Life In 19x19http://lifein19x19.com/ Values of moveshttp://lifein19x19.com/viewtopic.php?f=10&t=16073 Page 5 of 9

 Author: mitsun [ Mon Sep 17, 2018 4:38 pm ] Post subject: Re: Values of moves bernds wrote:The way I see it - if Black plays there, he has two points, and if not, it's fifty-fifty whether Black gets a point or not. So that would make it a 1.5 point gote. One could also imagine a situation where nothing else is on the board, in which case it would be a 1 point sente (from White's perspective).Yes, exactly correct! A few minor comments: O Meien and others prefer to calculate the value of a single move, rather than the difference between a pair of moves (B first versus W first), so they divide your result by two.If the move is truly sente for W, the probability of W playing first becomes 100%, as does the probability of B answering, so the value of the starting position is 1, and the value of the move is 0 or meaningless. If a W play here would be sente, but it is B turn to play, then for B to play here first is reverse-sente, and the value of the move is estimated as twice the normal gote value.

 Author: RobertJasiek [ Mon Sep 17, 2018 9:50 pm ] Post subject: Re: Values of moves daal wrote:As far as I gather, assigning a count means to claim that black has x points even though the position is unfinished. This is done by averaging the possible outcomes. In the above example, either black plays at a giving him two points or white plays there creating a second situation in which black could get either one or zero points. This second situation is interpreted as meaning that black has half a point. This value allows us to get an average for the original position, which is (2 + 0.5): 2, or 1.25. So when we look at this unfinished position, we can give it a value , a count, of 1.25. From this, we see that if black plays at a, he gains .75 points. Likewise, white would also gain .75 by playing there. That is what the move is worth, and that number can be compared with other similarly derived numbers to determine the biggest move. The count has a further significance which is that we can add them up to see who is ahead. Corrections welcome.This, and mitsun's description of it, apply for a local gote with one player's gote follow-up.

 Author: RobertJasiek [ Tue Sep 18, 2018 3:09 am ] Post subject: Re: Values of moves You assume that everybody would always use their specific degree of Mathlish but that is not so. A person can vary the degree from informal English via weak or strong Mathlish to formal mathematics.In the context of degree of Mathlish, you emphasise O as somebody using lots of diagrams to explain something. However, I tend to use more diagrams although my average degree of Mathlish is higher. There is no identity between degree of Mathlish and numbers of diagrams.You advertise your preference for a low degree of Mathlish but do not expect everybody to become a couch potato. For calculations and evaluations involving numbers, a high degree of Mathlish and mathematical annotation have the great advantages of clarity, accuracy and easy access to verification of correctness. Furthermore, variables and equations allow generalisation and recognition of application of general methods at a glance. When different kinds of numbers are calculated, the consistent naming of variables helps. Not by chance do we use T for the temperature in physics or T for the temperature in endgame theory.One of the worst achievements of some culture is appreciation of natural language versus scepticism of mathematical language. Its clarity is an advantage - not something to be ashmed of. Correctness of move values is something to be proud of.

 Author: bernds [ Tue Sep 18, 2018 3:49 am ] Post subject: Re: Values of moves mitsun wrote:bernds wrote:The way I see it - if Black plays there, he has two points, and if not, it's fifty-fifty whether Black gets a point or not. So that would make it a 1.5 point gote. One could also imagine a situation where nothing else is on the board, in which case it would be a 1 point sente (from White's perspective).Yes, exactly correct!Well, good. I was starting to doubt my own sanity, and possibly that of others.But I only looked at terminal positions, and as far as I can tell so did you, and I explicitly asked for an example where that is not possible, since Bill claimed that in general you can't.Quote:O Meien and others prefer to calculate the value of a single move, rather than the difference between a pair of moves (B first versus W first), so they divide your result by two.Yeah, OK, I had gathered that the point values come out as half of what you'd expect. What I'm trying to figure out is - why the emphasis on position values? What is being communicated when, for example, Bill says the value of A in the original position is zero? I'm assuming there has to be some deep insight because on the face of it it's just doesn't sound very helpful.Quote:If the move is truly sente for W, the probability of W playing first becomes 100%, as does the probability of B answering, so the value of the starting position is 1, and the value of the move is 0 or meaningless. If a W play here would be sente, but it is B turn to play, then for B to play here first is reverse-sente, and the value of the move is estimated as twice the normal gote value.It still seems strange to consider sente moves as having size zero. How do you compare them? The thing about reverse sente is of course standard and comprehensible to even someone like me who gets his endgame foundations from GSatE.

 Author: John Fairbairn [ Tue Sep 18, 2018 4:28 am ] Post subject: Re: Values of moves Quote:Its clarity is an advantage - not something to be ashmed of.You really don't get it, do you, Robert?It's NOT clear - to us. That's the whole point. It's only clear to you because there is a whole lot of axioms, principles, mathematical grammar, modes of expression, jargon, abbreviations and so on behind it that you mathlanders understand. We lesser mortals haven't got this apparatus and don't want it, not because we couldn't learn it but because we haven't got the time or the need to learn it.I could say something to you in Japanese. It would be totally opaque to you but it would be totally clear to me and other people who knew that language. That doesn't mean you can't learn Japanese, but you've already told us many times you haven't got the time, the need or the motivation to learn it. The result is you haven't got the apparatus to understand it. I respect that and therefore speak to you in English. But you talk to me in mathlish. Why? That's rude.It may, in context, be rude, but speaking mathlish is nothing to be ashamed of, no more than speaking Japanese is shameful, and no-one has said it is. What is shameful is stubbornly not listening to other people when they say what you say is hard to follow because it's in a "foreign" language.

 Author: RobertJasiek [ Tue Sep 18, 2018 6:38 am ] Post subject: Re: Values of moves bernds,in the go game, we have positions and moves. Each move transforms a position into another position. Therefore we study properties of positions and properties of moves. In endgame evaluation, we study values of (local) positions aka counts and values of moves. When studying a single move, it transform the count of the preceding position into the count of the resulting position and the considered value of the move is its gain. The gain expresses the change in counts caused by the move. When relating Black's move or sequence started by it to White's move or sequence started by it, we consider another kind of values of moves, the move value.Positions, moves and their values are all related. Therefore we consider them all. In particular, gains and move values are derived from counts.What is the result of a sequence? The count of the resulting position. We are still interested in its value so we do not discard it just because of having calculated the move value of the already executed move. For decisions among different moves, we would compare different resulting counts.Already for these basic applications, counts are essential.

 Author: Knotwilg [ Tue Sep 18, 2018 8:00 am ] Post subject: Re: Values of moves John Fairbairn wrote:Quote:Its clarity is an advantage - not something to be ashmed of.You really don't get it, do you, Robert?It's NOT clear - to us. That's the whole point.There is ability and there is motivation. Myself I'm a mathematician and can be expected to easily follow the discussion but I don't because precise endgame calculations do not interest me. I suspect you are interested in O Meien's endgame treatises and its impact on the world of (professional) go from a cultural perspective. When required to acquire "mathlish" to understand the technical discussions, you lose your motivation. I don't think my intellectual capabilities to acquire the lingo used here are substantially higher than yours. I think your motivation is substantially higher than mine and so I agree with Robert that you will find it quite easy to acquire the vocab of "count" or "follower".I recently studied Remi Coulom's paper that contributed to the AlphaGo shockwave, not because I have a predisposition for computer science, but because I was motivated to understand more about machine learning and the inner working of Leela Zero. There's a big amount of "complish" in there but I persevered.Incidentally, I don't second your comparison with Japanese. Other languages substitute each other. Math, mathlish or complish are extensions of English, necessary to convey the subtleties of the domain. Your desire to acquire expertise on the endgame through common English maybe essentially a frustrating one. I'm reminded of poor old Galilei's writings, which didn't have algebra at their disposal yet. Today, 14 year olds with only a fraction of Galilei's brain are better at expressing his laws to their peers than he was to his, because they have all acquired the language (algebra) without too many quibbles (well, ok).

 Author: Bill Spight [ Tue Sep 18, 2018 8:29 am ] Post subject: Re: Values of moves Knotwilg wrote:Incidentally, I don't second your comparison with Japanese. Other languages substitute each other. Math, mathlish or complish are extensions of English, necessary to convey the subtleties of the domain. Your desire to acquire expertise on the endgame through common English maybe essentially a frustrating one. I'm reminded of poor old Galilei's writings, which didn't have algebra at their disposal yet. Today, 14 year olds with only a fraction of Galilei's brain are better at expressing his laws to their peers than he was to his, because they have all acquired the language (algebra) without too many quibbles (well, ok).I once read some translations of what are now considered simple algebra problems from Arabic texts of the Middle Ages. They would start like this: "Heap, it's third, . . ." Mathlish makes it so much easier.

 Author: John Fairbairn [ Tue Sep 18, 2018 8:59 am ] Post subject: Re: Values of moves Quote:RJ: "rude", "stubbornly not listening": I do not join personal meta-discussion.Oh, yeah?Quote:RJ: You spread fear for principles to be learnt.

 Author: RobertJasiek [ Tue Sep 18, 2018 9:00 am ] Post subject: Re: Values of moves Bill Spight wrote:if a position is symmetrical for both players, its average value is 0.Maybe bernds was concerned with one side having 8 and the other side having 6 points; why would you get the average 0. Answer: choose a small locale in which the hane-connect endgame is counted; its average is 0; only afterwards add the extra points outside the locale. We do so because the hane-connect shape is always the same so we can recall its count 0. It does not matter whether the settled parts of the adjacent territory regions have different shapes in different occurrences of hane-connect.

 Author: Bill Spight [ Tue Sep 18, 2018 9:07 am ] Post subject: Re: Values of moves To continue, after a breather. John Fairbairn wrote:The reason I belabour this point is on behalf of people like daal and myself. O does not speak in mathlish. Mathlish is not just a dialect with different words. It's a different grammar, with a different way of ordering things. The people who mainly speak about boundary plays and counting here order things differently from daal, myself and O. The fact that people like yourself also understand O (and no doubt far better than I can) and also understand English does not mean you are not reverting to mathlish when you speak to us.It seems to me that O does speak in mathlish. That is, he introduces technical terms and defines them. As is appropriate to his audience and the purpose of his book, he does not delve into the difficulties that Robert, moha, myself, and others do in our discussions here. It is not mathlish per se that makes these discussions difficult, it is their subject matter. Quote:Here are a couple more of the ways of expressing himself that made O appealing to me. What he said is old hat to you. To me it was as if someone opened the curtains and let the daylight in.{snip}Quote:In cases where a reverse sente and a gote boundary play of apparently similar size are bound up together, the procedure is:[1] Calculate the deficit as a number of points disregarding sente and gote in the case where you play the reverse sente;[2] If it appears that you would recover that deficit with the next boundary play, play the reverse sente. If you would not recover it, make the gote play instead and so maintain your advantage.Sounds like mathlish to me. How does he define deficit, what does he mean by disregarding sente and gote? And anyway, the general question of when to play gote vs. reverse sente is difficult.Quote:It's quite rare in Japanese to have people writing mathematically about go. Here the mathematicians seem to predominate.I think that there may be cultural reasons for that. Without a long history in the West, ordinary people are not drawn to go. Hippies were, for a while, until, as a friend pointed out, they actually had to think. For some reasons, mathematicians were, as far as I can tell. In my case, despite being fairly strong for a Western amateur, and having written a go newsletter way back when, I cannot claim any real go expertise. But in the mathematics of go, I am a world class expert. Quote:PS As an example of English vs mathlish, take the following from earlier in the thread:Quote:If the move is truly sente for W, the probability of W playing first becomes 100%, as does the probability of B answering, so the value of the starting position is 1, and the value of the move is 0 or meaningless. . . .I confess I am not quite sure what is meant.

 Author: Bill Spight [ Tue Sep 18, 2018 9:10 am ] Post subject: Re: Values of moves RobertJasiek wrote:Bill Spight wrote:if a position is symmetrical for both players, its average value is 0.Maybe bernds was concerned with one side having 8 and the other side having 6 points; why would you get the average 0.I think he was talking about daal's original diagram and question.

 Author: RobertJasiek [ Tue Sep 18, 2018 9:19 am ] Post subject: Re: Values of moves Bill Spight wrote:in the mathematics of go, I am a world class expert.Aye.