Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Tuned Fruit 2.1 ahead of shredder at blitz ??

Author: Marc Lacrosse

Date: 16:38:32 07/09/05

Go up one level in this thread



>... Question is will it do that good at longer time control.
>anyway very happy Fruit is giving Shredder a run for the title as best
>engine software.
>
>Gerold.

I completely agree with you, Gerold.

Even though it was possible to tune Fruit so that it could be on a par with
shredder at blitz this would not in any case signify anything regarding longer
time controls.

However this could be interesting in more than one way :
1. Volker Pittlik recently published an interesting paper on his Winboard forum
according which relative strength of various programs is very constant
regardless of time control. Moreover he showed that one of Shredder stronger
points was its ability to score much better than other programs in standardised
tests at short time controls as compared with longer ones.
2. Another interesting point is the fact that this would confirm that Fabien
Letouzey released Fruit 2.1 in a hurry without almost any kind of parameters
tuning so that a lot of performance improvement is possible for us mere
non-programming mortals through simple tuning trials.

And finally as we all know, testing with less than several hundred games has
almost no signification at all from a statistical point-of-view. Being an
isolated tester, I can get relevant data at blitz in a few days on my two or
three PCs, but I could not get anything significant before months at longer time
controls.

Longer time controls testing and tuning has to be left for well-organised
associations (like the wonderful CEGT) or rich amateurs with dozens of PCs at
home ...


Marc



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.