Author: Tony Werten
Date: 05:07:04 08/01/01
Go up one level in this thread
On August 01, 2001 at 00:19:13, Dann Corbit wrote: >I am testing a tool written by Les Fernandez. It takes a given chess position, >and transforms it into rotated or mirrored images. For each position, the >evaluation should (obviously) be identical. Also, you would think that the >depth in plies reached would also be the same. Funny, but it does not always >work out that way. Sometimes, the chess programs I test will even choose a >different best move than the theoretically predicted one. > >How does your program do when analyzing these 4 positions which are "exactly the >same, only completely different"? > I spend some time on this "bug" a couple of months ago. The problem is that when the board position is different, the order of pieces is different and so will the order of movegeneration. Therefor different nodecounts and in worse case different scores can occur. cheers, Tony
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.