Author: Vincent Diepeveen
Date: 16:08:06 09/20/00
Go up one level in this thread
On September 20, 2000 at 18:11:54, Georg v. Zimmermann wrote: >On my way home today this problem came to my mind: > >even in a quite position in the middlegame a move does increase the eval by some >centipawns (if only by putting bishop on better square). From human viewpoint that's true. From the crazy computer viewpoint it's not true. Humans don't have horizon evaluative effects like computer programs have. >Now this doesn't matter if you do a full tree search, but as soon as you do >extensions (even worse with fractional ext.), will the program not try to chose >moves that extend just so much so that the final position eval()-uated is with >opp to move ? >One could maybe add a dynamic correction factor. Anyone tried that. Results ? I experimented for years with it, and i can't get it to work, also i can only conclude it plays worse if i use a bonus, even if i add all kind of conditions on the bonus and vary the bonus in all kind of forms. >Regards, >Georg v. Zimmermann
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.