Author: Bruce Moreland
Date: 11:12:17 08/25/99
Go up one level in this thread
On August 25, 1999 at 10:13:17, Jon Dart wrote: >I don't remember all the details because I tracked down and fixed >this problem in my program over a year ago. At the time I did a >lot of tracing to find exactly where the issue was. Crafty currently >does not adjust non-exact scores: maybe Bob can elaborate. Even if you want to adjust the exact scores I think it might make sense to convert the bounds to more conservative bounds. By this I mean that if you have >= mate in 3, turn it into something outrageous like >= mate in 1000. What do you guys gain by adjusting the scores anyway? If alpha and beta are somewhere near zero (and +2000 centipawns qualifies as near zero in this case), you'll cut off on all of these scores, and if they are set such that mate in 3 causes a cutoff and mate in 5 doesn't, doesn't this imply that the score at the root is mate in something? Why try to be efficient during the very last few moves of the game, it the price is bugs? bruce
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.