Antti Törmänen’s second book ’Rational Endgame’ published

[lichigo]

Super interesting book, difficult but great. I always studied endgame with the miai counting so it is hard to break this habit. What are the advantages of using the modern way(deiri counting)?
Is the miai counting less accurate?

[RobertJasiek]

lichigo, you confuse miai / deiri.

The advantages of modern endgame theory (and its move values) include:

- modern endgame theory is extremely more deeply developed than traditional endgame theory,

- for every application of modern endgame theory beyond the scope of traditional endgame theory, the former simply works while the latter then best works only after first converting traditional to modern move values, i.e., by actually using modern endgame theory,

- modern endgame theory simply allows application of the principle of usually playing in order of decreasing move values while traditional endgame theory first requires the still missing calibration of move values,

- modern endgame theory is very consistent: for simple gote with or without follow-up, the move value equals the gain; traditional endgame theory does not have this consistency but, if its move values are used anyway, relations need artificial factors for gote and case analysis because sente does not need the same factors for the same relations,

- modern endgame theory allows easier comparisons of every two values because they are already calibrated.

Traditional endgame theory has only one advantage: comparison of gotes is faster (because the calibration by means of division by 2 for gotes is not used) when comparing only(!) gotes to each other. For all other purposes, modern endgame theory with its calibrated move values works better. So much better indeed that traditional endgame has led to countless (accidental or systematic evaluation) mistakes while modern endgame theory leads to only infrequent accidental mistakes.

[Bill Spight]

Super interesting book, difficult but great. I always studied endgame with the miai counting so it is hard to break this habit. What are the advantages of using the modern way(deiri counting)?
Is the miai counting less accurate?

As Robert said, you have miai and deiri backwards. For simple comparison of plays, with the corrections to deiri, they are equivalent. Otherwise, the gain (miai value) is more accurate.

Here is an example like one that I tried to argue a 5 dan out of years ago.

Code: Select all

               A
              / \
             B   0
            / \
           10  5

/ stands for a Black move, \ stands for a White move. The scores are from Black's point of view.

The 5 dan correctly identified the move from A to B as worth 7½ pts. by traditional deiri values. He also correctly identified the move from B to 10 as worth 5 pts. by deiri values. So far so good. However, he thought that by playing from A to 10 in two moves Black gained 12½ pts. more than if White played from A to 0. All you have to do is look at the game tree to see that that is false. All Black gains by comparison is 10 pts.

Gains (miai values) give the correct answer with no confusion. The move from A to B gains 3¾ pts. and the move from B to 10 gains 2½ pts. for a total gain for Black of 6¼ pts. The move from A to 0 gains 3¾ pts. for White, so the gain for Black for playing twice versus White playing once is 6¼ + 3¾ = 10 pts. As advertised.

OC, the 5 dan could have seen directly from the game tree that 10 pts. is the right answer, but deiri counting confused him.

[lichigo]

Thank you for the answers, I'm consider to buy endgame 2 values to learn more about it. At the moment I'm not fully understanding the difference but in some parts I do.
Could you show some real examples (diagrams) of real difference between the 2 ways ?

[Bill Spight]

Thank you for the answers, I'm consider to buy endgame 2 values to learn more about it. At the moment I'm not fully understanding the difference but in some parts I do.
Could you show some real examples (diagrams) of real difference between the 2 ways ?

The diagrams make little difference. The difference comes when you try to do arithmetic with the values. That works with gains, but not ordinarily with swing values. Most of the time, people are content with simply comparing the values of two different plays. But where they get confused is when they try to add and subtract values. The example with the 5 dan illustrates that. Or suppose that someone counts the game and finds that they are 3 pts. behind on the board, but they have a play with a swing value of 4 pts. They think that by playing it they can win. Wrong! The gain is only 2 pts., so playing it leaves them 1 pt. behind.

What you are asking is to show the confusion of the players. You can't do that just in a diagram, unless it shows playing mistakes caused by the confusion.

[RobertJasiek]

Imagine a simple gote with Black's move achieving the count 1 or White's move achieving the count -1. (Exercise: put this on the board.) The initial position has the count 0.

Modern endgame theory: each move gains 1, which is the move value.

Traditional endgame theory: the move value is 2. Gains were neglected but if we considered them, we would find that they are inconsistent with the move value.

[lichigo]

Thank you, gonna buy your book for more studies ^^

[Pio2001]

Anyone care to comment on how they like the book? Good buy?

I've got both Antti Törmänen's Rational Endgame and Robert Jasiek's Engame 2 - Values. But I am still completely confused with the theory in them.
I can understand what happens with double gote moves, but the rest is still above me.

[Kirby]

I’d say I don’t actually have much practical experience in the endgame. Many of my games are decided by resignation. I guess I could think about move values before then, but I’m too busy trying to read variations.

Then again, maybe I’m just making an excuse not to learn the math.

[RobertJasiek]

How about reading the books slowly and taking your time? Endgame evaluation is not like most other go theory but we must learn its own conceptual framework from scratch. If you find simple gote easier than other types of local endgames, this may be because simple gote depends on only two follow-up positions while other types depend on more. Understanding more can be more difficult than understanding two so be patient with yourself.

[Bill Spight]

I’d say I don’t actually have much practical experience in the endgame. Many of my games are decided by resignation. I guess I could think about move values before then, but I’m too busy trying to read variations.

Then again, maybe I’m just making an excuse not to learn the math.

The purpose of the math is to make reading easier.

[yakcyll]

I’d say I don’t actually have much practical experience in the endgame. Many of my games are decided by resignation. I guess I could think about move values before then, but I’m too busy trying to read variations.

Then again, maybe I’m just making an excuse not to learn the math.

The purpose of the math is to make reading easier.

Personally, I don't think it makes reading easier; more like it dispenses with reading in favor of 'preparing' the endgame before the game begins in the form of analyzing and remembering particular formations and their values. I'm starting to believe that this is the way to go about it, given how complex some positions can be. If I got it correctly, Antti seems to share this sentiment when talking about finding the compromise between precise counting and managing time during tournament games. Calculating and manipulating a tree of formation-value pairs during a match doesn't quite strike me as natural.

[Bill Spight]

I’d say I don’t actually have much practical experience in the endgame. Many of my games are decided by resignation. I guess I could think about move values before then, but I’m too busy trying to read variations.

Then again, maybe I’m just making an excuse not to learn the math.

The purpose of the math is to make reading easier.

Personally, I don't think it makes reading easier; more like it dispenses with reading in favor of 'preparing' the endgame before the game begins in the form of analyzing and remembering particular formations and their values. I'm starting to believe that this is the way to go about it, given how complex some positions can be. If I got it correctly, Antti seems to share this sentiment when talking about finding the compromise between precise counting and managing time during tournament games. Calculating and manipulating a tree of formation-value pairs during a match doesn't quite strike me as natural.

Precise calculation of average values of complex positions is for analysis. During play I do not recommend that amateurs strive for more accuracy than ¼ pt., as a rule. Yes, situations do arise where the players can recognize that a very small difference matters, or may matter, and can do the math over the board. If precise calculation mattered much, you would not see so many math mistakes in endgame texts, Tormanen's and Jasiek's books being exceptions. But if the calculations did not matter, you would not see them in endgame texts at all.

Besides, there is more to endgame math than calculating average values.

[RobertJasiek]

If values are very different, the larger ones are obvious and calculation is not needed. If values are similar, calculation is necessary to decide them. This is often so because there are many local endgames and similar values can occur easily.

I have often reflected whether approximation might be useful for rather similar values but I have almost always come to this conclusion: It takes more time to verify that an approximation is correct, applicable and does not create mistakes in decision-making than it takes time to calculate precisely. If approximation rounds to, say, 1/4 and allows 1/4 point mistakes, this is unacceptable because they are too large by far as they can accumulate over the very many endgame moves. Verifying which rounding to the nearest 1/4 is correct is hard. Rounding to integers might be feasible concerning time-consumption but is totally useless imprecision whenever similar values are involved. Instead of searching excuses allowing approximation, it is better to practise correct calculations with fractions and accelerate them.

There are situations for which values replace reading and others for which values very greatly simplify reading. Reading does not become superfluous but, in general, values very much accelerate reading.

EDIT: The middle game is different. Endgame values or positional judgements can often be rounded to, say, 1/2 because many other imprecise aspects, such as fighting or influence, also have an effect on decisions. However, when everything compared is local endgames characterisable by values, then approximation is dangerous.