Author: blass uri
Date: 10:40:53 06/03/00
Go up one level in this thread
On June 03, 2000 at 12:49:03, Albert Silver wrote: >On June 03, 2000 at 10:41:58, blass uri wrote: > >>My observation is that programmers did not try to teach program some rules >>about endgames(for example about the fact that KRPPP vs KRPP is almost always a >>draw when all pawns are on the same side). > >That's less than obvious you know. Perhaps this position, but you'd be surprised >at what can be won at times. Check out Korchnoi's book on endgames to see some >of the positions he managed to win. Impressive technique. I do not have korchnoi's book on endgames. <snipped> >What rules would you suggest? There are some obvious ones, but there are also >very disputable ones that have too many exceptions to be of much use, other than >to a human player. I believe that programs can know everything that humans know if people teach them. I did not say to give 0.00 evaluation to position with one pawn advantage when you are not sure that there is no exception but something closer to a draw (it is possible to divide the regular evaluation by 4 if some conditions exist). The program may be weaker in games because of speed but it will be clear that it knows some rules and knows not to evaluate draw positions as a clear advantage. It may avoid some mistakes of trading for a drawn endgame. Uei
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.