Computer Chess Club Archives


Search

Terms

Messages

Subject: Stronger

Author: Harald Faber

Date: 12:50:30 05/25/02

Go up one level in this thread


On May 25, 2002 at 12:58:28, Torstein Hall wrote:

>I do not know if this counts statistically, but Fritz seems to kling to the top
>spot quite a lot. That, and of course that it has a small margin to the rest,
>makes me feel it is just a little bit stronger.
>
>Torstein

Usually I'd agree with you, this cannot be luck/statistics at all.
But: I don't see Fritz superior than other progs. I go even further and say
Fritz is in many positions blind (from the CHESS point of view) and inferior to
several other progs. And this really makes me believe that the SSDF testing
method is different than my understanding. Of course learning plays a role, but
I'd prefer a program to NOT make a mistake first because of positional
misunderstanding. And I don't understand why programs having learned from
hundred(s of) games are tested vs. programs with no learning values at all.
These are no equal conditions. And that is why I reset the learning values (book
and position) before each match. And that is certainly a main reason why I don't
"see" and superiority nor see results in my tests that could show it.

Ed Schroeder once wrote that SSDF finds out the best learner. And I absolutely
agree with this statement. Personally I am not interested in the best learner.
And looking at CHESS UNDERSTANDING Fritz is definitely not my first choice,
although Fritz 7 has made a significant step forward in chess understanding.

I would be very interested how the results look like if the learn files (also
book) were deleted before each new match.




This page took 0.02 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.