichi immediately when a hand becomes ready rather

than wait for a few turns to riichi. Basically, all riichi should be

insta-riichi.

Dama is not to riichi when having a closed ready hand. See Section

7.3 for cases where going dama might be better than riichi.

Yaku-less dama is when you have a ready hand with no yaku and

choose not to riichi. There are very few instances where doing

so is justiﬁable.

Genbutsu are tiles that are safe for a given player, either because they

were discarded by that player themselves or because they are

discarded by other players after that player called riichi. See

Section 8.2.1 in Chapter 8.

Chapter 8

Defense judgement

8.1 To push or to fold?

Knowing when to push and when to fold is another important

element of mahjong strategies. Push–fold judgement is a lot more

complicated than riichi judgement covered in the previous chapter.

In presenting defense strategies, I will ﬁrst describe a very simple

principle that tells you when to be defensive and when to be oﬀen-

sive, based purely on your hand. After understanding this principle,

the next step is to understand how to be defensive. The latter part of

this chapter introduces a set of defensive techniques.

8.1.1 A simple principle

A lot of variables can factor into our decision to push or to fold

against the opponents. You may want to consider, among other things,

whether or not you currently have a ready hand, the potential hand

value of your hand, the likely hand value of an opponent’s hand,

your current rank in the game, the opponent’s standing in the game,

just to name a few.

It is simply impossible to take into account these and other impor-

tant factors all at once in a limited amount of time. Instead, I suggest

you utilize the following shortcut for push/fold judgement.

190 CHAPTER 8. DEFENSE JUDGEMENT

Push/fold judgement

When another player has a ready hand,

Push if two of the following conditions are met:

1. Ready hand;

2. High scoring hand;

3. Good wait.

Fold if two of the following conditions are met:

1. 1-away (or worse) hand;

2. Low scoring hand;

3. Bad wait.

Let me explain each component of this principle in turn.

8.1.2 Guessing if an opponent has a ready hand

First, you need to guess if another player has a ready hand or not;

if your opponent does not have a ready hand, there is no point in

playing defensive. Of course, knowing whether an opponent has a

ready hand can be diﬃcult. Rather than spending too much time

trying to guess if they have a ready hand, let’s stick with rough but

simple shortcuts.

There are three possibilities to consider.

A. Riichi

This is the easiest case. You can be fairly certain that the op-

ponent has a ready hand. We will talk about how to defend

against riichi in Section 8.3.

B. Open ready hand

Knowing whether or not an opponent has an open ready hand

8.1. TO PUSH OR TO FOLD? 191

is a bit complicated. We will discuss this in Section 8.4.

C. Dama ready hand

We will completely ignore this case.

Assuming an opponent would not have a dama ready hand is ob-

viously not always correct. Nevertheless, this shortcut would be ac-

ceptable given that accurately guessing whether or not an opponent

has a ready hand is extremely diﬃcult. Part of t he reason why it is

OK to ignore the case of dama ready hand lies in the fact t hat riichi

is such a powerful tool in riichi mahjong that calling riichi is strictly

better than going dama in most instances; your opponents cannot win

a game if they keep choosing dama when they should call riichi (and

they are likely to know that).

8.1.3 Three conditions to push/fold

Note that, according to the principle laid out above, just because

(you think) an opponent has a ready hand, it does not automatically

mean that you must fold immediately. Speciﬁcally, you should still

push if two out of the three conditions speciﬁed above — ready, high

score, and good wait — are met.

The ﬁrst condition is fairly straightfor ward. Just remember that

a clear, ﬁrm line should be drawn between having a ready hand and

having a 1-away (or worse) hand. Pushing with a 1-away hand is

acceptable only when both of the other two conditions are met.

The second condition is also straightforward. We say a hand is

a high scoring one if the minimum hand value is 7700; otherwise it

is a low scoring hand (recall the discussion in Section 7.3.4 from the

previous chapter).

192 CHAPTER 8. DEFENSE JUDGEMENT

The third condition (good / bad waits) needs some explanation.

When you have a ready hand, this is straightforward. You can decide

if the third condition is met simply based on the waits classiﬁcation

discussed in Section 7.3.1 from the previous chapter. That is, wait is

good if it is at least as good as stretched single or semi side wait (2

kinds–6 tiles); dual pon wait, closed wait, edge wait, and single wait

are usually considered to be a bad wait.

The question then is: how do we judge if a hand has a good wait

when the hand is 1-away or worse? When a hand is 1-away or worse,

your judgement should be based on the best possible wait pattern you

can choose when the hand becomes ready in the worst possible man-

ner. An example will be helpful. Suppose you have the following

hand when another player calls riichi.

—‰ȷȷ““”¨˝$$$" $

Dora

This hand is not ready but has the minimum of 7700 point, so your

decision to push or fold depends on the third condition. Notice that

this is a perfect 1-away hand; no matter how this hand becomes ready,

you can always choose to have a side wait. In other words, the best

possible wait pattern you can choose when this hand becomes ready

is always a good one. If you draw or pon on ȷ or “ or chii ›-„,

you can have a side wait of ˜-˚. If you draw or chii ˜-˚, you

can have a side wait of ›-„. Therefore, you can push to the fullest

(zentsu) with this hand.

On the other hand, the following hand is also 1-away from ready

with a high scoring potential, but it will not always lead to a good-

wait ready hand.

8.2. DEFENSE BASICS 193

Red

‰‚‹˙“„«ˆ˜¨˚˚" ˜

Dora

Speciﬁcally, it will be a side-wait ready hand only if you draw ” ﬁrst.

If you draw —-ı ﬁrst (which will occur with a much higher proba-

bility than drawing ” will), it will be a closed-wait hand. Therefore,

the best possible wait pattern in the worst-case scenario is not a good

one. Therefore, you should fold with this hand when you are fore-

stalled by an opponent.

For another example, consider a chiitoitsu hand. A chiitoitsu

hand will always have a bad wait. This means that you should in

principle fold if an opponent calls riichi when your chiitoitsu hand

is not ready, even if you have two or more dora in your hand.

8.2 Defense basics

Once you understand the criteria to fold, the next thing you need

to know is how to fold. There are three main ways to identify safe

tiles to discard.

8.2.1 Genbutsu and other absolutely safe tiles

I introduced the term genbutsu in the previous chapter. Strictly

speaking, genbutsu tiles of player X refer to those tiles discarded by X

himself. However, if X has called riichi, then all the tiles discarded by

anyone after riichi (and passed up by X) are also called X’s genbutsu

tiles.

Genbutsu tiles of player X are 100% safe against X, but not nec-

essarily safe against the other two players. There are three kinds of

tiles that are 100% safe against all of the opponents.

194 CHAPTER 8. DEFENSE JUDGEMENT

• The tile that was just discarded by your left player.

• A fourth honor tile when there is no possibility of Thirteen Or-

phans.

• An absolute “no chance” tile.

The ﬁrst kind is fairly straightforward. Because of the furiten rule,

the tile just discarded by the left player is 100% safe for you to discard

in the present turn. That tile is not only genbutsu for the left player

but also a temporary genbustu for the right and the facing players.

Until their temporary furiten status is lifted, the right and the facing

players cannot call ron on it.

The second kind of absolutely safe tile is relatively simple. Sup-

pose all four of « are visible to you (among the discards, in your

hand, or as a dora indicator). Then, none of your opponents can win

Thirteen Orphans unless you discard «. In such situations, a fourth

honor tile is 100% safe. That is, ﬄ is 100% safe for everyone if all the

other three tiles of ﬄ are visible to you.

The third kind, absolute “no chance”, needs some explanation.

Let me just give you an example here. Suppose all four tiles of ˛, all

four tiles of ‹, and all three tiles of ‚ are visible to you. Then, the

fourth ‚ is 100% safe for everyone because this tile cannot be a part

of any set, run, or pair. I will explain more about “no chance” tiles in

Section 8.2.2 of this chapter.

Of course, it is not always possible to ﬁnd tiles that are 100% safe

for the player who has called riichi (let alone for all three opponents).

Therefore, we need to know how to identify relatively safe tiles by

relying on suji and kabe (blockade) theories. I will introduce t hese

two theories in turn.

8.2. DEFENSE BASICS 195

8.2.2 Understanding suji

When someone calls riichi, the possibility you need to be wary

of ﬁrst and foremost is that the opponent has a side-wait hand. It is

true that players will call riichi even when their wait is worse than

side wait. However, according to some statistics, about two thirds

of riichi hands have a side wait or better. This is partly because the

likelihood of choosing dama increases when the wait is bad. Another

reason is that players seek to retain side-wait protoruns over closed-

or edge-wait protoruns when choosing tile blocks to maximize tile

eﬃciency.

Suji defense is a defense tactic to avoid dealing into a side-wait

hand. A suji is a three-tile interval that corresponds to the wait of a

side-wait hand. For example, when a hand has a side-wait protorun

—, the wait is – or ‰. This wait combination of – and ‰ is

called –-‰ suji. There are 6 suji in each suit, giving rise to 18 suji

in total. All the six suji and their corresponding side-wait protoruns

are summarized in Table 8.1.

Table 8.1: Six suji

suji protorun suji protorun

1-4 suji — 4-7 suji ıȷ

2-5 suji ‰ 5-8 suji ȷﬀ

3-6 suji ‰ı 6-9 suji ﬀﬁ

When ‰is among a player’s genbutsu, we say –and ﬀ are suji

tiles. Suji tiles are safer than non-suji tiles because the furiten rule

prohibits a player from calling ron on – when ‰ is his genbutsu and

his wait is –-‰. Likewise, when ı is among a player’s genbutsu,

— and ﬁ are suji tiles and thus they are safer than non- suji tiles;

196 CHAPTER 8. DEFENSE JUDGEMENT

Table 8.2: suji tiles

Genbutsu suji tiles

‰ discarded – and ﬀ become safer

ı discarded — and ﬁ become safer

ȷ discarded  and ﬂ become safer

– and ﬀ discarded ‰ becomes safer

— and ﬁ discarded ı becomes safer

 and ﬂ discarded ȷ becomes safer

when ȷ is among a player’s genbutsu,  and ﬂ are suji tiles and

thus they are safer than non-suji tiles.

Although ‰ makes – a suji tile, the opposite is not true; – in

itself does not make ‰ a suji tile. ‰ is no safer than other tiles just

because – is among genbutsu. What – negates is the possibility

of the –-‰ side wait, but the ‰-ﬀ side wait is still a possibility.

‰ becomes safer only when both – and ﬀ are among a player’s

genbutsu. Table 8.2 summarizes combinations of genbutsu tiles and

tiles that are made safer by them.

Suji trap

Keep in mind that suji defense works only against side-wait hands.

Since players will call riichi even when their wait is worse than side

wait, we cannot rely too much on suji. When you wait for a tile

that is a suji tile of some tiles you have discarded yourself, we say

you have a suji-trap wait. In particular, when your wait is a suji

tile of the riichi declaration tile, we say it is an immediate suji-trap

wait. An immediate suji-trap riichi is a rather common occurrence

in riichi mahjong primarily because of double closed shape (e.g., 135,

246, 357, etc.). Consider the following hand.

8.2. DEFENSE BASICS 197

—‰ȷﬂﬂﬂ““˜¨˝$$$ ˜

Dora

If you call riichi by discarding the ȷ, the hand waits for , which

is a suji tile of ȷ (immediate suji-trap riichi).

In general, the reliability of suji is higher for tiles discarded ear-

lier in a hand. That is, suji tiles of early discards tend to be safer,

whereas suji tiles of those tiles that are discarded later are more

dangerous. In particular, suji tiles of the tile discarded upon riichi

is at least as dangerous as non-suji tiles.

For example, suppose an opponent calls riichi in the 7th turn with

the following discards.

"$ﬂ‰˛“-

riichi

There are three tiles in the discard that create suji tiles.

• ‰ makes – and ﬀ suji tiles.

• “ makes ‚ and « suji tiles.

• ˝ makes ˆ and ˘ suji tiles.

You might wonder about the reliability of suji for those tiles discarded after

riichi. There is in fact a big disagreement among professional players about

whether those tiles discarded after riichi make for safe or dangerous suji.

For example, (Nanba Shibukawa; NPM) argues that suji tiles of

those discarded after riichi are more dangerous than suji tiles of those dis-

carded before riichi, whereas (Nobuhiro Ishibashi; Saikouisen) ar-

gues the exact opposite. However, both schools of thought agree that the very

tile discarded upon riichi makes for dangerous suji. See http://osamuko.

com/identifying-dangerous-suji/ for some data analyses.

198 CHAPTER 8. DEFENSE JUDGEMENT

Note that ﬂand ˛do not create any suji tiles on their own. Among

these three tiles ‰“˝, suji tiles of ‰ are relatively safe, whereas

suji

tiles of

are rather dangerous. The reason why

suji

tiles of

˝ are dangerous is that players tend to keep double closed shape

such as ˝ˇ¯ or ´˜˝ until the hand becomes ready. Consider

the following 1-away hand.

Red

–––ȷﬀﬁ¸¸˙““´˜˝

We would discard “ rather than ˝. Then, if we draw ‹-” ﬁrst,

we do insta-riichi by discarding ˝, creating an immediate suji-trap

wait.

3. Understanding tile blockade (kabe)

Another defense tactic to identify safer tiles is to utilize a tile

blockade (kabe; wall). When chunks of a number tile are visible to

you, we say these tiles form a blockade; they block a formation of

runs that contain that tile. Suppose all four of — are visible to you,

either because they have been discarded, they are in your hand, or

they are used as a dora indicator. Then, none of your opponents can

have –-‰ suji wait, making – relatively safe. This is because no

one can have a protorun — in such a situation.

No chance

When all four of a number tile are visible, we say we have a “no

chance” situation, meaning that there is no chance that an opponent

has a suji wait that contains the tile that forms a blockade. In the

example above, – is a no-chance tile thanks to a blockade of —.

8.2. DEFENSE BASICS 199

No-chance tiles are safer than a non-suji tile, but keep in mind

that dual pon wait and single wait of a no-chance tile is still possible.

A nice thing about no-chance tiles is that t heir safety does not depend

on whether it is a suji tile or not. For example, when we have a block-

ade of —, – is safe regardless of whether ‰ is among genbutsu.

Note also that ‰ is not necessarily safe just because we have a block-

ade of —, for ‰-ﬀ suji wait is still a possibility. Of course, when

all four of — are visible to you and ﬀ is among a player’s genbutsu,

then ‰ becomes safer for that player.

Table 8.3 summarizes possible blockades

Blockade Safe tiles

– None

— –

 –—

‰ —

ı ﬀ

ȷ ﬀﬁ

ﬀ ﬁﬂ

ﬁ ﬂ

ﬂ None

Table 8.3: Blockades

and the resulting no-chance safe tiles. No-

tice that each blockade can produce at most

two sets of safe tiles. It should be easy to

see how a blockade of 1 does not make any

tile safer, a blockade of 2 makes 1 safer, and

that a blockade of 3 makes 1 and 2 safer.

However, a blockade of 4 makes only 2 and

3 safer, as it negates 2-5 and 3-6 waits that

contain 4, but it does not make 1 any safer.

Similarly, a blockade of 5 makes only 3 and

7 safer, negating 3-6 and 4-7 suji waits.

A blockade can negate non-suji waits as well. For example, if all

four of  and all four of ı are both visible, then a closed wait of ‰

is impossible. An opponent has to have a dual pon wait or a single

wait if he is to win on ‰. If, additionally, all three of ‰ are visible

to you as well, then the fourth ‰ is 100% safe. Waiting for ‰ in this

situation is simply impossible.

A blockade can also negate certain yaku, which decreases the

chance that an opponent has an expensive hand. For example, when

200 CHAPTER 8. DEFENSE JUDGEMENT

all four of  are visible to you. Then an opponent cannot have ittsu

(Pure Straight) in manzu (cracks). This information can help us de-

cide whether to discard a non-

suji

ﬁ

or a non-

suji

. The chance

of dealing into an opponent’s hand is equal, but the chance of deal-

ing into an expensive hand is lower with ﬁ. A blockade of  also

negates sanshoku of 123, 234, or 345. This information can help us de-

cide whether to discard a non-suji ˆ or a non-suji ˘. Again, the

chance of dealing into an opponent’s hand is equal, but the chance

of dealing into an expensive hand is lower with ˆ.

One chance

When only three of a number tile are visible to you, we have an

incomplete blockade, making for what’s called “one chance” tiles.

One-chance tiles are generally safer than non-suji tiles, but not as

safe as no-chance tiles. The reliability of incomplete blockades de-

pends on two things.

First, relying on an incomplete blockade is eﬀective in earlier turns

but not as much in later turns. Suppose that an opponent calls riichi,

and — is among his discards. Then, because other players are likely

to discard — if they have one, this tile may become an incomplete

blockade later on. However, an incomplete blockade formed this way

is not very reliable. When all three players have apparently folded

and the fourth — is still invisible to you, then it is highly likely that

the riichi’ed player has it. One-chance tiles would become almost as

dangerous as non-suji tiles in later turns in situations like this.

Second, one-chance tiles are more reliable when the incomplete

blockade that makes for a one-chance tile is known only to you, thanks

to a concealed set or a pair in your hand. On the other hand, one-

chance tiles t hat are created by an incomplete blockade in the discard

8.2. DEFENSE BASICS 201

pool are not particularly safe. This is because an opponent is more

likely to choose riichi over dama when one of his winning tiles is a

one-chance tile and appears safe.

When we have two incomplete blockade of consecutive number

tiles, we say they form a “double one chance” situation. For example,

if three of — and three of  are both visible, an opponent has to

have the fourth — and the fourth  to have –-‰ suji wait, which

is highly unlikely. Therefore, double-one-chance tiles are safer than

single one-chance tiles.

Tile blockade: Safety ranking

No chance > Double one chance > One chance (earlier turns) >

One chance (later turns) ≃ Non-suji

Combining blockade and suji

When we have a blockade of — and

Blockade Genbutsu Safe

— or  ﬀ ‰

 or ‰ ﬁ ı

‰ or ı ﬂ ȷ

ı or ȷ – ‰

ȷ or ﬀ — ı

ﬀ or ﬁ  ȷ

Table 8.4: Blockade and suji

ﬀ is among a player’s genbutsu, we

can deny not only –-‰ suji wait but

also ‰-ﬀ suji wait, making ‰ safe.

Combining the blockade and suji the-

ories like this might seem a bit compli-

cated at ﬁrst, but you will get used to it

as you play more games.

202 CHAPTER 8. DEFENSE JUDGEMENT

8.2.3 Safety ranking

Based on what we have learned so far, Table 8.5 below provides

a ranking of tile safety.

Table 8.5: Safety ranking

Rank

100% Genbutsu

AAA Fourth suji terminal; Fourth honor tile

AA Third suji terminal; Third honor tile

AA- Second suji terminal

A+ Second valueless wind tile; First suji terminal

A Second honor tile

BBB Suji 4,5,6; No-chance tile

BB+ Suji 2, 8

BB- Suji 3, 7; One-chance tile (earlier turns)

B First honor tile

CC Non-suji terminal

C One-chance tile (later turns)

DDD Non-suji 2,8

DD Non-suji 3,7

D Non-suji 4,5,6

There is not much diﬀerence between ranks if they are given t he

same alphabet. Tiles in the AAA ranking are dangerous only against

Thirteen Orphans. When Thirteen Orphans is not possible (i.e., there

are some terminals or honors that are already exhausted), they be-

come 100% safe. Fourth tiles mean that three of that tile have al-

ready been discarded. Likewise, third, second, and ﬁrst tiles mean

that two, one, or none of that tile have already been discarded, re-

spectively.

There is a diﬀerence between 4,5,6 tiles, 3,7 tiles, 2,8 tiles, and ter-

minals (1,9) because of the diﬀerence in versatility. Non-suji 4,5,6

tiles are the most dangerous because they can be caught by two dif-

8.2. DEFENSE BASICS 203

ferent suji waits. For example, 4 can be caught by a 1-4 suji and a

4-7 suji, making it doubly dangerous. 3,7 tiles are more dangerous

than 2,8 tiles because 3,7 can be caught by an edge wait, whereas 2,8

tiles cannot. terminals cannot be caught by either an edge wait or

closed wait. Suji 4,5,6 tiles are safer than suji 2,8 tiles because 4,5,6

make for bad candidates for dual pon wait or single wait.

204 CHAPTER 8. DEFENSE JUDGEMENT

8.3 Defense against riichi

Putting together what we have learned so far, the defense strategy

against an opponent’s riichi can be summarized as follows.

Defense against riichi

• Do not discard Rank D tiles against an opponent’s riichi

until your hand becomes ready

(unless your hand has a really good wait and a really high

score).

• If you need to push when your hand is 1-away from ready,

you can discard R ank C or safer tiles. Only if your hand

has a guaranteed mangan, you can discard Rank D tiles.

• If you need to push when your hand is 2-away from ready

or worse, you can discard Rank B or safer tiles.

• If you cannot satisfy the above criteria, you must completely

fold (betaori).

8.3.1 What to discard when you get stuck

When you cannot identify safe tiles at all, rely on the following

and try to be as safe as possible.

Tile chunks

Discard pairs and concealed sets. Once you get one tile passed

against a riichi’ed player, you can be safe for the next turn or two.

Avoid dealing into expensive hands

If you discard terminals, you can avoid dealing into a tanyao (All

Simples) hand. Also, try not to discard the dora indicator tile (when

8.3. DEFENSE AGAINST RIICHI 205

dora is a number tile) and any tiles close to dora, as well as the dora

tile itself.

Tiles outside early discards

Tiles that are outside those discarded in “early” turns are rela-

tively safe. Consider the following riichi.

"—$¯‚ˇD

riichi

This opponent discarded — in the 2nd turn, which is relatively early.

This suggests that he is not very likely to have –-‰ suji wait. If

he had a tile block ——, he would probably have kept it and dis-

carded something else. This line of argument is obviously not 100%

reliable. However, if you compare ﬂ and – in the current example,

–

is relatively safe.

8.3.2 Miscellaneous

Here are some additional factors you may want to take into ac-

count when deciding whether or not to be defensive, and how much

defensive you should be.

Your position in the game

You should be more defensive when you are ahead of the game,

while you should be more aggressive when you are behind. This

should especially be the case in the South round.

Turn

You can be more aggressive in earlier turns, whereas you need

to be much more defensive towards the end of a hand. Suppose

an opponent calls riichi in the 3rd turn, and your hand is 1-away

206 CHAPTER 8. DEFENSE JUDGEMENT

from ready. Since you have 15 more turns to draw, you still have a

good chance of making the hand ready. In such situations, it may

be worthwhile to be a little bit aggressive against riichi. However, if

you have only three more turns to draw (i.e., in the 15th turn) and

your hand is still 1-away from ready, the chance of making a ready

hand before the hand ends is very low. It is not worthwhile to discard

dangerous tiles at this point.

Moreover, in earlier turns, there are many suji that are “alive”,

which diminishes the probability of dealing into an opponent’s side-

wait hands. For example, suppose an opponent calls riichi with the

following discard.

"$ﬂ*

riichi

So far, only 2 out of 18 suji have been denied by the discards (ȷ-

ﬂ and ˆ-˝), leaving 16 suji alive. Suppose you are considering

whether to discard ¸. Assuming that the riichi’ed player has a side-

wait hand, the conditional probability of dealing into his hand with

¸ is only

at this point.

However, as the hand proceeds, the number of live suji waits

will decrease, making it more and more dangerous to discard a non-

suji tile. Suppose that the hand proceeds and the riichi’ed player’s

discard is as follows.

Since the riichi’ed player may not have a side-wait hand, the joint probability

that this is a side-wait hand and the hand waits for ¸ is even lower than

. The

total probability that the riichi’ed player is waiting for ¸ is a bit greater than

this joint probability because of the possibility of dual pon wait and single wait.

8.3. DEFENSE AGAINST RIICHI 207

"$ﬂ*›

¨‰“˚—”

ﬁ

Since as many as 16 suji waits have already been denied, t here are

only 2 suji waits that are “alive” (¸-‹ and ˝-˘). Then, the con-

ditional probability of dealing into his hand with ¸given that he has

a side-wait hand is now as high as 50%. This gives us an additional

reason to be more defensive towards the end of a hand.

Opponent’s style

If you know the type of opponent you are facing, you may want to

take that into account. For example, if you know that your opponent

is an old-fashioned player who calls riichi only when they have a

good-wait hand, you can rely heavily on suji theories.

However, if you know that your opponent understands the mod-

ern riichi strategies as described in the previous chapter, it is more

diﬃcult for you to guess whether he has a good wait or a bad wait.

This is because he would not shy away from riichi even with a bad-

wait hand. You cannot rely too much on suji theories in such situa-

tions.

208 CHAPTER 8. DEFENSE JUDGEMENT

8.4 Defense against open hands

8.4.1 Guessing if an opponent has a ready hand

To defend against an open hand, we ﬁrst need to know if an op-

ponent has a ready hand or not. Again, we will use some simple

shortcuts, which hopefully lead us to the right conclusion most (if

not all) of the time.

Defense against open hands

Assume an opponent has a ready hand in any of the following

situations.

1. He has three or more open sets / runs.

2. When he is doing a ﬂush hand, he starts discarding tiles in

the suit he is supposedly collecting.

3. He keeps discarding the tile that he draws.

8.4.2 Estimating the value of an opponent ’s hand

The next thing you need to know is how expensive an opponent’s

hand is. Although it is practically impossible to infer riichi’ed player’s

hand value, we can often estimate the hand value of an opponent’s

open hand. If you can easily see that an opponent’s hand is tanyao-

only or fanpai-only, there is no need to be defensive.

Open hands with dora

An obvious case of an expensive open hand is one with dora tiles.

For example, if an opponent has an open set of dora, clearly he has

8.4. DEFENSE AGAINST OPEN HANDS 209

a four-han (or higher) hand. Also, if you play with red ﬁves, scores

get expensive quite easily. For example, suppose the dealer has the

following open hand in East-1.

Red

ﬃﬃﬃﬃﬃﬃﬃ Gﬄﬄ ,`˚

Then, this hand has at least 5800 (seat and prevailing wind + red ﬁve).

Try not to push too hard against this player.

Flush hands

Flush hands (honitsu or chinitsu) tend to get expensive as well.

Red

ﬃﬃﬃﬃ )´´ ,˜` .ˇ˘

A hand like above has a minimum of 3900 and a maximum of haneman

if you deal into it with a souzu tile. You should fold when you draw

an unwanted souzu tile.

Value tiles

Open sets of value tiles also make for an expensive hand.

ﬃﬃﬃﬃ K## ﬂEﬂ ﬄﬄG ´

Dora

This is a pretty scary hand. You should not discard anything other

than genbutsu tiles.

210 CHAPTER 8. DEFENSE JUDGEMENT

Cheap hands

On the other hand, you can sometimes see that an opponent is

likely to have a cheap hand. For example, suppose an opponent is

doing the following.

ﬃﬃﬃﬃﬃﬃﬃ /˚˘ 7“” ´

Dora

We can see that honitsu and ittsu are impossible and sanshoku is

unlikely. If all of the dora tiles and red ﬁves are visible to you (in

your hand or among the discards), you can be pretty sure that this

hand is inexpensive. It is true t hat the following is still a possibility.

###$$$% /˚˘ 7“” ´

Dora

But, you may be able to rule this out if you check what value tiles are

still alive in the wall.

8.4.3 Identifying dangerous tiles against open hands

You can’t win your own hand if you completely fold ever y time

an opponent melds. Unless an opponent has an obviously expen-

sive hand (e.g., three open sets of value tiles, etc.), we would want

not to fold completely but to discard some tiles that are not particu-

larly dangerous. It is therefore important to identify dangerous tiles

against open hands.

For example, consider the following.

ﬃﬃﬃﬃ L$$ ﬂﬂE ‚5‚

8.4. DEFENSE AGAINST OPEN HANDS 211

This opponent probably has a toitoi (All Pungs) hand (otherwise, it

would be cheap so you can ignore it). If it is toitoi, suji theories and

blockade t heories are completely useless (remember, their purpose is

to avoid dealing into side-wait hands). Most dangerous tiles in this

situation are “raw” tiles (tiles that are completely invisible to you). In

particular, you should not discard raw value tiles. As single wait is

also a possibility, all honor tiles are generally dangerous (unless they

are the fourth tile).

For another example, consider the following.

ﬃﬃﬃﬃﬃﬃﬃ ﬂﬂE .ˇ˘ “

Dora

In a situation like this, one possibility is that the opponent has a dual

pon-wait hand with value tiles on the one hand and dora tiles on the

other, as follows.

““´ˆ˜$$ ﬂﬂE .ˇ˘ “

Dora

Especially when you are facing a “reliable” opponent, there would

be a good reason (such as having two dora tiles) why he rushed by

melding a side-wait protorun ﬁrst.

212 CHAPTER 8. DEFENSE JUDGEMENT

8.4.4 Discard upon chii

We can sometimes identify tiles that are relatively safe or rela-

tively dangerous against an open hand by paying attention to what

an opponent discarded upon calling the last chii or pon. Consider

the following t hree cases.

Case 1 chii → discard a tile in the same suit

An opponent who had an open set of $ called chii on ‚ and

discarded ‹.

ﬃﬃﬃﬃﬃﬃﬃ 5‹› L$$

and discarded ‹

In this case, it is unlikely that this opponent has a wait in the neigh-

borhood of ‹, so pinzu (dots) tiles such as ˛‚›“” are rela-

tively safe.

Case 2 chii → discard a tile in a diﬀerent suit

An opponent who had an open set of $ called chii on ‚ and

discarded ¨.

ﬃﬃﬃﬃﬃﬃﬃ 5‹› L$$

and discarded ¨

In this case, this opponent’s wait is very likely to be in the neigh-

borhood of the t he last discard, ¨. In particular, ˆ-˝ suji and

˜-˚ suji are extremely dangerous, and a closed wait of ˇ is also

8.4. DEFENSE AGAINST OPEN HANDS 213

a possibility.

Case 3 pon

An opponent who had an open set of $ called pon on ‚ and

discarded ¨.

ﬃﬃﬃﬃﬃﬃﬃ 5‚‚ L$$

and discarded ¨

In t his case, this opponent’s wait can be in any suit; we cannot iden-

tify which tiles are particularly safe or dangerous.

What makes these diﬀerences? These readings are based on an

assumption that the opponent has a good 1-away hand before calling

the last chii or pon. In Case 1, t he opponent has the following perfect

1-away hand before the last chii.

–—ﬁﬁ‹‹›˜¨ L$$

Then, after calling chii on ‚ the opponent discards ‹, making the

neighborhood of ‹ relatively safe.

On the other hand, in Case 2, the opponent has t he following

perfect 1-away hand before the last chii.

–—ﬁﬁ‹›˜¨¨ L$$

Then, after calling chii on ‚ the opponent discards ¨, making the

neighborhood of ¨ dangerous. In this particular case, the resulting

214 CHAPTER 8. DEFENSE JUDGEMENT

wait is ˆ-˝ suji. If the ˜¨¨ block were ¨¨˝ instead, the

resulting wait is ˜-˚ suji. If the ˜¨¨ block were ¨˚˘, the

resulting wait is

Finally, in Case 3, the opponent has the following perfect 1-away

hand before the last pon.

–—ﬀﬁ‚‚˜˜¨ L$$

Then, after calling pon on ‚ the opponent discards ¨, making the

wait unrelated to ¨. Notice that, if the opponent calls chii on ȷ

with this hand and discards ˜, the neighborhood of the last discard

becomes dangerous (just like Case 2). Similarly, if the opponent calls

chii on ˆ with this hand and discards ˜, the neighborhood of the

last discard becomes safe (just like Case 1).

Discard upon chii

• chii → discard a tile in the same suit

⇒ the neighborhood of the last discard is safe

• chii → discard a tile in a diﬀerent suit

⇒ the neighborhood of the last discard is dangerous

• pon

⇒ wait can be anything

8.5. GLOSSARY 215

8.5 Glossary

Zentsu is to push to the fullest, usually against an opponent’s riichi.

Betaori is to fold to the fullest.

Suji is a three-tile interval that corresponds to the wait of a side-

wait hand. There are six suji: 1-4, 2-5, 3-6, 4-7, 5-8, and 6-9.

See Section 8.2.2.

Suji tile is a tile that is made safe against side wait when a certain

tile(s) is among an opponent’s genbutsu. For example, when

¨ is among genbutsu, ´ and ˇ are safe against a side-wait

hand.

Suji trap is when the wait is a suji tile. When this happens, the

wait is either dual pon wait, closed wait, edge wait, or single

wait.

Blockade (kabe; wall) is formed when three or four of a number tile

are visible to you. When all four of a number tile are visible,

they form a complete blockade, making for no-chance tiles. When

three of a number tile are visible, they form an incomplete block-

ade, making for one-chance tiles.

No-chance tile is a tile that is made safe by a complete blockade.

There is “no chance” that an opponent has a protorun that in-

cludes a tile that is blocked.

One-chance tile is a tile that is made safe by an incomplete block-

ade. There is “one chance” that an opponent has a protorun

that includes a tile that is blocked.

Chapter 9

Melding judgement

9.1 To meld or not to meld?

Melding decisions — to call pon/chii or not to call — depend on

a lot of variables. The most important criteria of all are t he following

two.

When not to meld

Do not meld if one of the following two holds.

1. The hand is both cheap and far from ready.

2. Melding signiﬁcantly reduces the hand value.

I will discuss each of the two in tur n, and then discuss exceptional

situations that justify melding even when the two conditions above

are satisﬁed.

Throughout this chapter, I assume that you are playing with rule

sets that allow open tanyao. Although older EMA rules did not

allow it, revised EMA rules (eﬀective as of April, 2016) now allow

tanyao to be an open hand. I will also assume that you are the Sout h

player in the 6th turn in East-1 unless otherwise stated.

9.1.1 When not to meld 1: cheap and slow

Melding is acceptable only when your hand is either expensive or

fast. Melding with a cheap and slow hand is one of the two biggest

sins in riichi mahjong.

As we lear ned in Chapter 7, t he other big sin is meaningless dama.

9.1. TO MELD OR NOT TO MELD? 217

When we say a hand is “slow”, we mean that the hand is 2-away

or worse after melding (i.e., 3-away or worse before melding) and wit h

a bad wait. Let’s see a few examples.

Cheap and slow hand

ıﬀ‚‹““”«˜˝## –

Dora

An opponent discarded # just now.

You should not call pon in this situation. Even after calling pon, the

hand is still 2-away from ready with a couple of bad waits, as follows.

ıﬀ‚‹““”˜˝ K## –

Dora

The probability of winning this hand any time soon is not very high.

What if an opponent calls riichi now? You will have nothing but sim-

ple tiles between 3 and 7 to discard. It is not worthwhile to discard

such tiles against riichi when you have a cheap 2-away hand. When

your hand is cheap and slow, you should worry more about keeping

safe tiles such as # than about winning the hand.

Slow but expensive hand

ıﬀ‚‹““”«˜˝## “

Dora

An opponent discarded # just now.

This time, you can call pon on # then discard «. This hand is still

slow (after all, the hand shape is exactly the same as before), but it

has a potential to be 7700 (White Dragon + sanshoku + two dora)

even when you open it. When you see a high score potential, you

can meld even with a slow hand.

218 CHAPTER 9. MELDING JUDGEMENT

Cheap but fast hand

ıȷ‚‹““”«ˆ˜˜## ‰

Dora

An opponent discarded # just now.

Again, you can call pon on # then discard «. This hand is cheap

and it will be 2-away after the pon, but all the remaining blocks have

a good wait. When you can expect to win a hand in no time, you can

meld even with a cheap hand.

Melding judgement 1

Don’t meld with a cheap and slow hand!

9.1.2 When not to meld 2: big gap in hand values

You should also refrain from melding when doing so signiﬁcantly

reduces the hand value. More speciﬁcally, do not meld when the

hand value drops from 7700 or above to 2000 or below.

Cheap but fast hand

‰ıȷﬀﬀ˛‚‚‹‹¨˝˝ ˚

Dora

The left player discarded ˜ just now.

Do not meld with a hand like this, at least until the 13th turn or so

(the third row in the discard). Although calling chii on ˜ will make

the hand ready with a good wait, the hand value reduces to 1000. If

you keep the hand closed and call riichi, the hand can potentially be

When t he hand value reduces from haneman to mangan, or from baiman to

haneman, melding is acceptable. This is because mangan is already expensive

enough.

9.2. MELDING CHOICE: EXAMPLES 219

a game-deciding hand with a realistic possibility of getting mangan

or haneman.

Fast and expensive hand

Red

——»ȷﬀ››“˜¨¨˝˚ ¨

Dora

The left player discarded ” just now.

Calling chii on ” will make the hand ready with a good wait and a

high score (7700). It is true that this hand can be even more expensive

if you keep it closed. However, 7700 is already pretty expensive. An

additional han does not improve the hand value as much beyond 4

han. We should thus call chii on ”, especially after the 9th turn or

so.

Melding judgement 2

Don’t meld if melding signiﬁcantly reduces t he hand value!

9.2 Melding choice: examples

We will now see more examples of melding judgements, some of

which will describe an exception to the two conditions introduced so

far.

9.2.1 Eliminating bad waits

One of the purposes of melding is to eliminate a bad wait in a

hand to enhance speed. When you can call pon or chii to complete

a bad-wait block in your hand it often makes sense to do so. More

Riichi + tanyao + pinfu + iipeiko + dora = mangan ron or haneman

tsumo.

220 CHAPTER 9. MELDING JUDGEMENT

speciﬁcally, when you call chii with an edge-wait or closed-wait pro-

torun to make the hand ready, you should meld. Consider the fol-

lowing hand.

ﬁﬂ”„«´ˆˆ˝˚ˇ$$ !

Dora

The left player discarded ﬀ just now.

You should call chii on ﬀ and discard ´. It is true t hat doing so

means that t he hand value will be 1000 (Green Dragon only) and that

the hand can be won only with $. However, notice that the hand

value is not very high anyway if you keep the hand closed. Even if

you draw ﬀ and call riichi, the hand value is 2600 if you win on $

or 1300 if you win on ˆ.

Keep in mind also that winning a cheap hand like this is not to-

tally meaningless. This is because doing so also means you prevent

your opponents from winning their (possibly expensive) hands. You

do not want to make your mangan hand into a 1000 hand, but t he

hand above is not a mangan hand.

Moreover, this hand has pretty low tile acceptance (4 kinds–12

tiles, ﬀˆ˜$); the chance of making the hand ready without meld-

ing is not very high, either.

Melding judgement 3

If you can eliminate a bad wait and make the hand ready, meld!

9.2.2 Improving the wait

It sometimes makes sense to meld even when your hand is al-

ready ready, as long as doing so improves the wait and/or the scores.

9.2. MELDING CHOICE: EXAMPLES 221

It may also make sense to meld to make a bad-wait 1-away into a

good-wait 1-away one.

Improving the wait

—‰»ȷﬀ‚‹˙´´ %%M

The left player discarded ı just now.

The hand is already ready, waiting for . However, you should call

chii on ı with ȷﬀ and discard —, so you can upgrade the wait

to a side wait of -ȷ. With melded hands, it is important to think

about the possibilities of improving the wait and/or scores by meld-

ing further. In the current example, calling chii on ı-ﬁ or pon on

´ will improve the wait from a closed wait to a 2-way wait.

9.2.3 Conﬁrming yaku

Ideally, we would like to complete a bad-wait block by melding

so that we can have a good-wait block as the ﬁnal wait of the hand.

However, it sometimes makes sense to complete a good-wait block

by melding if doing so conﬁrms a certain yaku in a hand.

Conﬁrming yaku

—‰˛˛“”´ˆ`˚ˇ˘¯ ¯

Dora

The left player discarded ¨ just now.

Calling chii on ¨ completes a side-wait block in this hand, leaving

the hand 1-away with one edge-wait and one side-wait protoruns.

However, this is acceptable because calling chii on ¨conﬁrms ittsu

in this hand. Getting ittsu with this hand requires that we have ¨,

222 CHAPTER 9. MELDING JUDGEMENT

not ˇ, to complete t he protorun ˝˚. We should thus think of this

protorun more as an edge-wait protorun rather than a side-wait pro-

torun. Calling

chii

is tantamount to eliminating a bad wait in

this case.

Conﬁrming yaku

˛˛›“”„˜¨ﬄﬄﬄ $

Dora

The left player discarded ˛ just now.

If you call pon on ˛ and discard „, the hand is ready. However,

doing so only gives you a 1300 hand. Instead, you should discard ¨

after calling pon to have a 1-away honitsu hand, as follows.

›“”„˜ﬄﬄﬄ ˛4˛ $

Dora

You can further call chii on any of ‹“”« or call pon on to get

5200 or above.

9.2.4 Concealed set of value tiles

There are situations where melding with a cheap and slow hand

may be acceptable. Recall that one of the reasons why melding is not

worthwhile with a cheap and slow hand is that we will lose safety

tiles if we meld. When that is less of your concern, melding may be

an option even with a cheap and slow hand.

Note that you should discard the —, not the ‰, after melding. This is because

we will discard the “” block if we draw the red ».

9.2. MELDING CHOICE: EXAMPLES 223

Concealed set of value tiles

‰ȷﬁ˛‹´ˆ˝˚˚%%% !

Dora

The left player discarded ı just now.

This hand is both cheap and slow. Even after calling chii on ı, the

hand is still 1-away with a bad wait. However, this hand has a con-

cealed set of %, which can be used as three safe tiles when someone

calls riichi. In such cases, you can meld as long as doing so advances

the hand. That is, you can chii any of ıﬀ‚˜. You should not

call chii on ¨-ˇ or pon on ˚, because doing so does not advance

this hand from 2-away to 1-away or improve the wait/scores.

Melding judgement 4

If you have a concealed set of value tiles, you can meld with a

cheap and slow hand.

9.2.5 When it is OK to meld with cheap & slow hands

There are a few more instances where melding with a cheap and

slow hand is acceptable, summarized as follows.

1. You are ahead of t he game in South-4.

The hand value is not of your concern in such a situation. You

can meld with a cheap hand; you can also meld even when

melding signiﬁcantly reduces the hand value.

2. There are two or more riichi bets on the table.

Winning any hand guarantees a minimum score of 3300 points

in such a situation, as you get at least 1000 (your hand) + 2000

for riichi bets + 300 for continuation. This is not much diﬀerent

from winning a 3900 hand.

224 CHAPTER 9. MELDING JUDGEMENT

3. You are losing and you are the dealer.

You should aim for calling riichi as soon as possible in order

to delay the opponents’ attack. However, when you think you

cannot make the hand ready for riichi soon enough, calling pon

or chii early may serve the same purpose.

9.3. CALLING KAN 225

9.3 Calling kan

There are three ways to call kan (kong) — making a concealed

quad ( ankan), making an open quad (daiminkan), and extending an

open set to an open quad (kakan). I will discuss decision criteria for

each of the three cases in turn.

9.3.1 Concealed quad (ankan)

The beneﬁts of making a concealed quad includes:

• another chance to draw a tile;

• increased minipoints; and

• possibilities of getting more dora.

Conditions to justify making a concealed quad includes:

• 1-away, where at least one block has a good wait;

• 2-away, where all the blocks have a good wait;

• you need more dora or more minipoints to improve the place-

ment (especially in South-4);

• you are losing badly.

With this in mind, consider several examples. (Dora is ! in all

the examples.)

(1) —‰ȷ˛‚›“”ˆˆ""""

(2) ‰‰˛‚›“”ˆˆ""""

(3) ‰‰˛‚“””ˆˆ""""

226 CHAPTER 9. MELDING JUDGEMENT

(4) ı¸‚›“”ˆˆ""""#

(5) ı¸¸‚““”""""##

(1) This hand is 1-away, and one block has a good wait and another

has a bad wait. You can call kan.

(2) This is a perfect 1-away hand, so you can call kan.

(3) This is a perfect 2-away hand, so you can call kan.

(4) This hand is 1-away, but all the remaining blocks have a bad wait

(closed wait). You should not call kan.

(5) This hand is 2-away with a bad wait. You should not call kan

in normal situations. However, if you are in South-4, and you need

2000 points to win the game, t hen you should kan immediately. As

the hand will have at least 60 minipoints, you can get 2000 points

with one han (White Dragon).

Kan judgement 1

In principle, your hand needs to be close to ready to justify mak-

ing a concealed quad.

When not to make a concealed quad

Calling kan also comes with some cost, including:

• you may lose safety tiles;

• the new dora may go to the opponents.

When the following conditions are present, you should refrain from

calling kan.

• the hand is close to ready for chiitoitsu as well;

9.3. CALLING KAN 227

• one of the four tiles can be used a good ﬂoating tile;

• you lose some yaku if making the set into a quad.

Concealed quad?

‰‰‰‰ﬁﬁ‹›“”¨˝˚˘

It is your turn. What would you do?

If you call kan, the hand will be 1-away and the wait will not be ter-

ribly bad; it can be made ready if you draw any of

ﬁ˛‚‹›“”„«˜˚ˇ¯, and you may draw one of these

tiles as a rinshan tile (the bonus draw after kan). However, the result-

ing hand will be either riichi only or riichi + tanyao only, sometimes

with a bad wait.

If you choose not to call kan and discard ˘, you can treat one

of the four tiles of ‰ as a ﬂoating tile that could form a side-wait

protorun. The hand will be a side-wait ready hand if you draw any

of ı‚›“„. Moreover, if you draw ı or ȷ, the hand will be

ready for sanshoku of 456. If you call kan, on the other hand, the hand

will lose the ability to accept —ıȷ that would otherwise make

the hand ready. Therefore, you should not call kan at this point and

simply discard ˘. You can call kan later if the hand becomes ready

by drawing a pinzu (dots) tile.

Concealed quad?

‰‰‰‰ﬀﬁﬂ‹›´ˆ˜ˇ˘

It is your turn. What would you do?

If you call kan, you will lose pinfu. Moreover, if you draw a tile that

228 CHAPTER 9. MELDING JUDGEMENT

completes one of the two side-wait protoruns after calling kan, the

hand becomes a single-wait hand. You should thus discard ‰. Then,

if you complete one of the side-wait protoruns ﬁrst, you can discard

another ‰ to make the hand ready for pinfu.

Kan judgement 2

There are cases where we should not make a concealed quad

even when the hand is (close to) ready.

9.3.2 Open quad (daiminkan)

Conditions to justify making an open quad are a little bit more

demanding than the conditions to justify making a concealed quad.

You can make an open quad in any of the following situations:

• the hand is ready with a good wait, and the hand value is be-

tween 2000 and 5200 points;

• you need more dora or minipoints to improve the placement

(especially in South-4);

• you are losing badly.

Open quad?

ȷﬀ„„´ˆ˜%%% 5¸˛ ˆ

Dora

The right player discarded % just now.

This ready hand is currently worth 2000 points with a good wait.

Calling kan on % is therefore justiﬁable. If any one of the tiles in your

hand becomes dora, the hand value increases from 2000 to 5200. If

you get rinshan tsumo in addition, it will be mangan.

9.3. CALLING KAN 229

Open quad?

ﬁﬂ„„´ˆ˜%%% 5¸˛ ˆ

Dora

The right player discarded % just now.

This ready hand is currently 2000 points with a bad wait. Calling kan

on % is not justiﬁable when t he wait is bad.

Open quad?

—„„´ˆ˜%%% 5¸˛ ¯

Dora

The right player discarded % just now.

This ready hand is currently 1000 points with a good wait. Calling

kan on % is not justiﬁable. Even when one of the tiles in your hand

becomes new dora, the hand value only increases from 1000 to 2600

points.

9.3.3 From an open set to an open quad (kakan)

When you draw the fourth tile of an open set, you have an oppor-

tunity to extend the open set to an open quad. Conditions to justify

extending an open set to a quad are more demanding than those for

a concealed quad but less demanding than those for a regular open

quad. Doing so is less foolhardy compared with a regular open quad

because you are not losing four safety tiles. At the same time, this is

riskier than making a concealed quad because you may be running

the risk of getting chankan (Robbing the Kong).

You can extend an open set to an open quad in any of t he follow-

ing situations:

• the hand is 1-away or better with a good wait, and it has two

230 CHAPTER 9. MELDING JUDGEMENT

han or more;

• the hand is 1-away or better, and there are not many turns left

to draw tiles;

• you need more dora or minipoints to improve the placement

(especially in South-4);

• you are losing badly.

Open set to quad?

—¸¸¸‹‹ˆˆ˜ %M% 

Dora

You drew the fourth % just now.

This 1-away hand has two han and a good wait. Calling kan on % is

justiﬁable.

Open set to quad?

—¸¸‹›ˆ˜¯¯ %M% ¸

Dora

You drew the fourth % just now.

This hand has three han and a good wait. However, since it is 2-away

from ready, calling kan on % is not justiﬁable.

9.4. MISCELLANEOUS TIPS FOR MELDING 231

9.4 Miscellaneous tips for melding

9.4.1 Think ahead

When you call pon, you have to say “pon! [p´ON]” out loud imme-

diately and nothing else. There is no such call as “Wait!”, and you

will have to forgo your call if (1) the next player has already drawn

their tile before you call pon or (2) another player has already called

chii before you do.

This means that you need to think ahead and

make up your mind about what tile to call before the tile is discarded.

That is, you should think about what tile(s) can improve the wait

and/or the scores of your hand all the time. For example, consider

the following hand.

Thinking ahead

—‰ııı¸˛‚‹‹ ##K

What tile(s) are you waiting for?

The hand is ready, waiting for . However, you should also be pre-

pared for melding further to improve the wait and/or the scores. If

you draw or call pon on ‹ and discard —, the wait will be upgraded

to an irregular 3-way wait of -ȷ ‰. Moreover, if you draw or call

pon on the red » and discard —, not only the scores get better but

also the wait will be upgraded to a side wait of -ȷ.

Relatedly, think about what to discard upon melding before you

call. If you are unsure about what to discard upon melding, it prob-

ably means you should not make the call.

A pon call takes precedence over a chii call, but only if calls are made simulta-

neously. If the chii call was made well before the pon call, the chii call should

take precedence.

232 CHAPTER 9. MELDING JUDGEMENT

9.4.2 Be ready for dora

You should also think about how to utilize dora when melding.

Consider the following hand.

Utilizing dora

ﬁﬁ‹›“”„ˆ˜˜ $L$ ›

Dora

The left player discarded “ just now.

We should deﬁnitely call chii on “ to make the hand ready, but

the question here is: should we chii with ‹› or with ”„? Let’s

compare the resulting hands in each of the two possibilities.

ﬁﬁ“”„ˆ˜ 8‹› $L$ ›

ﬁﬁ‹›“ˆ˜ 8”„ $L$ ›

Dora

Notice that the ﬁrst hand can accept another dora. That is, if you

draw ›, you can keep it and discard „ to improve the hand value

from 2000 to 3900. With the second hand, you will have to discard

the dora when you draw another.

Calling chii with ‹› is better also from a perspective of de-

fense. Having to discard „ against an opponent’s riichi is much bet-

ter than having to discard ›. › can be captured by both ˛-› and

›-„ suji, whereas „ can be captured only by ›-„ suji. More-

over, even if you deal into an opponent’s ›-„ wait, the hand value

would be lower if you discard „ than ›, on average.

Of course, this is unless an opponent has sanshoku of 678.

9.4. MISCELLANEOUS TIPS FOR MELDING 233

9.4.3 Be mindful of the seating

Each time you take a tile from the facing player (toimen) with a

pon (or kan) call, the left player’s (kamicha) turn gets skipped. Like-

wise, each time you take a tile from the right player (shimocha) with

pon/kan, the facing player’s turn gets skipped. At the same time, the

right player will have an additional chance to draw a tile in either of

the two instances. In this sense, your act of calling pon/kan beneﬁts

the right player while penalizing the left and the facing players. It is

useful to keep this in mind in making a melding choice, especially

when the beneﬁt of melding only slightly outweighs the beneﬁt of

keeping the hand closed in terms of tile eﬃciency.

For example, when you are North, you should not meld as ag-

gressively because doing so will beneﬁt the dealer. Likewise, when

you are South, you should try to call pon from the facing player (North)

rather than from the left player (East), so that you can penalize the

dealer. The same is true when there is a clear front-runner in the

game. When your right player is much ahead of the game, you should

try to have a closed hand rather than a melded hand. On the other

hand, when your left player is leading the game, you should meld a

bit more aggressively so you can penalize him.

Seating-related tip 1

When your right player is the dealer and/or the front-runner, try

not to call pon too much.

Applying the same logic, you do not want your right player to

call pon from your left player. This means that, if you plan to discard

something that can be pon ’ed by the right player, you should do so

sooner rather than later. For example, suppose you are East, and you

are deciding which one of the three valueless wind tiles !" to

234 CHAPTER 9. MELDING JUDGEMENT

discard in the 1st turn. In this case, you should discard ﬁrst. If the

South player calls pon on you discard, that would be much better

than if he called

pon

from the North player. Moreover, if you discard

, there is a good chance that another player may do the same in

the 1st turn as well, lowering the chance that the South player builds

a pair of in later turns and calls pon.

Seating-related tip 2

When discarding valueless wind tiles, discard the right player’s

wind ﬁrst, then the facing player’s wind next.

Chapter 10

Grand strategies

The most important goal in mahjong is to win a game or generally

improve the placement. I do not deny the inherent joy of winning an

expensive hand with rare yaku. However, we should always keep

in mind that winning a hand is just a means to an end; sometimes

dealing into an opponent’s (cheap) hand can serve our purpose of

winning the game. In this chapter, I will discuss strategies to improve

the placement.

10.1 What do do in South-4

Most mahjong rule sets adopt some type of uma system where

players get some extra bonus / penalty points according to the place-

ment. For example, EMA rules award 15000 points to the ﬁrst ranked

player, 5000 points to the second ranked player, −5000 points to the

third ranked player, and −15000 points to the fourth ranked player.

Such systems make it clearer t hat getting a better placement is gen-

erally more important than simply winning hands.

Suppose you are currently ranked fourth in South-4, and that the

third ranked player has 1800 more points than you do. In such a sit-

uation, winning a 1000 hand is not very meaningful. If you manage

to add just one more han and win, you will not only get 2000 points

directly but also get an extra 10000-point bonus for coming in third,

a total of 12000-point gain. This is as big as winning a haneman hand.

In a situation like this, the tradeoﬀ between speed and hand value

is qualitatively diﬀerent than usual. For example, when choosing

You will get −5000 points instead of −15000 points, resulting in a net gain of

10000.

236 CHAPTER 10. GRAND STRATEGIES

between a good-wait one-han hand and a bad-wait two-han hand,

you should deﬁnitely choose the latter. After all, you are essentially

comparing a good-wait 1000-point hand with a bad-wait 12000-point

hand. On the other hand, when choosing between a good-wait two-

han hand and a bad-wait three-han hand, you should choose the for-

mer. Increasing the (virtual) hand value from 12000 to 13900 would

not be worthwhile if doing so signiﬁcantly diminishes the chance of

winning.

Suppose further that the third ranked player is the dealer. Then,

you will have another option to improve the placement. That is, if

anyone other than t he third ranked player gets a mangan tsumo (or

above), you will come in third. This is because the dealer pays 2000

more points than a non-dealer in case of a mangan tsumo. When this

happens, you will lose 2000 points for the mangan payment but gain

10000 points for the placement bonus, resulting in a net gain of 8000

points. This is as big as winning a mangan hand yourself.

Suppose yet further that the second ranked player is the right

player, who is behind the ﬁrst ranked player by 10000 points. Then,

he will try to get a mangan tsumo because doing so puts him in the

ﬁrst place. If he is obviously pursuing a honitsu hand, you may want

to discard tiles in the suit he is collecting so he can meld his hand to

get it ready.

If he indeed gets a mangan tsumo, you will get 8000

points; if he wins by ron from the third ranked player, you will get

10000 points.

That being said, assisting other players in hopes of their getting a

mangan tsumo is more like a last resort. What you should think about

Of course, you assist him only until he gets ready. You need to be careful not

to deal into the mangan hand you helped him make.

10.1. WHAT DO DO IN SOUTH-4 237

ﬁrst and foremost is winning your own hand that is just expensive

enough to improve your placement, which we will now turn to.

Improving the placement by ron / tsumo

As the discussion in the previous section illustrates, you need to

be extra conscious about your placement in South-4. If you are cur-

rently ahead of t he game, your top priority is to maintain your place-

ment. If you are behind, you should do your best to improve your

placement as much as possible.

In South-4, the ﬁrst thing you need to do before the hand begins

is to ﬁgure out the point diﬀerences between you and other players.

When playing online on Tenhou, this can be easily done any time by

mouseovering the middle board, as illustrated in Section 1.4.4. When

playing oﬄine, each player should count and report their points be-

fore the hand begins.

Once you ﬁgure out the point diﬀerences, you then need to know

how expensive your hand has to be to improve your placement. In

doing so, you need to ﬁgure out the required hand values under

three possibilities, as follows.

1. ron from anyone

2. tsumo

3. direct hit ron

The ﬁrst possibility to consider is winning your hand by ron from

anyone (that is, not from the very player you are trying to overtake).

For example, suppose you are currently ranked second, and the ﬁrst

ranked player has 3400 more points. Then, winning a 3 han–30 mini-

points hand (3900 points) by ron from anyone is suﬃcient to improve

238 CHAPTER 10. GRAND STRATEGIES

your placement. You should thus aim to have a 3-han hand. Since

you don’t need a 40-minipoint hand, melding is also an option.

The second possibility to consider is winning your hand by tsumo.

For example, suppose you are currently ranked second, and the ﬁrst

ranked player (non-dealer) has 9500 more points. Then, getting a

mangan tsumo is suﬃcient to improve your placement because you

gain 8000 points while the ﬁrst ranked player loses 2000 points, in-

ducing a 8000 + 2000 = 10000 point diﬀerence. You should thus aim

to have a mangan hand and try to win it either by tsumo or by ron

from the ﬁrst ranked player.

The last possibility to consider is winning your hand by ron from

the very player you are trying to overtake. For example, suppose

you are currently ranked second, and the ﬁrst ranked player (non-

dealer) has 15200 more points. Then, even getting a haneman tsumo

is not enough. You need a mangan ron directly from the ﬁrst ranked

player. This is suﬃcient because you gain 8000 points while the ﬁrst

ranked player loses 8000 points, inducing a 8000 × 2 = 16000 point

diﬀerence.

Among these three possibilities, your hand value judgement should

be based primarily on the ﬁrst possibility (i.e., winning it by ron from

anyone). This one requires the highest hand value but it is the most

realistic. Given that the player you are trying to overtake will try

hard not to deal into your hand, making your hand value judgement

based solely on the third possibility (direct hit ron) is too much of a

wishful thinking.

With this in mind, consider the following hand. Assume that you

are the North player in the 6th turn in South-4. You are currently

10.1. WHAT DO DO IN SOUTH-4 239

ranked second, and the ﬁrst ranked player (South) has 5100 more

points.

South-4: 5100 points behind, dora ȷ

–—ȷ¸˛‚‹‹ˆ¨˝˚ ﬁ

Draw

What would you discard?

If you keep ﬁ and discard ˆ, the hand is ready. However, doing

insta-riichi with the current hand is not ideal. Since the hand value

is only 2600 (2 han–40 minipoints), winning it by ron from the third-

ranked or fourth-ranked player will not improve your placement (un-

less you get ippatsu or ura dora). Also, getting tsumo will only give

you 1000-2000 (3 han–30 minipoints), generating only a 4000 + 1000 =

5000 point diﬀerence. This is not suﬃcient to improve the placement.

You should rather keep the hand 1-away by discarding ﬁ. If you

draw ´ or ˜, you can then do insta-riichi to get riichi + sanshoku =

at least 5200 (3 han–40 minipoints). Winning it by ron from anyone is

now suﬃcient to improve the placement. If you draw ı or ﬀ, you

can also do insta-riichi to get riichi + pinfu + dora. Winning it either

by tsumo or ippatsu ron is suﬃcient to improve the placement.

Point diﬀerence induced by tsumo

Getting the correct point diﬀerences induced by tsumo can be

a bit complicated. For example, suppose you are the North player,

currently ranked second in South-4. The West player is leading the

game, having 6300 more points. In this situation, would winning a

3 han–30 minipoints hand by tsumo be enough to improve the place-

Drawing ﬀ means you are in furiten, but you should still do insta-riichi.

240 CHAPTER 10. GRAND STRATEGIES

ment? What about winning a 3 han–40 minipoints (= 4 han–20 mini-

points) hand by tsumo?

To calculate the point diﬀerence induced by tsumo, we add the

points you gain and the points your rival (the ﬁrst ranked player)

loses. For instance, the point diﬀerence induced by a 3 han–30 mini-

points hand is: 4000 (your gain) + 1000 (your rival’s loss) = 5000

points. The point diﬀerence induced by a 3 han–40 minipoints hand

is: 5200 (your gain) + 1300 (your rival’s loss) = 6500 points. In this

example, getting a 3 han–40 minipoints tsumo is suﬃcient to improve

the placement, but getting a 3 han–30 minipoints tsumo is not.

It would be extremely tedious if we have to do these calculations

for several possible hand values all in our head in South-4. It would

be more eﬃcient if we memorize the induced point diﬀerences for

several representative cases; that way, we can use our time and en-

ergy thinking about other important things during the game.

Tables 10.1–10.4 below summarize induced point diﬀerences for

limit hands and those with 30, 40 (20), and 50 (25) minipoints. In each

table, the second column shows the induced point diﬀerences against

another non-dealer, whereas the third column shows those against

the dealer. Since a dealer pays twice as much as a non-dealer, the

induced point diﬀerences against a dealer are greater. In addition,

for each counter (continuation) placed on the table, the induced point

diﬀerence will get bigger by 400 points.

Note that these four tables assume that you are a non-dealer. When

you are the dealer, you do not usually need to do these calculations

because you get to continue the game if you win a hand anyway.

However, when playing with a bankruptcy rule or with time limits,

the dealer may not be able to continue the game, in which case even

10.1. WHAT DO DO IN SOUTH-4 241

Table 10.1: Limit hands

Tsumo Non-dealer Dealer

mangan 10000 12000

haneman 15000 18000

baiman 20000 24000

yakuman 40000 48000

Table 10.2: 30 minipoints

Tsumo Non-dealer Dealer

300-500 1400 1600

500-1000 2500 3000

1000-2000 5000 6000

2000-3900 9900 11900

Table 10.3: 40 (20) minipoints

Tsumo Non-dealer Dealer

400-700 1900 2200

700-1300 3400 4000

1300-2600 6500 7800

Table 10.4: 50 (25) minipoints

tsumo Non-dealer Dealer

800-1600 4000 4800

1600-3200 8000 9600

the dealer has to consider if winning a particular hand improves the

placement. Tables 10.5 and 10.6 at the end of this chapter provide a

summary for a dealer as well.

Memorizing these tables would be way more important than mem-

orizing, say, scores for 70-minipoint hands. With these tables in mind,

consider the following hand. Assume that you are the North player

in the 6th turn in South-4. You are currently ranked second, and the

ﬁrst ranked player (South) has 3300 more points.

South-4: 3300 points behind, dora ı

––ıﬀﬁ‚‹›“”„ˆ˜ ´

Draw

What would you discard?

You wanted to draw ȷ ﬁrst so that you can have riichi + pinfu + one

dora = 3900. Winning that hand by ron from anyone would improve

your placement. However, now that you drew ´, what should you

242 CHAPTER 10. GRAND STRATEGIES

do?

Recall that a 700-1300 tsumo would induce a 3400 point diﬀer-

ence. Since t his is a pinfu hand, getting riichi + pinfu + tsumo

gives you exactly 700-1300 tsumo. You should thus do insta-riichi by

discarding ı. Once you call riichi, you can do either (1) ippatsu

ron from anyone, (2) direct hit ron from the ﬁrst ranked player, or (3)

tsumo to improve the placement.

Consider a more complicated example that involves some mini-

point calculation. Assume that you are the North player in the 6t h

turn in South-4. You are currently ranked second, and the ﬁrst ranked

player (East) has 4700 more points. What are the conditions under

which you can improve your placement with the following hand?

South-4: 4700 points behind, dora ı

———¨¨$$ :„„ ˇˇ/

What are you waiting for?

Winning this hand on $ by ron from anyone or tsumo satisﬁes the

condition because it gives you 5200 points (toitoi + Green Dragon

with 40 minipoints). Winning it on ¨, however, only gives you 2600

points. You can still improve the placement if you get a direct hit

from the ﬁrst ranked player, but not if it is from other players. Even

though the ﬁrst ranked player is the dealer, you cannot improve the

placement if you draw ¨, either. Declaring tsumo on ¨ gives you

Whether or not you should let it go when the third or fourth ranked player

discards your winning tile depends on the point diﬀerence between you and

the third ranked player. Unless it is greater than 12000 points, you should call

ron and hope to get one ura dora.

10.1. WHAT DO DO IN SOUTH-4 243

700-1300, which induces only a 4000 point diﬀerence even against the

dealer.

However, if you manage to draw „ or ˇ, you should extend the

melded set to a melded quad. Doing so not only gives you a chance

of rinshan tsumo or new dora but also enables you to improve the

placement when drawing ¨. This is because the hand will have 50

minipoints if you tsumo: 20 for the base minipoints + 8 for a melded

Kong of „ or ˇ + 2 for a melded set of ˇ or „ + 4 for a concealed

set of — + 2 for a pair of $ + 4 for a concealed set of ¨ + 2 for tsumo

= 42, rounded up to 50 minipoints. A 2 han–50 minipoints tsumo

induces a 4800 point diﬀerence against the dealer.

Maintaining your placement

If you are ahead of the game in South-4, you should do your best

to maintain your current rank. Trying to win a cheap but fast hand

to end the game is an option, but be extra careful not to deal into an

opponent’s expensive hand. For example, suppose you have 15200

more points than the second ranked player. If neither you nor the

second ranked player is the dealer, he cannot defeat you even with

a haneman tsumo. Then, what you need to be wary of the most is

to give him a direct hit mangan ron. You will lose not only the 8000

points for t he mangan payment but also the 10000 bonus points for

the placement, a total net loss of 18000 points.

In order to ﬁgure out what exactly you should do when you are

ahead of the game in South-4, try to imagine what each of your oppo-

nents aims to do. Recall the situation I described in discussing riichi

judgement in Section 7.3, reproduced below.

Let’s think about the incentive structure for each of the other three

players in turn. First, the fourth ranked player (West) should try to

244 CHAPTER 10. GRAND STRATEGIES

East (you) 39000 South 22900

West 13000 North 25100

have a mangan tsumo, for that would put him in the third place. The

third ranked player (South) would need a 500-1000 tsumo or 2600 ron

to get the second place, which is a realistic goal to pursue. In order

for him to get the ﬁrst place, he would need eit her a haneman tsumo

or a direct hit haneman ron from you. Finally, in order for the second

ranked player (North) to get the ﬁrst place, he would need a haneman

tsumo or a direct hit mangan ron from you. Given that he has only

2200 more points than the third ranked player, the second ranked

player should rather aim to win whatever hand possible to maintain

the current rank.

So, what should you do in such a situation? What you should be

afraid of the most is a haneman tsumo by South or North. However,

notice that South and North are in a ﬁerce competition among them-

selves. Take advantage of this. If winning a fast hand yourself does

not seem possible, you should try to assist the South player. Since

the South player is your right player, you should discard versatile

middle tiles (3–7) so that he would call chii on them, possibly with

a red ﬁve (because South needs 2 han). Recall that even giving him a

direct hit mangan ron will secure you the ﬁrst place.

10.2 What to do by South-3

It is never too early to start paying attention to your placement. If

you are behind other players, the target point diﬀerence you should

achieve before the beginning of South-4 is 10000 points or fewer.

Overtaking with a 10000 point diﬀerence in a single hand is a tough

but not unrealistic goal; you can do so either by mangan tsumo or

10.2. WHAT TO DO BY SOUTH-3 245

haneman ron.

Suppose the dealer wins a mangan tsumo in East-1. Now he is

leading other players by 16000 points, which is a bit depressing. How-

ever, instead of trying to overtake him with a single expensive hand,

you should aim to reduce the point diﬀerence from 16000 to 10000 by

the end of South-3. This is a more realistic goal; keep calling riichi

with a pinfu-only hand, and you will eventually get tsumo + one ura

dora which induces a 6500 point diﬀerence.

On the other hand, if you are ahead of the game, you should aim

to have a 10000 or greater point diﬀerence with the second ranked

player in South-4. For example, consider the following hand. Sup-

pose you are the North player in the 6th turn in South-3. You are

currently ranked ﬁrst, and you only have 1000 more points than the

second ranked player (West).

Riichi judgement in South-3: dora ȷ

—‰ȷﬀﬁ‚‚‚¨˝ˇˇ"

Riichi or dama?

The choice here is between keeping the hand dama to maximize your

chance of winning the hand or calling riichi to aim for a bigger point

diﬀerence. You should do insta-riichi in such a situation. If you

win this hand by dama ron, the point diﬀerence will only be 3600 in

South-4. Having a 3600 point diﬀerence is not much diﬀerent from

having a 1000 point diﬀerence from the perspective of the second

ranked player. However, if you win this hand with riichi, the point

diﬀerence will be at least 6200 in South-4, 9000 if you get one ura

dora, and 11000 if you tsumo. Having a point diﬀerence of 11000 in

246 CHAPTER 10. GRAND STRATEGIES

South-4 signiﬁcantly increases your chance of winning the game.

If you will be the dealer in South-4, the target point diﬀerence

with the second ranked player is 12000 or more, not 10000. This is

because a mangan tsumo by a non-dealer induces a 12000 point dif-

ference against the dealer.

You should also be mindful of induced point diﬀerences by noten

penalties in South-4 and South-3. The maximum point diﬀerence in-

duced by noten penalties is 4000 (1-player noten and 3-player noten).

This means that you should aim to have at least 4000 point diﬀer-

ence by the end of Sout h-3. For example, suppose you are leading

the game in South-4 and you are the dealer. If the hand ends in ex-

haustive draw and the point diﬀerence between you and the second

ranked player is more than 4000 points. Then, you should declare

noten (not ready) and terminate the game even when you have a ready

hand. You will not have this option if the point diﬀerence is fewer

than 4000 points.

10.3. TABLES FOR INDUCED POINT DIFFERENCES 247

10.3 Tables for induced point diﬀerences

Table 10.5: For non-dealer

Tsumo Non-dealer Dealer

300-500 1400 1600

400-700 1900 2200

500-1000 2500 3000

700-1300 3400 4000

800-1600 4000 4800

1000-2000 5000 6000

1300-2600 6500 7800

1600-3200 8000 9600

2000-3900 9900 11800

2000-4000 10000 12000

3000-6000 15000 18000

4000-8000 20000 24000

6000-12000 30000 36000

8000-16000 40000 48000

Table 10.6: For dealer

Tsumo Non-dealer

500 2000

700 2800

1000 4000

1300 5200

1600 6400

2000 8000

2600 10400

3200 12800

3900 11600

4000 16000

6000 24000

8000 32000

12000 48000

16000 64000

Appendices

248

Chapter A

Manners for oﬄine playing

Manners are meant to make the game of mahjong a pleasant ex-

perience. They are a collection of small tips and techniques the fore-

runners have developed to avoid unnecessary troubles. I present

manners for four diﬀerent phases of a game — (1) dealing tiles, (2)

drawing and discarding, (3) calling, and (4) winning a hand. Each

entry is given a rank, from one star (

∗

) to three stars (

∗∗∗

). Three-star

manners are more important; you should try to acquire three-star

manners ﬁrst, and then move on to practice two-star and one-star

ones.

A.1 Dealing tiles

1-1: Shuﬄing

∗∗

I recommend a 2-step shuﬄing approach. First, shuﬄe tiles re-

ally hard. Don’t worry about keeping the tiles face down at t his point.

This will guarantee that sets, runs, and pairs from the previous hand

are really broken apart. Second, put all the tiles up-side down and

shuﬄe them face down gently. This will guarantee that no one re-

members the locations of certain tiles.

1-2: Push the wall forward

∗∗

Once you build a wall, push it forward a little so that the facing

player can easily reach your wall. If you push it forward too much,

you will lose the space for discards.

1-3: Tilt the wall

∗

When pushing the wall forward, it would be better if you tilt the

wall a little, as follows. This will make it even easier for the facing

A.1. DEALING TILES 251

player to pick a tile from your wall.

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

1-4: Split the wall

∗

In addition to tilting, some players like to mildly split the wall

into three blocks upon building the wall. You get six tiles on the left,

ﬁve tiles in the middle, and another six tiles on the right, illustrated

as follows.

ﬃﬃﬃﬃﬃﬃ

ﬃﬃﬃﬃﬃ

ﬃﬃﬃﬃﬃﬃ

This will make it signiﬁcantly easier for the dealer to identify the

breaking point in the wall.

1-5: Break the wall

∗

After rolling the dice, the dealer should break the wall himself.

When the dice indicate a number k that is greater than 7, it is easier

to count 15 − k tiles counterclockwise rather than counting k tiles

clockwise after identifying which wall to break. For example, when

the dice roll is 9, the dealer should count 6 tiles counterclockwise

(i.e., 6 from the left edge), leaving 6 tile pairs on the left of his wall.

Likewise, when the dice roll is 10, the dealer should count 5 tiles

counterclockwise, leaving 5 tile pairs on the right of the right player’s

wall (right from the dealer’s view, left from the right player’s view).

1-6: Put the rinshan tile down

∗∗

After breaking the wall, the rinshan tile (the ﬁrst replacement

tile) should be preemptively put down. This is to prevent it from

falling over. This should be done by the player who has the dead

wall in front of him.

252 APPENDIX A. MANNERS FOR OFFLINE PLAYING

1-7: Tur n over the dora indicator

∗∗∗

Immediately after putting down the rinshan tile, the dora indi-

cator should be turned over. Doing so is way more important than,

say, separating the dead wall from the end of the wall (which is com-

pletely unnecessary especially at the beginning of a hand). In my

experience, European players somehow like to do the latter ﬁrst and

don’t open the dora indicator even after they ﬁnish the dealing.

1-8: Look at the tiles

∗∗

As you take tiles from the wall during the initial dealing, you

should start taking a look at them. Don’t wait until you get all thir-

teen tiles; doing so is a waste of time not only for you but also for the

other players.

1-9: Dealer’s ﬁrst discard

∗∗

The dealer should not discard a tile until the North player gets

all thirteen tiles. This is to give everyone a roughly equal amount of

time to decide whether to call pon / chii / ron on the ﬁrst discard.

A.2 Drawing and discarding

2-1: Don’t use both hands

∗∗

During the play — after the dealing is done and before the hand

ﬁnishes —, you should use only one hand. If you are right-handed,

you should not put your left hand on the table, either. Don’t do

things like drawing with your left hand and discarding with your

right hand. This is to prevent (the appearance of) cheating. The only

occasions where using both hands during the play is acceptable are

(1) when sorting the tiles in your hand, and (2) when revealing your

hand upon winning or in cases of exhaustive draw and four-riichi

abortive draw.

A.2. DRAWING AND DISCARDING 253

2-2: Arrange the discards

∗∗∗

Discards should be arranged in an orderly way (six tiles in a row).

2-3: Discard before sorting

∗

You should discard a tile before you put the tile you draw into

your hand. Putting the newly drawn tile into your hand upon tsumo

is a serious violation. To avoid it, you should make a habit of not

putting a newly drawn tile into your hand immediately.

2-4: Let go of the discard

∗∗

Upon discarding a tile, you need to let it go immediately and not

keep your ﬁnger on the tile. This is to guarantee that all the other

three players can see which tile you discarded all together.

2-5: Don’t take an overly long time

∗

Keep in mind that t hree other players are waiting for you; you

should not take an overly long time to draw / discard. In particular,

beginners may want to pay attention to the following.

• When your turn comes, draw a tile immediately (unless you

need to think about whether to call the last discarded tile).

• Once you make up your mind about what to discard, discard it

immediately.

2-6: Don’t speed other players

∗∗∗

Yes, it could be irritating if someone is taking a long time, but be

considerate. You should not press other players to be faster than they

could.

254 APPENDIX A. MANNERS FOR OFFLINE PLAYING

A.3 Calling

3-1: Vocalize clearly

∗∗∗

When you call pon [p´ON], chii [tS´I:], kan [k´2N], riichi [r´I:tS], ron

[r´ON], or tsumo [ts´umo], utter the word clearly so the other three play-

ers can hear you.

3-2: Vocalize before taking an action

∗∗∗

When calling pon, say “Pon.” ﬁrst before taking the tile. Likewise,

when you call riichi, say “Riichi.” ﬁrst before discarding a tile and

placing a riichi bet (see also 3-4 below).

3-3: Wait before calling chii

∗

When calling chii, wait for 1 second before uttering the word.

On the other hand, when calling pon / kan / ron you must do so

immediately. If someone says chii ﬁrst (after taking 1 second), other

players should not be able to call pon / kan / ron. Pon, kan, and ron

should take precedence only if calls are made concurrently.

3-4: Calling riichi

∗∗∗

The procedure to call riichi is as follows.

1. Say “Riichi.”

2. Discard a tile, rotating it sideways.

3. Conﬁrm that no one calls ron on the discarded tile.

4. Place a riichi bet.

The most important point is t hat you say “Riichi.” before discarding

a tile. This is because the opponents’ choice of what to do with your

EMA rules allow a pon call to occur even after a chii call is made. I think this

should be changed.

A.4. WINNING A HAND 255

discard (i.e., whether or not to call pon on it, etc.) may be diﬀerent if

you riichi.

A.4 Winning a hand

3-1: Vocalize clearly

∗∗∗

When winning a hand, you need to say ron or tsumo clearly. It is

also OK to say “mahjong” instead.

3-2: Don’t put the winning tile into the hand

∗∗∗

When winning by tsumo, don’t place the winning tile inside the

hand. Just place the winning tile right next to your hand. This is

important because scores (yaku and minipoints) may be diﬀerent de-

pending on which tile was the one to complete the hand.

3-3: Don’t take the winning tile

∗

When winning by ron, some European players grab the winning

tile and place it right next to their hand. Don’t do it. You should re-

frain from doing this to prevent (the appearance of) cheating. People

do this on TV, but they do so only for the camera.

3-4: Sort the tiles before revealing your hand

∗∗

You need to sort the tiles before showing your hand, so that other

players can easily check your hand’s score and possible furiten vio-

lation. Do not split the hand into constitutive groups. Doing so may

actually obstruct other players’ vision.

3-5: Declare yaku

∗∗

After revealing your hand, reveal the ura dora if you have called

riichi. You need to show the ura dora to all the other players even

when you don’t get any of them. This is to make sure that you are

256 APPENDIX A. MANNERS FOR OFFLINE PLAYING

not underreporting your hand value.

After that, you should declare

all the yaku in your hand.

3-6: Declare the score

∗∗∗

You need to declare the score of your hand yourself. It is OK

to get other players’ help on scoring, but you need to be the one to

declare it. When declaring tsumo scores, say the payment by a non-

dealer ﬁrst, followed by the payment by the dealer. For example,

when declaring a 300-500 tsumo, say “Three hundred, ﬁve hundred.”

rather than “Five hundred, three hundred.”

3-7: Conﬁrm the score

∗

When one player wins a hand, the other three players must also

see the hand and conﬁrm the declared score. You should also check

if the hand was not furiten.

3-8: Payment

∗

A standard stick set would include four kinds of sticks, as follows.

100 point 1000 point

5000 point 10000 point

In addition to these, I suggest you prepare a set of four 500-point

sticks. I usually use green 100-point sticks that I bought in Japan,

It may sound odd, but there are situations where you have strategic incentives

to underreport your hand value. Trying to avoid bankruptcy of another player

when you are still ranked second or third is one obvious example. For another

example, players may not want to change the placements of other players in a

game if they are competing for ranking at a tournament. Underreporting the

hand value is usually illegal.

Some people may say that you only need to declare the score and that declaring

yaku is either unnecessary or even undesirable. I personally think it’s unnec-

essary, especially when playing with experienced players. However, given that

not everyone at the table can quickly identify all the yaku in another player’s

hand, it would be prudent if the winner declares all the yaku.

A.4. WINNING A HAND 257

which look like: .

If you don’t have any green sticks,

you can use anything (e.g., coins, poker chips, etc.) for substitute.

Having 500-point sticks would make stick payment much more eﬃ-

cient.

To streamline the payment, you should try to minimize the num-

ber of sticks exchanged on the table. Here are two examples of eﬃ-

cient method of payment.

3900 ron

When a player wins a 3900 hand, the player who discarded the

winning tile gives the winner one , and the winner

gives back 1100 points with one and one .

5200 tsumo

When a player gets a 1300-2600 tsumo, the most eﬃcient and beau-

tiful method of payment is as follows.

1. The ﬁrst non-dealer (the one sitting closer to the winner) pays

the exact amount with one and three .

2. The second non-dealer gives the winner 1500 with

one and one .

3. The winner gives back the second non-dealer 200 with

two of the three he got from the ﬁrst, which en-

sures that the second non-dealer pays 1300.

4. The dealer gives the winner 5100 with one and

one .

The image of 500-point stick was created by someone known as “381654729”

(Tenhou ID: ). I thank him for letting me use it in this book.

258 APPENDIX A. MANNERS FOR OFFLINE PLAYING

5. The winner gives back t he dealer 2500 with the two

he got from the two non-dealers and the one he

got from the second non-dealer, which ensures that the dealer

pays 2600.

6. After all the exchanges, what remains on the table is exactly

5200 with one and two .

For this to work out perfectly, everyone needs to be on the same page.

It may sound complicated at ﬁrst, but it sure feels good when the four

players manage to make it happen together.

3-9: Exhaustive draw

∗

In case of exhaustive draw, the dealer should be t he ﬁrst one to

declare whether or not he has a ready hand. If he wants to declare

ready, he has to show the hand and say “Tenpai.”; if not, say “Noten.”

or “Not tenpai.” without showing his hand. Then, South, West, and

North declare tenpai or noten in that order.

The order of declaration could make a diﬀerence in some (rare)

occasions. Declaring ﬁrst is advantageous in some instances and dis-

advantageous in others. Suppose the dealer is ranked ﬁrst in South-

4, having 2900 more points than the second ranked player. Suppose

further that he has a ready hand. In such a situation, the dealer has

an incentive to make a declaration after the second ranked player. If

the second ranked player declares noten, the dealer would want to

declare noten to terminate the game.

On the other hand, if the sec-

ond ranked player declares tenpai, the dealer would want to declare

tenpai and continue the game. This is because the induced point

diﬀerence in case of tenpai–noten is either 3000 (2-player tenpai) or

Continuing the game means he r uns t he risk of losing the placement bonus

(10000 in EMA rules) and the oka points (if any).

A.4. WINNING A HAND 259

4000 (1-player tenpai or 3-player tenpai), each of which exceeds the

current point diﬀerence of 2900.

Declaring ﬁrst can be advantageous only when playing with a

bankruptcy rule. Suppose one player is at the verge of bankruptcy,

having only 1300 points. Suppose further that both he and the ﬁrst

ranked player have already declared noten. In such a situation, if the

second ranked player declares tenpai ﬁrst, the third ranked player

would have to declare noten even when he has a ready hand. Oth-

erwise, the fourth ranked player goes bankrupt and the game is ter-

minated.

Because of these advantages and disadvantages of declaring ﬁrst,

we should stick with the predetermined order for the sake of fairness.

Chapter B