Author: Robert Hyatt
Date: 17:15:27 08/02/99
Go up one level in this thread
On August 02, 1999 at 11:38:14, Mark Higgins wrote: >On August 02, 1999 at 09:06:40, Robert Hyatt wrote: > >>On August 02, 1999 at 04:34:03, Scott Gasch wrote: >> >>>Sorry for flooding the board with my questions... :) >>> >>>I have another one. I am running into slow middle game performance because of >>>high numbers of captures and deep qeval extensions. What are some of the ways >>>people are using to limit the number of captures considered by qeval? Is there >>>an easy way to throw out captures that don't make any sense in order to speed up >>>the program? >>> >>>Thanks, >>>Scott >> >> >>I use a static exchange evaluator to purge captures (in the q-search only) >>that seem to lose material (ie QxN, PxQ). You can also do alpha-based >>pruning... if the current material score is way below alpha, and you are >>sure your positional score can't pull the score up to alpha, and a capture >>can't do so either, then the capture can be ignored. IE material_score == >>-500 (rook down), alpha=100. You know your largest positional score is +300, >>and you are considering a move like NxP. material+300+100 is < alpha. No >>sense in searching that NxP. > >It might be NxP Checkmate !!! > >Ciao > >Mark If you want to do checks in the q-search, you simply don't do this "culling" if the move is a check (did this in Cray Blitz). IE if it is a check, always include it. Unless it can't be mate because some previous move in the q-search was a plain capture and the opponent could stand pat after it...
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.