Life In 19x19
http://lifein19x19.com/

Endgame 3 - Accurate local evaluation
http://lifein19x19.com/viewtopic.php?f=57&t=16836
Page 1 of 1

Author:  explo [ Wed Aug 07, 2019 5:04 am ]
Post subject:  Endgame 3 - Accurate local evaluation

Last December I briefly reviewed Endgame 2. I bought the pdf versions of Endgame 3 and Endgame problems 1 in early June but later picked up on Robert’s offer to review them against a (physical) copy.

The books were very, very well packed and it took me a few minutes to unpack them. The printing and binding quality is also very good.

I had started Endgame problems 1 before the books arrived. I left it aside to get a full hang on the theory before going back to practice.

The introduction (chapter 1) quickly caught my attention. The positions below are the first examples of the book and part of the introduction.
Click Here To Show Diagram Code
[go]$$Bc Example 1
$$ +-------------
$$ | X X X X O .
$$ | . O O O O .
$$ | X X . O . .
$$ | . O O O O O
$$ | X X X X X O
$$ | . . . . X .[/go]

Click Here To Show Diagram Code
[go]$$Bc Example 2
$$ +---------------
$$ | O O O O . X .
$$ | O O O O . X .
$$ | . X X X X X .
$$ | O O O , X . .
$$ | O O O . X X .
$$ | . X X X X . .
$$ | O O O O X X .
$$ | . O . O O . .[/go]

In the first example, is it sente for black to connect 2 stones ? In the second example, is it sente for white to connect 6 stones ? I would (have) answer(ed) my opponent without question in both cases. Actually, both positions are local gote. Correct judgement of these kind of positions is the topic of chapter 3.

Chapter 2 quickly deals with the basic of miai counting, or modern endgame as Robert would call it. It’s a a good reminder if you forgot about some of the formulas, or you know about miai counting but you need to learn about Robert’s notations or vocabulary. It’s too short if you’re completly new to the topic, but to be fair you should be reading Endgame 2 before Endgame 3 in that case.

Chapter 3 is about judging whether a local endgame is sente or gote. We learn about 4 possible ways to do that with counts and move values. Each way is treated several times in examples. Overall the chapter was very clear to me, and determining the values and result for easy positions like the two examples was easy. Each method looked similarly efficient when working out examples and problems with pen and paper.

Chapter 4 is about discussing sente and gote options. We have positions where a gote and a sente sequences are both making sense. The objective is to decide which one is optimal. This chapter was also very clear and examples/problems were tackled without much trouble. The same position with a couple of stones moved here and there was used many times so that it was clear using instinct was a poor way to guess the answer.

Chapter 5 is a mix of several topics. First, privileges are briefly discussed. Nothing new for dan players I hope but worth mentioning for weaker players. Also good for the sake of clarity, as some examples uses the concept later in the book.
Next comes « double sente ». We use examples to see that it’s not (really) a thing, but we get practical advice on how to deal with situations we situation we perceive as double sente.
Last, we get introduced to traversal sequences, which will be the topic of the two remaining chapters. Traversal sequences are sequences that should be play in one go : interrupting the sequence can lose points. The follow up idea is to determine when play should be interrupted to play elsewhere on the board.

Chapter 6 presents a sure way to handle long sequences. The downside is that the calculus was a lot of work even for basic positions. I mostly got the examples and problems right so I think I understand the theory correctly, but I needed some time and some paper to do it properly. It was definitly too much time to consider trying it in a real game at my current capacity.

Chapter 7 presents another way to handle long sequences, which doesn’t always work if I understood correctly. That chapter wasn’t untirely clear to me and I started the first problems unsure of where to look at. I wasn’t as lost when doing the last problems but it’s still a bit fuzzy. I’ll give it some time to digest and I’ll give it another chance after. The method was clearly requiring less calculus, but I’m still unsure of being able to do it without a pen and paper.

I was satisfied with Endgame 2 and I am satisfied with Endgame 3. No rough start this time, as the last chapter was the only one giving me trouble. Even if I don’t think I can make use of everything in a real game, solving some of the positions was satisfying on its own.
The part about double sente is concrete advice which I think I can apply in my games.
Distinguishing gote and sente sequences also seems doable in my games during the microendgame and may give me the chance to grab some extra points over the course of a game.

In his review Robert said « The book is not for you if you die on seeing explicit calculations. » and I very much agree. The book doesn’t offer any tactical difficulties, but you’ll see a lot of values and variables. This is a theory book, but you should try to work out the values by yourself as much as possible so it’s quite a lot of work for a theory book.
On the other hand I don’t want to oversell the difficulty. The math level required isn’t high. Everyone who doesn’t want to run away after reading Endgame 2 or reading some of the endgame topics here should be fine.



Finally, the last paragraph is about chapter 3 and adressed to Robert and the other readers of Endgame 3. On page 50 it is said « If we prefer application of conditions 1, we need to learn these relations by heart » and it wasn’t my experience.
Conditions 2 was my intuitive view of local sente/gote so the relations made sense instantly.
Conditions 1 and 4 came next and they are closely tied in my eyes. The defender chooses if it is a local sente or a local gote. So on conditions 1, he picks the count most favorable to himself and declares the locale as such. On conditions 4 he attributes the lowest move value to his opponent.
Conditions 3 are not intuitive to me. I understand that in a local gote M_G<M_S (conditions 4) and F<M_G (conditions 2) thus F<M_S. I can do a similar reasoning in a local sente, but I can’t explain with words why the relation is as it is.

Arnaud Boucherie, somehow 2 dan at the French go federation and about 4/5 dan on Fox/Tygem.

Author:  RobertJasiek [ Wed Aug 07, 2019 7:15 pm ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

explo wrote:
« If we prefer application of conditions 1, we need to learn these relations by heart » and it wasn’t my experience. [...] on conditions 1, he picks the count most favorable to himself


You are lucky having a better experience with conditions 1. Have you already had it before reading the book or have you acquired it due to its advice of choosing the count most favorable to himself? It was about the last thing I added to the book. Before, my only applicable aid (the method of multiples would not help me in practice) was learning conditions 1 by heart. Then I discovered the clue just in time.

Author:  Bill Spight [ Wed Aug 07, 2019 10:29 pm ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

As Robert points out, the method of multiples is cumbersome over the board. For me, it is mainly a way of illustrating concepts. Con permiso, I thought it might be fun to show how it works for the first example. :)

The method of mulitples is a way of finding the average value of a position, not a play, per se. In this case, if the position is a Black sente, we know that its average territorial value is -8 (i.e., 8 pts. for White). But I'll go ahead and show the sente option, anyway. 4 copies of the position are enough to find the gote value. :)



The comparison of results shows that White does better not to reply to Black's play, so the position is gote. (We can tell by inspection that it is not a White sente. ;)) To find the average value of the position we divide -36 by 4 to get -9. :) Then if White plays first the result is -14, so we can figure how much a gote gains, namely -9 - (-14) = 5 pts. :)

We can actually use the idea of the method of multiples without having to visualize 4 copies and the play in them. All we have to do is calculate the local scores of these outcome positions.

Click Here To Show Diagram Code
[go]$$Bc Outcome 1
$$ +-------------
$$ | X X X X O .
$$ | B O O O O .
$$ | X X . O . .
$$ | B O O O O O
$$ | X X X X X O
$$ | . . . . X .[/go]


Local score = 0

Click Here To Show Diagram Code
[go]$$Bc Outcome 2
$$ +-------------
$$ | X X X X O .
$$ | W O O O O .
$$ | X X . O . .
$$ | B O O O O O
$$ | X X X X X O
$$ | . . . . X .[/go]


Local score = -8

Click Here To Show Diagram Code
[go]$$Bc Outcome 3
$$ +-------------
$$ | X X X X O .
$$ | . O O O O .
$$ | X X . O . .
$$ | W O O O O O
$$ | X X X X X O
$$ | . . . . X .[/go]


Local score = -14, which we count twice. :)

Author:  explo [ Thu Aug 08, 2019 2:49 am ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

RobertJasiek wrote:
You are lucky having a better experience with conditions 1. Have you already had it before reading the book


Before reading the book I only knew about conditions 2. I had the intuitive idea of conditions 2 for some time and it seems I picked the explicit conditions in Endgame problems 1 where it is mentionned and I had read that part in June. Correct me if I'm wrong but I can't find any condition in Endgame 2 so I couldn't pick it there. Endgame 2 had clear positions where determining sente/gote intuitively worked just fine.

RobertJasiek wrote:
or have you acquired it due to its advice of choosing the count most favorable to himself?

I can't find that advice in chapter 3. When reading page 50 I remember being surprised by your sentence about memorising the conditions. I had already acquired by myself the idea of picking the most favorable count or move value but I was worried that it would not work in more complicated positions. It worked just find for positions with both players' follow ups and I become more confident about it.
It later found echo in the pages about multiples. You use the phrases "sente strategy" and "gote strategy". The word strategy make a lot of sense with the idea of picking the more favorable count.
The advice of choosing the more favorable count is a the start of chapter 4: "the creator choosing the option of the more favorable count helps us to recall the conditions" (page 72). I was confused because the same sentence about the defender would have fit well in page 50! Page 72 it could also be emphasized that it works with the second condition: the creator picks the more favorable move value.

Author:  RobertJasiek [ Thu Aug 08, 2019 3:37 am ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

Chapter 3 does not have such a statement because, as it becomes clear in the section on multiples, it is related to the second moving player's choice. IMO, adding a statement at the beginning of the chapter would have confused more than help, but your opinion differs.

Endgame 2 avoids all such conditions and uses examples with obvious gote / sente types indeed because the conditions and related theory appear in Endgame 3. Yes, Endgame Problems 1 uses conditions 2 / 3 (and some conditions for long sequences as necessary.)

Author:  explo [ Thu Aug 08, 2019 6:02 am ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

RobertJasiek wrote:
Chapter 3 does not have such a statement because, as it becomes clear in the section on multiples, it is related to the second moving player's choice. IMO, adding a statement at the beginning of the chapter would have confused more than help, but your opinion differs.

My opinion may not be different. I understand that such a statement at the start of chapter 3 could bring more confusion than help. But in my case, your statement about needing to learn conditions 1 by heart brought some confusion (others may not be confused). I don't see how it offers some help. In my opinion, not saying anything might have been better. If people can't make sense of any of the conditions by themself, you've already hinted towards learning conditions 2 and 3, the ones Bill Spight emphasizes (page 37).

Upon writing the review and reflecting on the book, that was the only thing that bothered me (slightly). In my review of endgame 2 there was a lot more stuff that bothered me, and that I would have removed or moved elsewhere in the book.

Author:  RobertJasiek [ Tue Dec 31, 2019 12:30 am ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

Reference: https://lifein19x19.com/viewtopic.php?p=252408#p252408

xela wrote:
Click Here To Show Diagram Code
[go]$$B White to play
$$ X X X O O O
$$ . . . 1 2 .
$$ -------------[/go]

because for black to play 1 and then abandon the position is a loss for black. [...]
I haven't read it yet, but I believe Robert Jasiek's Endgame 3 has some theory on "long sequences" which goes into this in more depth.


See Problems 1 + 2 (pp. 155+156) in Chapter 5.12 and the theory (pp. 159-168), Example 4 and Diagram 4.3 (pp. 176-179) in Chapter 6. Bill has also done some analysis of the shape here. The (ko threat) exchange loses 1/6 of a point.

Author:  RobertJasiek [ Thu Oct 15, 2020 2:49 pm ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

In https://www.lifein19x19.com/viewtopic.p ... 70#p260570 Gérard TAILLE asks for evaluation of double sente moves. I have described their evaluation on pp. 100 - 120 in the book Endgame 3 - Accurate Local Evaluation. One possible approximative calculation is

M = (B - W) / 4,

where B and W are the counts after two successive Black or White plays, respectively. In an environment with temperature T, one can then compare

M ? T

and have a fair exchange if M ~= T, answer locally if M >> T or ignore if M << T.

Author:  RobertJasiek [ Sat Oct 17, 2020 7:44 am ]
Post subject:  Re: Endgame 3 - Accurate local evaluation

Counter-example 2 on p. 101:

Click Here To Show Diagram Code
[go]$$B no local double sente
$$ -----------------------
$$ | . O X . X . O . O X .
$$ | . O X X X . O O O X .
$$ | . O X . . . . . O X .
$$ | . O X X X . O O O X .
$$ | . O O O X X O X O X .
$$ | . O O O O O X X X X .[/go]

Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/