Author: scott farrell
Date: 00:40:50 11/28/02
Go up one level in this thread
On November 27, 2002 at 09:57:30, Bob Durrett wrote: > >>Has anybody yet explored the option of using complex evaluation subprograms >>[which do not rely primarily on the use of search algorithms] for position >>evaluation? > >Essentially, the idea being hinted at is that maybe positions could be >decomposed into sub-elements [using "position fragments" is a special case] and >each evaluated separately. Then the overall position evaluation would be >obtained by some synthesis process. There may be other ways to evaluate >positions too, other than the current practice of relying on extensive >searching. > >Bob D. This intrigues me. I have often thought that many combinations within a part of the board repeat themselves much more than the entire board does. Currently when a pawn is moved on the other side of the board, away from the main action, then you re-evalute/re-search the entire board, and generally do all the same extensions, say leading closely to mate. On the other hand a human knows that moving the pawn (not near the action) is unlikely to have any effect, and as such does look down the line at all. If you could recognise patterns, of sub positions within the board, much like the pawn hash table, you could get bigger cache hits (adimitedly of a smaller part of the board), and you could also use it for pruning. Scott
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.