Author: Robert Hyatt
Date: 10:01:09 09/20/05
Go up one level in this thread
On September 20, 2005 at 00:05:36, A. Cozzie wrote: >[D] 6k1/6p1/6K1/7P/8/5B2/8/8 w - - 0 1 > >Zappa (incl. 1.1) has the old 'bad bishop in the corner' trick that humans have >used to embarrass computers for years. However, this position has a twist: > >White can play Be4 Kh8 Kf7 g5 (only move) hxg6. This gets White's pawn off the >h-file. Of course, it also stalemates Black, but you can't have everything. >Zappa doesn't detect stalemate in eval(), so its strategy is always "shuffle N-5 >moves, then play Be4 Kh8 Kf7 g5 hxg6 and get the pawn of the square, and Black >has no move to find he's stalemated". Crafty detects stalemate in the >evaluation, so it gets this one. > >kburcham++ > >anthony I don't remember when this came up, but it was probably within the last year or two. Basically the same "theme" where there is either a way to force the h-pawn (wrong color due to the bishop) off the h-file by forcing the opponent to have to push his g-pawn as his only legal move, but then you make the wrong evaluation because if you take his g-pawn, you now don't have a drawn ending due to wrong-colored bishop, but you stalemate the opponent who now has no legal move, the reason he had to push his g-pawn in the first place. I simply added code to handle this one special case, which solved the problem. I think this came up in one of the many B+RP (wrong color) discussions that pop up from time to time, particularly those with multiple rook pawns of the wrong color that many get wrong where they get the single wrong-rook-pawn case correctly. Crafty was getting them all right since I don't care how many rook pawns there are due to the bitboard evaluation trick of just looking for a pawn (or pawns) on just the wrong rook file, so someone came up with that one which promptly broke my code. I took it as a challenge to fix it, and the fix was easy and didn't cost anything in terms of speed...
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.