Author: Dieter Buerssner
Date: 10:40:42 11/04/03
Go up one level in this thread
On November 04, 2003 at 02:49:22, Daniel Shawul wrote: > In my chess program i keep track of two killer moves and a history table. > I am a bit confused when to update them. > It seems to be better when I update them only with beta fail moves. > Is that the right way? I know most chess programs do it each time >a best move is found(like they do their hashtables). Perhaps, you misinterpreted something here. I would not believe, that most chess programs update the killer move, when a best move is found. At least not in general, when the best move is a fail low move (and this will happen very often). Many chess programs won't store a move in hashtables, when the best move failed low (UPPER_BOUND), either. So, actually, it is really like updating hashtables with moves (not scores). Perhaps, I misunderstood your message, because your last was right. Whether to update the killer move in an "EXACT" (score inside search window) situation might be debatable, and I guess for this, many chess engines really differ. Whether allowing killer moves to be captures or not might be worth trying. In my experience, there is no big difference. But it might depend very much on the engine. Handling killer moves and history moves (about) the same, seems correct. Also note, that typically you will have rather rarely EXACT score, so updating only in fail high case, or also in exact case seems from start not to be a major difference. Ignoring fail low cases for history/killer (not necessarily hash) seems almost a must. Regards, Dieter
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.