Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Looking for a half-crazy but stable, free UCI engine

Author: Tord Romstad

Date: 06:25:20 01/13/06

Go up one level in this thread


On January 13, 2006 at 08:44:51, Pro Deo wrote:

>Hi !
>A question-please:
>What about the strength of this version of your engine ?

As an author, I don't like giving very precise replies to questions
like this concerning my own program.  I have released a lot of
versions of my own engine over the last couple of years, and I
have seen many releases of engines by other authors.  More
often than not, the authors (including myself) are too optimistic.
The results of the released versions are usually slightly inferior
to the results in private, pre-release tests.

I think the reason is that we all have our favorite conditions for
testing.  I, for instance, always play at a time control with an
increment, I always play with pondering disabled, and I always
disable opening books and learning.  Even if no conscious tuning
is done, it is hard to avoid that the engine gets optimised for
the particular type of match settings you prefer.  My time
managment, for instance, probably works better in time
controls with an increment than in old-fashioned
x moves / y minutes time controls.

When a new engine version is released, most testers will of
course test it in conditions which are slightly different from
the author's favorite test conditions.  This means that the
results will often be a bit worse than the author expected.

Therefore, even though I have run hundreds of test games
with versions very similar to Glaurung 1.0, I prefer not to
post the results or my rating estimate.  I leave the task of
estimating the strength to independent testers.

The best thing you can do at the moment is probably to
have a look at the CEGT rating list:

http://www.husvankempen.de/nunn/cegtrating4040best.html

You will find Glaurung 051213, a one month old private
version, on this list.  I hope that Glaurung 1.0 is a little
bit stronger, but the difference cannot possibly be very
big.

Patrick Buchmann has played some blitz games with
Glaurung 1.0 beta 1, which is very similar to Glaurung 1.0:

http://perso.wanadoo.fr/lefouduroi/tournois/tournois.htm

The estimated rating in Patrick's test is on the level of
Spike 1.0, but it's only 84 games, and I think Glaurung was
rather lucky.

Tord




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.