Author: Robert Hyatt
Date: 05:40:53 09/03/99
Go up one level in this thread
On September 01, 1999 at 21:24:30, James Robertson wrote: >On September 01, 1999 at 19:52:10, Heiko Mikala wrote: > >>On September 01, 1999 at 12:41:41, Alan Grotier wrote: >> >>> >>> >>>At what negative evaluation figure is it reasonable to assume that a chess >>>program can nolonger defend the position and will lose the game? >>> >> >>My program at the moment (for playing on FICS) is set up to resign, if for 4 >>moves in a row the score is <= -8.0 >> >>The score -8.0 is adjustable via an .ini file though. >> >>This seems to work fine. It doesn't resign too early but also doesn't frustrate >>human players by forcing them to play out a game through a boring endgame up to >>the mate. >> >>Another thing I'm using is, that the program won't resign, if it sees, that it >>will be mated soon, assuming that the human sees the mate too and that humans >>feel especially good after having mated a computer. >> >>The only problem is, if the opponent is in time trouble... you will start to >>pray, that the opponent loses on time, before your engine resigns ;-) > >That problem had me disappointed so much that I eventually made my program >resign only when >12 points down! > >James > > >> >> >>Greetings, >> >>Heiko. then make a dynamic resign threshold, like I did my draw threshold. Start it at N, if your opponent has (say) 25% as much time left as you have, subtract another M from it. If the increment is zero, subtract another M from it. And if his time is below some critical level, turn it off altogether. Crafty won't resign, for example, if you are playing 5 min chess and have less than 1 min with no increment.
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.