Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To Christophe and Ed: a bit more SelSafety seems to make GT II stronger

Author: Ed Schröder

Date: 02:06:22 04/20/01

Go up one level in this thread


On April 20, 2001 at 05:03:40, Fabio Barrettone wrote:

>Well, at last I think so from my experiments with an Athlon 950 256Mb of RAM.
>I've done 400 games against Yace (at 40/10m, sorry not too much time for longer
>time controls) with default settings and the result was 59% vs 41% in favor of
>Gambit II (I don't know about Chess Tiger 14.0 yet), using ERT (24Mb of hash for
>both the engines, and a book prepared by me for Yace).
>
>I've redone the same match (overwriting learning of Yace and Gambit) with
>SelSafety=12 (too much seems to be at the contrary overproductive) and the
>result has been 82% vs 28% in favor of Gambit. Now, I don't think 400 games and
>a so large difference is a casuality.
>
>I've also done the complete Nunn test against Fritz6b (sorry, don't have Deep
>Fritz) and also in the Sveshnikov, where GT seems to have big problems against
>Fritz, with SelSafety=12 it was able to win with white and draw with black,
>where with the default settings it lost with both white and black.
>
>Well, I've not had the time to test further this theory, but maybe, Christophe,
>with your home-made benchmarks, you could tell if my theory is wrong or not.
>Just my little 2$ on a thing I've noticed (for a casuality).
>
>Now I'm actually testing with Junior and Hiarcs, next with Chessmaster (Kurt and
>default settings).
>
>In the end a question: in the Chessbase version there's a ct.ini? I hope so,
>since I have only a computer and it's a real pain to enter the moves manually
>for engine vs engine matches.
>
>Sincerely,
>Fabio.


Thanks Fabio, I am sure Christophe will have a look at your suggestion.

Ed




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.