Author: Robert Hyatt
Date: 18:21:31 08/13/99
Go up one level in this thread
On August 13, 1999 at 18:06:40, Scott Gasch wrote: >Another question: my chess program is starting to play relatively well... >recently I was playing it against another program on xboard and my program ended >up in a king & rook vs king endgame. Well, needless to say this should have >been an easy win. However, my program never seemed to see deep enough (it was >searching to 12+ ply) to make mate. Instead it just liked to keep checking the >other guy's king. :( > >Since then I looked at the evaluation routines for some open source programs >hoping to find some little heuristic to help the pieces move more aggressively >and towards a mate. I did not see anything. > >I am thinking about coding a large term of the endgame eval routine based on the >king's mobility in the hope that this will encourage the program to "trap" the >other guy's king and reduce the size of his sandbox until it's checkmate. Has >anyone done something like this? How are other people developing the killer >instinct in won endgames? > >Scott Just create a piece/square table for kings where the losing king wants to be in the center and _not_ on the edge, and when on an edge, it wants to avoid the corners. Then your program will push him to the edge where it will be able to see the mate...
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.