Hol Igor Attempts On Sente Turned Gote/ Discussion

Gronk: Suppose I wanted to understand these endgame ideas, with proper notation and with the hope of applying some of them in my play. Suppose further that what I find on SL is too brief and/or skips too many steps and/or doesn't start basic and slowly enough. What should I do? What should I read? Does a book or article exist that might help? Thanks.

Charles Well, I think that pretty well all treatments in print are pretty bad. There may be some reasonable magazine articles, somewhere, on parts of it. Amateurs tend to switch off when confronted with detail on it, though; so I suppose most people gain understanding by working it out for themselves.

No reason that SL shouldn't have better treatments, of course. One reason why I broached this whole area.

Gronk: By the way, what I think would be perfect is something along the lines of the bookCountingLibertiesAndWinningCapturingRaces. I think that is a very nice book, taking one more-or-less coherent topic and going through it at reasonable pace in a way that is (IMHO) reasonably accessible. It has problems to work out along the way to check one's understanding.

Bill: Gronk, when you talk about "endgame ideas", that's pretty general. There are lots of endgame ideas, and different kinds. It seems to me that the existing literature is very good on endgame tesuji, so-so on endgame counting, and very poor on order of play, except as it relates to tesuji. Can you be more specific about what you would like to see better explained? Thanks. :-)

Gronk: Tesuji--no, too easy to learn in other ways (just by watching or playing others, for example). Counting--yes, but just a bit since at some point this comes down to a combination of tesuji knowledge and reading and hingest on ==> Order of play--YES, and in great detail, with problems. This is hard to find elsewhere, not trivial to figure out on one's own, and not easy to pick up in casual play. I was under the impression that this is the topic about which HolIgor wrote, more or less, when to take reverse sente, gote etc.

Charles Well, yes. I get the impression that the Japanese literature for amateurs doesn't care (the Korean too, IMX, which is more surprising to me). I may have this wrong, but isn't the first problem in Get Strong at the Endgame misleading about this?

Charles I think endgame is in a sense harder than counting liberties (which is a paradox). As all this discussion shows, it is hard to make unqualified true statements; and if ko is involved it all gets much worse. We know that Richard Hunter wrote his book by collating existing Japanese literature, and filling in the gaps, to get a sound treatment. What Bill is saying (and I agree, FWIW) is that if you did the same for the close study of the endgame, there would be a sort of void in relation to the kind of question I at least find basic.

I was actually wondering about writing something about this on Wikipedia, since the mark-up there admits more mathematical notation (and it might progress go coverage there from endless debate about the rules).

Bill: Unlike Charles, I do not find this topic to be basic. If the answer to the question of the order of play is to read it out, well, that's basic, I suppose. But when you introduce ideas to avoid just reading it out, things get advanced pretty quickly. The idea that miai cancel out if there are no kos is basic. The idea that the biggest play is best is basic, but not always right. Usually right, but not always. After that, the idea of tedomari introduces exceptions. That leads us to the study of go infinitesimals, and that is already a bit tricky, but rewarding. And we have just scratched the surface.

Gronk: What I like about what Richard Hunter has done in CountingLibertiesAndWinningCapturingRaces is he's started quite basic and worked up slowly to the harder stuff. Anyone can stop reading at any point and still come away with something useful. Put another way, the book isn't front-end loaded with a lot of theoretical concepts that don't get used until later. From page 1 it is all value added. Perhaps the same (or similar) cannot be done with order of play in the endgame (or whatever), but since I do not understand the topic, I cannot say.

Charles Anyone who wonders why we're still discussing this after much previous posting to SL on endgame topics can compare traditional analysis of small endgame and CGT analysis of same small endgame. Actually my attitude has hardened since then. I think the CGT approach isn't really suitable (yet) for anyone who isn’t a grad student in a mathematical discipline, and prepared to work through a pile of definitions before having any chance of seeing a practical improvement. Things have improved since they always used to say 'read Winning Ways and then come back'. In conversation with Elwyn Berlekamp he told me that WW is essentially irrelevant to mathematical go; and I think that's correct to first order. CGT is clearly not irrelevant – it is the 'ultimate' theory we are likely to have. But it has been a pedagogic disaster in relation to most go players; and I have to say the coverage here is not all it could be. I pulled together the current path on it, mostly because the remarks were dotted all over the site. It all needs a thorough edit. I have to say that although I'm aware of what atomic weight is supposed to do for you, and have looked into it more than once, I still couldn't really explain it to anyone else.

So, I feel with on the one side the Devil – the paternalistic attitude fairly obvious in the Asian literature, that you get to 5 dan and then you can start worrying about the endgame – and on the other the deep blue sea of CGT, the kind of modelling I prefer to work with gets very little look-in. I would say that it corresponds to what Gronk is requesting.

Bill: I agree that the CGT approach is not very useful, if that means learning CGT first, and then applying it to go. OTOH, I think that applying the lessons of CGT to go does not require all that, and that doing so without getting bogged down in the math can help make the traditional approach clearer and simpler. I do not see any real tension between the traditional approach and CGT. (If the kind of modeling Charles is talking about is exemplified in stacks of coins, it is the kind that I also worked with for many years. But I have switched to a model based on temperature. And that's a very small change, since temperature was implicit in my original model, anyway.)

DougRidgway Things I've found useful: CGT, CGT becomes hard currency (Matthews), the theory articles from Setpiece kos (Matthews), Generalized thermography (Mueller Berlekamp Spight, includes a nice but concise intro), Economist's view of combinatorial games (Berlekamp), and even Martin's talk on temperature discovery and Bill's talk on NTE. I don't recommend Mathematical Go Endgames unless you have a copy of Winning Ways open in front of you. It's probably worth emphasizing that thermography and environmental go is an approximation, and cannot be strictly derived from CGT and the game tree. That means that it's occasionally wrong, but it's also easier to understand and apply than the exact theory at T=1 with infinitesimals etc.

Gronk: Thanks! So, these will help me with my endgame play? That's my goal. I don't care one lick if I understand CGT but if that is what is required to achieve my goal, well...I asked for it. :)

Bill: Gronk, assuming that you understand miai and counting reasonably well, I would suggest taking a look at tedomari, go infinitesimals, and difference games. I wrote that material, not with an eye to math so much, but rather with an eye to practical go. I would be interested in your reactions to that material. :-)

Charles The best actual practical advice I know, as a single point and relating to the endgame, is not to damage your endgame prospects by counting too much on middlegame attacks, i.e. play the fighting on the assumption that the endgame will really come around. This seems to be important around shodan level.