Author: Robert Hyatt
Date: 15:09:19 12/09/01
Go up one level in this thread
On December 09, 2001 at 16:31:41, Oren Avraham wrote: >Hello everyone. > >I need some info about history heuristics, the concept etc.. >moreover, have anyone heard on a chess program implemented with PROLOG ? > >10x in advance... In reverse order. Prolog would be a neat language to use, but it is _way_ too slow. The idea of the history heuristic is like a generalized killer move heuristic. The idea is to set up an array of 4096 counters and initialize them to zero. When a move causes a fail high, or becomes a PV move by getting backed up to the previous ply, you increment the counter element for the move causing this. I use the combined from/to square values, which is a number 0 <= N <= 4095. When you order moves, move those moves that have a history count up in the list. You discover if a move has ever been "good" by taking the from/to squares, and using that as an index into the history table. If the value is non-zero, the move is a good one. The bigger the history value, the more times it was a good move, and the more likely it is to be good now.
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.