Author: James Long
Date: 13:43:51 11/08/98
Go up one level in this thread
On November 08, 1998 at 15:07:36, Peter Fendrich wrote: >On November 08, 1998 at 12:15:30, James Long wrote: > >> >>I recently received a game via email that Tristram played >>against Gromit. Tristram had 1 queen, 1 rook and 3 pawns >>vs a lone king, and allowed Gromit a stalemate. I've >>tried and tried to reproduce this, but can't seem to get >>Tris to replay this move. >> >>My first thoughts were that I should've implemented that >>stalemate guard in my endgame eval. I read about this >>in the "Chess 4.5" chapter of "Chess Skill in Man and >>Machine." Doing this is painfully slow, though. I ran >>some suites with it last night, and it's ugly. >> >>Seems to me the stalemate should've been caught in the >>search. >> >>Do any of you have a stalemate guard in your evaluation? >>Has anybody tried this and tossed it out? >> >> >>--- >>James > >Stalemates should definitively be discovered in the search. When there are no >moves and the King is not in check it is stalemate. It costs nearly nothing... Right, and that's pretty much the conclusion I had come to. I just can't figure out why the program gave up the stalemate, because I do look for this in the search... not easy to debug when you can't reproduce the error. >I don't think it's a good idea to detect stalemates in the evaluation function >if the program uses quiescence search, because then the >evaluation function almost always is called from such a search and the very rare >cases of stalemate isn't worth the effort to detect. After looking at what it did to my test suite results, I don't think it's a good idea either. :-) I thought I could ease the blow by placing the stalemate check in the endgame evaluator, which is only called when a few pieces are left on the board. The ugly part is, you have to generate *all* possible moves and try them until you find a legal one to see if the position is indeed a stalemate. But I think you're right... it's not worth the work. I'm gonna throw it out. --- James >The quiescence path isn't optimal anyway and it's hard to use it for more >sophisticated logic than what a capture search give you and maybe a few other >things. > >//Peter
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.