Author: Christophe Theron
Date: 20:45:40 07/02/98
Go up one level in this thread
On July 02, 1998 at 12:33:32, Mark Young wrote:
>On July 02, 1998 at 05:17:28, Shaun Graham wrote:
>
>> What's the skinny(i've been dying to use that phrase) on this new program?
>>According to the Shep's Louguet test it delivered a knockout 2610 rating, and
>>performing better on the combinational, and endgame part of the test better than
>>any other program! Also solving them faster!
>
>I would not put to much stock in the louguet test. I was the one who sent in the
>2615 Nimzo98 results (he put 2610 by mistake) with some setting changes. The
>setting are killer at solving test positions, but sucks when it comes to winning
>a game! I would want to see some games at slow time controls vs. some other
>strong programs before I would get to excited.
Mark, you are right. I could tune Chess Tiger to get a really huge score on the
LCT, but in this case my program would play a lot weaker.
Chess Tiger is not tuned for LCT, but tuned to play the best in "real games". I
use NO test suite to evaluate my program and decide if a change is good. I
simply play games.
I noticed over the months that when my program gets better, and wins more games,
the LCT score gets better too. But I know it wouldn't work the other way around.
Christophe
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.