Author: Will Singleton
Date: 14:25:13 03/03/99
Go up one level in this thread
On March 03, 1999 at 14:33:04, James Robertson wrote: >I just fixed a bug in my q-search. The wierd thing is that the program seems to >play worse after the bugfix!! > >After SEE, the bug would eliminate (randomly?) certain moves from the qsearch if >there was a move that gained more by SEE. This hurt it's performance in tactical >suites like WAC where odd captures and stuff are important. However, it sped up >the search in regular chess positions, and therefore could sometimes see a >little farther ahead. > >My question is: my qsearch is just alpha-beta, with moves that lose material by >SEE being pruned without searching. Are there any refinements to this method? I >read somewhere that EXchess does something with somevariable called delta, but >that is all I know. I am hoping that the more I cut down my qsearch, the more >advantage the non-bug version will have over the bugful version. > >Thanks for any help. >James I think you're talking about the method employed by Crafty and others, which is to exclude captures that are judged to be futile, even if they are good captures. For example, if you're down a queen in a qsearch capture sequence, there's not much reason to win a pawn. Will
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.