Author: Uri Blass
Date: 06:56:51 12/23/02
Go up one level in this thread
On December 23, 2002 at 08:33:09, Vladimir Medvedev wrote: >I have read "version history" files of some open source programs. >Almost all authors mention solving WAC002 as some of milestones on their way :) > >Currently my program GreKo can solve WAC002. The common-known clue is to give >some bonus for two far connected passers. Movei can solve it without bonus for passed pawns and without bonus for connected passed pawns. > >The question is, what is the compromise value for this bonus? > >When I set this value to 350, my engine finds the solution very quick (at depth >of 8-9, depending on other parameters), but this leads to some "strange" moves >in practic games. > >Now I use default value of = 275, and GreKo still can solve WAC002 - but at >depth of 12. I think that "strange sacrifices" in middlegame became much more >rare. Based on your words: If your target is to be better in test suites then use 350 If your target is to be better in games use 275 I cannot tell you which bonus to use because it is dependent in other evaluations in your program. > >What is your experience with different values of this parameter? I guess that 350 means 3.5 pawns but you cannot learn from experience of other programmers who use a different evaluation. other programmers may use different evaluation for rook,bishop,knight and different piece square table and different mobility evaluation. Uri
This page took 0.01 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.