Author: Uri Blass
Date: 04:51:05 04/10/03
Go up one level in this thread
On April 10, 2003 at 07:11:15, Pat King wrote: >On April 09, 2003 at 10:24:57, Uri Blass wrote: > >>On April 09, 2003 at 09:18:37, Pat King wrote: >> >>>On April 08, 2003 at 19:30:24, Robin Smith wrote: >>> >>>>The following position is a rather trivial (for a human) perpetual check. Why >>>>does it take some programs so long to see it? >>> >>>Most programs use some kind of "contempt" score for draws, so in a materially >>>balanced position, they will reject draw scores, so as to provide active play. >>>In a position like this, they will keep playing to win, having no means to >>>"realize" the truth. Since this logic doesn't cause losses, there is no reason >>>for programmers to spend time making programs smarter in this respect. >>> >>>>Robin >>> >>>Pat >> >>No >> >>Most program use no contempt in analyze mode > >Good point. My program doesn't yet make this distinction. > >>and the poster did not talk about >>scores of +0.15 that are normal contempt factor but about scores of winning for >>white. > >So small? I've been using much larger, >1 pawn. Perhaps I'm missing something on >how contempt is supposed to work. I use no contempt and I plan to use no contempt in the near future(not in playing mode and not in analyze mode). contempt is used only to avoid draws. If you use value of more than 1 pawn then it means that your program will prefer to be a pawn down and not to draw by repetition. I also think that the importance of contempt is almost 0 for games unless you want to do your program 100 elo weaker by using crazy values of contempt. Uri
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.