Author: Bas Hamstra
Date: 17:01:05 06/11/01
Go up one level in this thread
I hope it isn't blasphemy, but I hate endgames. I get totally bored and depressed when queens are traded. On the other hand, *losing* a bloodless endgame makes me want to die. Since that's worse, I have to work on endgames. The thing that I want to fix is exactly this: I have a passer, you have a passer. But yours is miles away and when it starts to walk my king *must* go there and your king eats all my pawns. Game over. My program sees that sort of thing too late. And when it's a matter of distant 3-2 majority, it doesn't see it at all. When you see it happen you immediately see that such a situation is hopeless. But when you try to pinpoint exactly under what circumstances it is hopeless, it is not that easy. Crafty is good at this, so is YACE. Best regards, Bas. On June 11, 2001 at 11:55:22, Robert Hyatt wrote: >On June 11, 2001 at 11:27:18, Bas Hamstra wrote: > >>Can anyone point me to an endgame suite that is better than Fine? I have a Fine >>suite with 48 positions, of which Tao solves 42 or so. Still, in games against >>YACE and Crafty it gets killed because if has no idea of outside passers and >>majorities. And no table bases, but I think that is no big deal. And hardly a >>KPK evaluator. >> >>I made a first start, in giving passers boni proportional with filedistance from >>the enemy king in a pawn endgames. Is there more to it? >> >> >> >>Best regards, >>Bas. > > >there are lots of things to think about. > >where is the passer? Other pawns? IE a passer on the a file is easier to win >with (pawns on say g and h files too) than if the passer is on the e file with >pawns on g and h files. (distant passer). Ditto for candidate passers if you >don't have passers. Their value should go up as pieces come off, the value >should not be "static". > >If this is a king and pawn ending, then "split passers" are far better >than connected passers (many totally overlook this... I did too until a >GM pointed it out one day a year or so ago). IE make a position with one >side having two pawns on the d/e files, the other side having pawns on >the b/g files. Which side do you prefer? If you prefer the d/e files, you >have tuning to do. :) > >Another important issue is "where does the king belong"? This is pretty hard >to answer with simple tests, so go for getting it right, rather than worrying >about the cost. If you have all pawns on the queenside, your king can't be >on the kingside... > >You can't just do "pawn-count = majority" either. You need to recognize >a majority that is immobolized. IE you are black with pawns on a7, b6 and >c5, with your opponent having pawns on b5 and c4. To use that majority will >require help from your king, which might take it away from where it is >needed. > >It is complex. But not impossible to do at all...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.