Author: Russell Reagan
Date: 17:58:33 04/29/04
Go up one level in this thread
On April 29, 2004 at 18:40:26, Mike Byrne wrote: >Crafty Vs Ruffian 2.10 with random evaluation factor and piece values. >The target column represents the predicted average if the facotors selected were >purely random based on the random code. > >The average column represents the average score when Crafty won. > >Not based on very many games but interesting: > > Target Winning Average >queen 900 932 >rook 500 490 >knight 300 302 >bishop 300 300 >pawn 100 98 >pawn block scale 100 111 >king safety tropism 100 103 >king safety scale 100 102 >pawn scale 100 100 >pass pawn scale 100 95 >in check extension 52.5 51 >pawn push extension 45 47 >recapture extension 45 44 >mate extension 45 43 >king safety asymmetry -20 -14 So both Crafty and Ruffian used random values for each game? Or only Crafty? Obviously you could add that feature in Crafty (if it isn't in there already), but I had never heard of commercial programs with a feature like that (not saying they don't exist). Interesting experiment. Do you have any plans to test Crafty with those winning average evaluation values?
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.