Author: Uri Blass
Date: 21:22:31 03/31/01
Go up one level in this thread
On March 31, 2001 at 19:55:23, Dieter Buerssner wrote: >On March 31, 2001 at 14:57:27, Frank Schneider wrote: > >>On March 31, 2001 at 09:22:57, Ferdinand S. Mosca wrote: >> >>>Hi everyone, >>> >>>A position from an actual game which caught my interest while >>>playing an engine match at G60. >>> >>> 3r4/5k2/5p2/6pp/5p2/1p6/p2R4/K3R3 b - - 0 1 >>>Is there any program that can evaluate this simple ending logically. >>> >>>The trap here is the move ...Rxd2, this may not win but it only draws. >>>There is perpetual check and stalemate if black plays ...Rxd2. Many programs >>>probably find the perpetual coming, but at what time do your favorite programs >>>give a 0.0 or near 0.0 score? >> >>Gromit 3.8.2 avoids Rxd2 after <1sec and prefers Rb8. > >Frank, this really is impressive. If you would like to enlighten us, how you can >achieve this with Gromit, I am certain, that you will have very attentive >listeners :-) > >-- Dieter I am interesting to know if programs can prove a draw by selective search when they let only check moves for white and evaluate positions when white is to move and cannot do checks as a draw only if it is a stalemate. It is clear that a good evaluation function can also help here(programs do not need to see the draw in order to avoid Rxd2 and it is possible to see that white has stalemate chances without search(it has only a rook to move)). A better evaluation can help programs not to search lines when white takes a pawn by Rxb3 because the program can evaluate the relevant positions as a win for black without search. Here is an example: [D]8/8/5p2/7p/4kp2/1p4R1/p2r4/K7 w - - 0 1 Program likes lines with Rxb3 in this position at least in the first plies. 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.