Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF Rating List 2003-10-17

Author: Mike S.

Date: 11:17:16 10/19/03

Go up one level in this thread


On October 18, 2003 at 18:58:26, Ed Schröder wrote:

>On October 18, 2003 at 02:54:37, Thoralf Karlsson wrote:
>
>Great job Thoralf, as usual.
>
>Just one wish, get CM9000 tested, without this top program your list is
>deflated.

And make sure to test it with reasonable custom personality settings. For
example, to have an easy compromise I'd recommend Sel.=12 and king safety
150/150. These are the most important changes (and larger hash size, i.e. 32
MB). I guess this would achieve 80%...90% of the improvement effects of various
custom settings, for King 3.23. On some rating lists, only selectivity and hash
are adjusted.

I think The King is an exception from the usual (and useful!) SSDF rule, only to
test defaults or programmer-recommended values (but J.d.Koning has written he
thinks Sel. 12 is good, anyway). It's obvious - reality in computerchess is,
that The King just isn't used with default setting values among the fans and
experts. So you would test for an audience that doesn't exist. (Average
non-expert customers may use the defaults, but neither read nor need SSDF
ratings.)

Maybe SSDF can consider to buy a few Chessmasters in a shop? :-))

The Fritz GUI can provide the Auto 232 function. The Wb2Uci adaption to use King
in Fritz is "custom", IOW not supplied by the manufacturers, but obviously most
people use it without problems. When you'll test The King in the Fritz GUI, I
think it would be sufficient to use a solid "neutral" opening book generated
from a good collection of IM/GM games, not too big move depth. In no case use a
converted OBK book, because - as tests have shown (IIRC also reported here) -
the conversion doesn't work for the latest CM books. It produces "something"
though, but which is totally broken. You've probably read about it here.

Basically it's just that a major engine of the top class is missing in the SSDF
list, that's why people are demanding to include it... I guess the ranking
itself wouldn't be too surprising. I don't doubt that King 3.23 is among the
Top-5 (with 3rd rank within reach occasionally, maybe not constantly). But who
knows.

I think there is no kind of law, that programmers or publishers have to give
permission to include a program in a test or have control the testing
conditions(especially not when the organisation is independent!).

Regards,
Mike Scheidl



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.