Life In 19x19

Review: Get Strong at the Endgame
Page 1 of 1

Author:  RobertJasiek [ Wed Sep 12, 2018 6:03 am ]
Post subject:  Review: Get Strong at the Endgame

The following disclaimer copied from the bottom to here. -JB
Disclaimer: Robert Jasiek is a researcher in the endgame and other go theory, author of endgame books and other go books, and go teacher.

Review: Get Strong at the Endgame

General Specification

* Title: Get Strong at the Endgame
* Author: Richard Bozulich
* Publisher: Kiseido
* Edition: 1997 (first printing)
* Language: English
* Price: EUR 19.50
* Contents: endgame
* ISBN: 4-906574-57-2
* Printing: good
* Layout: intermediate
* Editing: very weak
* Pages: 198
* Size: 146mm x 209mm
* Diagrams per Page on Average: 6
* Method of Teaching: examples
* Read when EGF: 10k - 1k
* Subjective Rank Improvement: --
* Subjective Topic Coverage: --
* Subjective Aims' Achievement: +


When I first read the book in 1997 as a 4d, it had no effect on my endgame skill, I suspected it to have quite a few mistakes but, at that time, my endgame calculations were too weak to be sure whether the mistakes were the book's or mine. Although people like to point out errata in books, so far the errata page on Sensei's Library for this book has mentioned only one significant mistake and a few typos. On David Carlton's book review webpage, Georg Snatzke (then 1d), suspected a lot of mistakes but has not been sure about his own understanding, either.

To clarify whether the book contains just a few or a lot of mistakes, in 2018 I (now 5d) have read the book for the second time after having educated myself about endgame theory. Usually, I need at most one day to read a typical book for single digit kyus. However, I have spent two weeks on my second reading of this book. Half of the time, I have invested on a) solving every problem carefully versus b) understanding and verifying correctness of the short texts in the book.


This is a problem book with 291 problems and answers but essentially without theory. For that, we are referred to Ogawa's The Endgame and its variety of traditional endgame theory under territory scoring.

The book has four chapters and a short appendix, which compares professional versus amateur endgame play in one particular game. The answers of chapter 1 appear late in the book. This gives the reader options of how to use the initial "test". He can consult the answers immediately, or read chapters 2 and 3 about tesujis and evaluation, then solve the problems of chapter 1 again before reading its solutions. If, however, the reader wants to use this as a measure of his improvement (as the author advertises), he would have to recall his first solution of the problems in chapter 1 and compare to his second solution.

37 of the 42 problems of chapter 1 and the 28 problems of chapter 4 are copied from Kano Yoshinori 9p's Yose Jiten (Endgame Dictionary). I have recognised, and solved within a second, quite a few of the problems of the chapter 2 about tesujis because their shapes appear in various classical problem collections or Ogawa's book. On the other hand, the selection of endgame tesujis (or sometimes ordinary moves behaving like tesujis) is representative. The problems in chapter 3 about evaluation show standard shapes (such as late follow-ups of josekis), are compiled easily but useful and relevant training

Typically, there are two diagrams for the answers of chapter 1 (correct versus failure variation), three diagrams for the answers of chapter 2 (correct, failure, comparison to ordinary moves) and two diagrams for the answers of chapter 3 (Black versus White starts). The problems of chapter 4 must be solved for Black's or White's start so typically the answers show two diagrams per problem.

Conceptually, the book is best assessed by perceiving two parts. Part I comprises chapters 1 + 2 and their answers, the problems of chapter 3, and the problems and answer diagrams of chapter 4. Part II comprises the answers of chapter 3 and the answer texts of chapter 4. Part I trains reading, tesujis and endgame evaluation. Part II conveys the author's applied sequences and calculation of endgame evaluation. I do not know whether the answer text in chapter 4 originates from Kano or Bozulich.

Part I

We have small board (usually 11x11) positions during the late or very late endgame. Each problem combines a few local endgames and we must find a correct move order to achieve the best result. The reader must spot tesujis and apply tactical reading. Move values might guide or deceive him so this chapter is hardly about evaluation. For the intended single digit kyu reader, effort is required. A low dan player might classify the problems as easy to intermediate. A beginner of endgame skill may find some of the problems hard, given the task of finding optimal play.

Chapter 4 continues chapter 1 but now we apply knowledge acquired during earlier chapters, must solve every problem twice (once for either starting player), the problems are of an intermediate (sometimes advanced) level, we are also asked to calculate move values and spend much time on solving. Even dans can find these problems interesting.

A reader not having seen a tesuji in chapter 2 before must invest effort on some of the problems. However, recognising a familiar shape makes a solution very easy. Similarly, some readers profit little from the tactical reading exercises in the book if they have already acquired a reasonable basic reading skill - some other readers might significantly improve if this should be their first, or an early, relevant source for training tactical reading or tesujis.

The problems of chapter 3 are good for training basic endgame calculation if the reader already knows some theory to apply it here and does not look at the answers. Circa half of the problems is easy and half is of intermediate calculation difficulty. Tactical reading or tesujis only play a minor role in this chapter.

If the book only consisted of part I, my assessment would be: it is useful for single digit kyus to train their tactical reading, become aware of tesujis and train their basic endgame calculation. On the other hand, since the texts provide rather little information, one can as well read the diagrams of Asian endgame books possibly without understanding any text. This book offers little different because too many examples or shapes are copied from other sources.

Even if we ignore part II, the typos permit an only intermediate assessment of the editing. The layout is also not a highlight. For example, repeating the same problem task 14 times in chapter 4 reduces it to a layout element.

The answer diagrams in chapter 4 have the following three mistakes. Answer 271 Black starts: White 14 is a mistake resulting in Black's win by 1 point (not in a tie, as the book's typo suggests) and must be played at 17, threatening a wedge as a follow-up and achieving a tie. Even with the dame filled, White need not reinforce the lower left region. Answer 271 White starts: likewise, White 11 is a mistake; played at 14 improves the score by 1 point for White. (Furthermore, "White 1 is the most efficient move." is wrong because White 5 is more efficient and different places in the book teach such details of local efficiency relevant in other positions.) Answer 283 White starts: unlike the book suggests, the moves are not "straightforward". Black 4 results in White's win by 3 points. Instead, Black 7 is correct and results in White's win by 1 point so is 2 points better.

The book does not mention most alternative, correct variations, which are not just equal options. The book should have a few more pages showing relevant, important, difficult variations. Just one correct diagram for a starting player is not always enough.

Part II

For almost all of the easy problems in chapter 3, the answer diagrams and texts are correct. Otherwise, the best that can be said about part II is that the reader should simply ignore it. This presumes his awareness of this advice.

Each of 'the answers in chapter 3' and 'the texts of the answers of chapter 4' contains many mistakes. Georg Snatzke nails it neatly when saying that there was a really big flaw in this book, bigger than in any other go book he knew, a lot of the solutions in the Bozulich-part of the book must be simply wrong, he could never trust the given answers, slowing him down and diminishing many of the positive aspects of the book. What are those mistakes?

We might tolerate typos affecting values because by far most of the mistakes are more severe. We can also tolerate intentional rounding if a number with a fractional component is rounded to the nearest integer. However, there is also rounding to the wrong adjacent integer or rough rounding some 2 points off resulting from guessing or lazy calculations. While the reader is asked to calculate move values, sometimes the author only gives rough lower or upper bounds.

In chapter 4, many move values are small so also many mistakes are small. When teaching endgame calculations, small mistakes (which are not intentional, correct rounding) must also be avoided so that the reader learns well.

The author cannot decide how to assess a basic ko. He sometimes calls it a half point ko, assesses it as one third of a point (as one would do under modern endgame theory), states the difference 1 between Black or White winning the ko or estimates a ko as some unspecified fraction.

The author could have eliminated circa two thirds of his mistakes in endgame evaluation by spending another 3 to 6 weeks on editing and proofreading. He would have found the trivial mistakes and those mistakes for which application of endgame theory approaches his limit of understanding at the time of writing the book.

Such mistakes are especially important. For part of the easy or difficult value calculations, the author shows his scope of understanding. With more effort, he could have applied his knowledge and avoided many of those mistakes approaching his limit of understanding. It mainly affects two kinds of mistakes: calculation when a local endgame has not so easy follow-ups and the distinction of gote and sente. The former is a consequence of Ogawa's weak description of theory for follow-ups. The latter is a major problem of traditional endgame theory.

A couple of mistakes lie beyond the author's understanding of endgame evaluation, reach his limits where his understanding is unsure and can easily result in mistakes, or hit the gaps of traditional endgame theory. For a responsible writing of the book, he should have studied endgame theory and practical application in professional games for, say, three additional months and thereby enabled himself to avoid most of the remaining mistakes related to calculation of follow-ups or gote / sente distinction.

This would have restricted remaining mistakes to those beyond the scope of traditional endgame theory: reassessing local double sentes and their move values in view of their inexistence; distinguishing gote from sente when it is particularly difficult; continuing local sequences for too long when they must be interrupted by plays elsewhere; neglecting a careful study of gote versus sente options and so on. As a consequence of not knowing how to verify such aspects, additional mistakes sometimes occur as a side effect. For example, if gote and sente are confused, or a sequence is too long, wrong move values are calculated.

There are also various other, infrequent kinds of mistakes, such as not knowing which option is correct and therefore taking the average of the results of two variations.

Not everything that looks like a mistake is one. For one advanced kind of follow-up calculation, I have spent one hour on trying to understand the short text of a particular answer to find out the author's implicitly used method of calculation and verify its correct application. This addresses another major problem of the too short answer texts: a few numbers are mentioned, the reader has to decrypt the text to identify what those numbers express and then invest much effort in understanding how these numbers are combined in a calculation, provided the text does not contain any mistake. "Difference" is used informally when two diagrams are compared like pictures and we shall spot and count those black and white points that are different. When three diagrams are compared, interpretation is much more difficult and made even more difficult by ambiguous references: does the author mean the position before a sequence, after a sequence or (as in some other answers) after move 1 of the sequence shown? Next, the reader must understand whether to add or subtract, and which number from which other number. Decrypting the texts is an extra puzzle. The reader is supposed to already understand all incarnations of the theory fluently, even those parts hardly described anywhere before.

In other words, what this book really lacks is an explanation of the underlying theory or at the very least a careful, step-by-step application during the calculations. It is not the sheer number of problems that can make one stronger but it would be: good explanations paving understanding of what numbers are being calculated, why, how they are combined when the calculations proceed and how to recognise methods of calculation for application to other examples. Instead, the author adds more confusion when restricting considered intersections (locales) without specifying them or when even changing them.

When a problem asks how much move 1 is worth, the disappointing answer calculates the value of the move as part of its sequence. The author might not be aware that a move itself can have a different value. The book is more disappointing whenever it calculates a move value but the interesting, difficult part of calculation would be the verification of gote versus sente or successive versus interrupted local play. Calculating a double sente move value is often easy but the book does not offer the much more demanding calculation as gote or sente.

I cannot recommend a book with so many, relevant mistakes and gaps in explanation and calculation. A reader trusting the author might even become weaker from applying wrong endgame evaluations in his own games. A beginner might learn at least calculation in easy examples, provided he can distinguish them from the intermediate problems of evaluation, whose answers should be ignored. The book requires a major correction the most urgently.

Disclaimer: Robert Jasiek is a researcher in the endgame and other go theory, author of endgame books and other go books, and go teacher.

EDIT: punctuation for disambiguation.

Author:  xela [ Sat Oct 19, 2019 1:24 am ]
Post subject:  Re: Review: Get Strong at the Endgame

Thanks Robert, I'm pleased to find this detailed review. This book has bothered me for a long time: I felt that it had potential to be very useful, but I simply didn't understand a lot of the solutions. Your description of the mistakes is reassuring to me.

Author:  Gomoto [ Sat Oct 19, 2019 1:39 am ]
Post subject:  Re: Review: Get Strong at the Endgame

Thanks for the review, an example and diagram of a typical mistake in the book would probably add to the reading experience.

slightly offtopic ;-)

I need an entirely different kind of end game book.

Me calculating the end game: If I play this 2 point gote move I have won the game. Oh, In this other position, I have to connect once more AFTER all the dame are played. Me playing the connection move ...

Author:  RobertJasiek [ Sat Oct 19, 2019 4:04 am ]
Post subject:  Re: Review: Get Strong at the Endgame

Example of typical calculation mistake in the book:
I would need some hours of time for that but I am very busy. Identifying a mistake takes little time but explaining it carefully and editing the diagrams takes hours. It does not help you at the moment that writing a complete commentary on all evaluation mistakes in a couple of endgame books is one of too many of my projects in the pipeline. Currently, I can only mention that the 2nd printing of 2016 is very similar to the first printing.

Your idea of an endgame book:
It is a good idea but I still have to realise at least three other endgame books, of which each is more important to write. So likely, someone else should fulfil your wish. There are related discussions in a) Get Strong at the Endgame, 11x11 problems, b) Endgame 2, chapter Microendgame, c) my webpage and d) a number of Bill Spight's endgame exercises here but neither meets your idea exactly.

Author:  xela [ Sat Oct 19, 2019 3:56 pm ]
Post subject:  Re: Review: Get Strong at the Endgame

I've been going through the small board endgames in old issues of Go World. There are quite a lot of them! The indexes on the GWA DVD make it easy to search. Too early to say yet if this is useful, but it feels like it should be doing me some good.

Author:  Gomoto [ Thu Oct 24, 2019 1:34 pm ]
Post subject:  Re: Review: Get Strong at the Endgame

(Robert, did you have a look already at KataGo and the correctness and accuracy of its score estimates in the endgame?)

Author:  RobertJasiek [ Thu Oct 24, 2019 9:43 pm ]
Post subject:  Re: Review: Get Strong at the Endgame

No (because no time). Earlier experience with AI endgames: if they win, they maintain their win but they often lose points. Therefore, we still should not learn correct, score-optimising endgame from AI. If they are behind, I suppose their endgame is better but I have not studied yet how good or bad it is then. The safe advice is to learn endgame from correct human study and therefore from sources teaching (mostly) correctly: Spight, Meien, Törmänen, Jasiek, Berlekamp / Wolfe, CGT theorists (if you want maths) and Davies (examples in Go Monthly Review, Go World).

Author:  Gomoto [ Fri Oct 25, 2019 2:04 am ]
Post subject:  Re: Review: Get Strong at the Endgame

Robert, thanks for the answer and the advice.

Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group