Author: Uri Blass
Date: 15:00:04 05/15/03
Go up one level in this thread
On May 15, 2003 at 17:33:38, Ryan B. wrote: >Computer vs. Computer games prove very little to me. If you remove most of the >chess knowledge from craftys eval it will actually gain a few elo against >computers. If the chess knowledge is correct then I do not believe it. My common sense tells me that a better evaluation can lead to a better order of moves and better branching factor so it should make the program search deeper when the time control is long enough. Evaluation can reduce the speed of the program only by a constant. better order of moves helps exponentially so even if you think only from search depth point of view then it is logical to guess that removing chess knowledge is not productive. It is possible that some knowledge can help against humans and not against computers but I guess that in this case the knowledge is not correct and humans are simply not good enough to take advantage of the wrong knowledge. Crafty has antisimmetric evaluation. This knowledge may help against humans but it is not correct knowledge. I see no reason to care if you can get 70% or 80% against humans. The real competition is the competition against other computers. Uri
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.