Author: Sune Larsson
Date: 02:57:28 05/25/01
Go up one level in this thread
On May 24, 2001 at 23:28:27, Miguel A. Ballicora wrote: >On May 24, 2001 at 22:53:46, Uri Blass wrote: > >>On May 24, 2001 at 19:37:02, Miguel A. Ballicora wrote: >> >> >>>[D]5rk1/ppp3p1/7p/3p2P1/3Pb3/5RPP/PPP2RK1/8 b - - >>> >>>Black to move and win with h5. No zugswang here, white can play for ever >>>Rf1-f2-f1. Now there is no zugswang, but too deep to see a victory, only >>>evaluation can make it. Black King is loose to place itself in a good position >>>to trade everything in f3 to win a pawn endgame. Can pepito see it? >>>If it can, can win the game if you play against it? >>>Any other program? >> >>I am going to play h5 in a game because I am sure that black is not losing but >>I am not sure if black can win after h5 h4. >> >>White is a pawn up in the pawn endgame and I suspect that white can put the >>pawns in good squares by b3 and c3 so black cannot trade pieces in the right >>way. > >place the king in d6, play b6, c5, Kc6, Kb5, a5, a4. Create a hole in c4 >or force a passed pawn. I think that black wins. Anyway, it does not matter. >h5 is the only chance to keep an advantage. > >>I suspect that no top program can find h5 and the only programs that can find >>h5 are weak programs and the fact that they can find it is because of some bug >>or bad design decision that may be productive in this position but counter >>productive in other positions. >> >>Your previous test position was already too hard to find by search and this >>position is even harder to find by evaluation because in the previous position >>white can see by evaluation that only pawns can move when in this case the rook >>can also move. > >That is the idea... > >>I do not understnad why do you expect programs to see things that almost never >>happen in games. > >You don't need to undersant it, because I don't expect anything. Why >do you assume that I expect them to find these things? In fact, I expect >them to fail. The reason of my post was curiosity, because maybe there >is a program that sees it. > >Anyway, this motif _happens_ in games. A pin that paralyzes the opponent >is not so uncommon. In fact, I extract the position from a real game. >On the other hand, the problem is not this motif, is even bigger, it >is the lack of long term planning that hurts. That hurts most of the programs >many times in the endgames, where calculation sometimes is secondary and >the most important thing is logical thinking (kind of retrograde thinking). > >>Programmers usually care about winning games and not about your positions. > >"My" positions? I did not invent anything! In fact, you came up with >a position of your own. We are having fun here. At least I do. >Anyway, the programmers want to win. That is right, but if they were interested >in A.I. they would also approach some problems that are not that common. > >My point of wiew is simpler, I have fun with chess. > >Regards, >Miguel I join up with you, Miguel, regarding this attitude or philosophy. There is so much fun to have with this game. Even a bit strange sometimes how certain positions can catch your attention. Thanks for the position you sent. Sure it's a win for black - he will brake through on the Q-side with pawns+king. These kind of positions awake my dream of extended pattern recognition in chess programs. As you mention a kind of retrograde thinking where first a possibility of a known (and plausible) position is recognized - followed by pondering about the ways to reach this position. Knowledge about how to handle different pawn formations - attacking patterns (both on the Q-and King side) - and ,of course, endgame patterns. For example in an ending good knight+pawns vs bad bishop+pawns, it revolves around knowing where to put the pieces, rather than calculating variations. GM:s have a lot of these known patterns in their minds. And there are very good tutorials out there, f.e. Strategy 2.0 - which could be something of a gold mine for chess programmers. Just the other day I was reminded of my own lack of knowledge in chess. Played a game vs Gambit Tiger - opened with 1.d4 (which never happened in my tournament games) - found myself in a Leningrader Dutch with f5, d6, g6 - and when black played e5 I had to choose between a)dxe5 b)d5 c)leave the pawn on d4. And I simply didn't know... Okey, all 3 ways were surely playable, but I'm sure practice has showed the best way to continue for white, in this specific position. Not knowing meant I had to calculate, use my instinct and try to invent the wheel once more - and this in a 30 min game vs GT...;) Sune > > > >> >>Uri
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.