Author: Ingo Lindam
Date: 15:30:20 07/11/03
Go up one level in this thread
>My best guess is that after the search reaches this position: > >[D]2k3b1/1pPp2q1/1P1Pp2p/4Pp1p/5P1P/7P/8/7K b - - 0 5 > >it will try a whole bunch of shuffling around moves because it thinks they are >better (more than a queen up!) than taking the draw. For instance, at ply 29 >from this second position Ruffian still thinks black is winning (-10.09). It >takes 9 ply to even get to this second position, so for Ruffian to see the draw, >we're talking about a _deep_ search. > >I think this could be solved with better evaluation (or at least better >understood, maybe with a score of around -1.00 or close to 0.00). Well, isn't atleast this above position a classical deadlock situation, where both sides are unable to make any progress? And deadlock preventing is not something completely new in programming, is it? (I remember a programming homework from my second year of computer science, preventing deadlock situations for a number of "worms" crossing a nxn chessboard from several sides to the opposite, which looks not much easier than above draw detection.) A possible solution would perhaps lead into a direction of having atleast a single winning plan in mind before claiming a winning score. Internette Gruesse, Ingo
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.