Author: Don Beal
Date: 13:36:29 10/05/04
Go up one level in this thread
On October 05, 2004 at 09:21:54, Steve Maughan wrote: >So instead of doing a null move search around beta, you're doing a null move >search at Beta - WACBOOST, and then adding WACBOOST so as to potentially get a >cutoff with any fail high from the modified search. I assume that the modified >search will fail high 'easier' than a normal Null move search and the assumption >is that there should be a move that has an incremental value of > WACBOOST - is >this correct? > >I haven't seen this idea before. Is it new? Has it been documented before? Do >you have any data on how it performs in tactical positions and gameplay? It's not new. Many programmers will have tried this. IIRC the first mention in print was Goetsch and Campbell in the Marsland & Sheaffer book Computers Chess and Cognition, but I'm posting this from home, so can't look it up. Your description of the way it works is right. It's a generalisaton of the null move idea. Null move pruning is a bet that the real score will be more than the score from doing nothing. Adding the margin I called WACBOOST above is a bet that the best score will be not only better, but better by the specified margin. This will still be a good bet for very small values of margin, but as the margin increases the bet will be more and more likely to fail. On the other hand, pruning will be more and more likely to occur. If one sets the margin at large values, the effect of the pruning is to prune away move sequences in which the two players battle it out for positional advantage, but it leaves unpruned major (i.e. tactical) threats and defences. So it could be used to see how good a program is at finding combinations, which I think is more or less what Stuart is trying to do.
This page took 0.01 seconds to execute
Last modified: Thu, 07 Jul 11 08:48:38 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.