Author: Robert Hyatt
Date: 20:44:36 01/29/01
Go up one level in this thread
On January 29, 2001 at 16:42:17, Gregor Overney wrote: >One major advantage of using commercial programs is the ability to select a >personality, such as an opponent playing at 700 who makes a blunder here and >then. - Well, just someone a hobby chess player has a chance to win against, at >least from time to time. > >One way to make this work is to add random functions to the evaluation functions >of Crafty, and, ultimately, screwing up the whole program. > >There must be a more predictable approach for this. Reducing the thinking time >does not make any sense on today?s CPU's. Truncating the q-search seems also not >the solution since one does not get the personally aspect by just reducing the >search depth. > >Creating a state-machine that is only executed if a personality has been >selected might be possible. This state-machine could be tight into the >evalalution engine. But where to start? > >Gregor There are three easy ways to tone crafty down: 1. use the "extension" command to reduce or completely disable the various search extensions. This will make it less tactically "aware". 2. turn off null-move completely. 3. use the "evaluation" command to scale some of the 'knowledge' back. IE reduce things like pawn structure, passed pawns, king safety, and so forth. If you find something you like, setting wise, send it to me and I will add a "personality" feature to make it easy to invoke your settings...
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.