Author: José Carlos
Date: 09:25:04 07/22/04
Go up one level in this thread
On July 22, 2004 at 09:57:37, martin fierz wrote: >On July 22, 2004 at 08:24:32, Anthony Cozzie wrote: > >>On July 21, 2004 at 10:20:30, Albert Silver wrote: >> >>>Hi, >>> >>>This is probably old news to many, but I ran across the pages of Michael Buro >>>(http://www.cs.ualberta.ca/~mburo/), and saw an article on ProbCut, highly >>>recommending it, and even mentioning its inclusion in a version of Crafty 18.15. >>> >>>"ProbCut works in chess on top of null-move search! Download >>>mpc_crafty_18.15.tgz to play with it. We encourage all chess programmers to >>>experiment with ProbCut!" >>> >>>One can download the article "ProbCut: An Effective Selective Extension of the >>>Alpha-Beta Algorithm" on his page of publications >>>(http://www.cs.ualberta.ca/~mburo/publications.html) as well as a follow-up >>>article "A.X. Jiang and M. Buro, First Experimental Results of ProbCut Applied >>>to Chess", Proceedings of the Advances in Computer Games Conference 10, Graz >>>2003. >>> >>>For new programmers looking for material, this is certainly one, plus it might >>>be added to the links in the Computer Chess Resource Center. >>> >>> Albert >> >>To me, ProbCut just seems wrong. How can you throw out an 8 ply search based on >>a 4 ply search and expect to get things to work, unless your margin is just >>huge? (Null move is obviously completely different here). > >i suppose this depends on the game. e.g. checkers is much more benign in terms >of evaluation - you can hardly misevaluate a position seriously if you simply >count material. if you have won a man, you win the game (there are some >exceptions of course...). => using probcut there makes a lot of sense. >for chess, i don't see why it shouldn't work at all. of course nullmove is >different, but both are methods to realize when you can stop wasting your time >on useless positions. i'd say probcut is much closer to the human way of >reasoning than nullmove. when i play a game of chess i stop searching at some >point and evaluate the position, because i think it's safe to do so. i never >think "now if the opponent could make two moves in a row....". Actually, I think we do sort of null move. In a certain complicated position, I analyze a line, see I win material, then stop analyzing and ask myself "can my opponent threaten me anything?". Maybe he can push a passed pawn. I let him play again (assuming it's my turn in the final position) and do a quick search to verify he can't hurt me. Not exactly a null move, but a similar idea. José C. >probcut will work on the majority of positions where one side has bludered >material. to get it working for the cases where one side sacced material for a >deadly attack is going to be the problem! > >cheers > martin
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.