Author: Uri Blass
Date: 08:01:36 04/09/04
Go up one level in this thread
On April 09, 2004 at 07:56:07, Brian Kostick wrote: >On April 08, 2004 at 08:56:29, Jonas Bylund wrote: > >>Is there a rule of thumb when it comes to selectivity, for example for long >>analysis (and i do mean long) would it increase the engines performance to >>increase the selectivity value? if one where to use TheKing for long analysis, >>would it do better with sel=16 than say sel=12? >> >>Regards >>Jonas > > >Ok, from the Chessmater 8000 manual (note that CM9000 has other settings and >limits): > >1: Selective Search determines how sensitive the personality is to non-useful >plies, when analyzing the possible positions in a game. > >2: It sets the number of plies (from 1 to 12) that the personality looks at >before making is next move. > >3: Unless told to do otherwise, the personality considers all possible plies >(which can be infinite in number) before making its next move. > >4: This makes for slow playing by the personality. > >5: The default is 6, which means the personality selectively disregards >unpromising lines for the first six plies of the search. No I think that it is only in the last 6 plies of the search. It is illogical to be selective only in the first plies of the search because it means that you may miss a good move only because your selective search does not consider it. I also think based on experience with chessmaster6000 that selectivity means null move pruning. Otherwise I cannot understand the decision not to be selective in pawn endgames because endgames are the first candidate for real selective search if you try to use human knowledge to prune moves. Based on my experience selectivity is simply null move pruning and I believe that the reason that chessmaster is strong is not better pruning than other programs but better extensions. > >6: Beyond the selective search, it does a brute-force search of all possible >remaining moves. > >I broke it down by sentence since that is what I had to do to try and interpret >the information. I'd suggest that 1, 5, and 6 give some useful impression. As I >understand it, if you have infinite time, then the Selective Search should be >lowered, not raised. FWIW, BK No I think that with more time bigger selectivity is better. Uri
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.