Author: Steven Edwards
Date: 13:51:54 05/28/05
Go up one level in this thread
On May 28, 2005 at 15:46:44, Matthew Hull wrote: >I'm curious -- do you do any game post mortems? I've seen several games where >GNUChess 5.07 (notalent) is down in score to Symbolic in the middlegame, by >sometimes more than 3 pawns, yet it seems to recover and win later in the game. >I guess this is where the speed difference is showing (700 MHz vs. 1400 MHz), >and Symbolic is getting significantly outsearched when the pieces start coming >off the board. Some of this is just chance, and some of it recently was due to a (now changed) tablebase probe trigger algorithm that was causing Symbolic to spend too much time on disk probes. This may need some more adjustment; however, it's only a toolkit issue and will have no effect in the cognitive search with a target of only 1,000 position nodes per search. I also note that Gnuchess still has the ply limit overflow problem that causes it to hang in some KPK endgames. It will lose a game on time but will be awarded a draw due to the opponent's lack of mating material.
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.