Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Results of 112 engines in test suite "WM-Test" (100 pos) for download

Author: Mike S.

Date: 14:43:15 08/17/02

Go up one level in this thread


On August 17, 2002 at 14:49:55, Vincent Diepeveen wrote:

>On August 17, 2002 at 05:11:01, Uri Blass wrote:
>(...)
>>Did they check carefully that the test is correct?

>No they didn't, they are too stubborn for that. (...)
>Look these guys i couldn't find on any rating list. (...)
>They are too bad in chess to analyze themselves even! (...)

That type of comment isn't helpful. I think it's the nature of *very* difficult
test positions, that there will be some doubt then and when which continuation
really is the best, and that it's not easy to prove it by analysis, especially
if two variants seem to be (nearly) equal. I'm also not 100% convinced by some
of the positions/solutions - which may well be perfectly correct though (but
that still doesn't mean they are suitable for computer tests always), etc. - but
I'm also not convinced of my own analytical skills (you also won't find me on
any rating lists :o)). But it should be possible to avoid unfair critizism.

Don't forget, even if you don't trust some of the positions: Most probably you
will trust the majority of these 100 (!) positions. They offer *thousands* of
results for comparison, including the ply depth info, from 112 progs... You can
remove the positions you don't like, and still have an *enormous* amount of
quality testing data. All provided by one man with one computer only.

>Of course i'm not going to do it.

It would take a lot of time and endurance (I did it, but in much smaller
scales).

>Of course they are not
>going to test movei or any other 'amateur' engine.

A *lot* of amateurs have been tested. I think, even the majority of of the
engines tested are amateurs (I didn't count them).

>Because just *suppose*
>that one of the engines is very aggressive tuned and scores real high
>on their testset.

Actually some do, i.e. Gromit or Goliath are ahead of some commerical engines in
the WM-Test results (I don't know if aggressiveness is the reason).

>How's CSTII doing on this testset, speaking of an aggressive but very
>weak engine?

An excerpt from the WM-Test results (the last value are the solutions):

1	  Fritz 7d  (7,0,0,8)	eng 19.05.02	256	2.698	70
2	  Fritz 7c  (7,0,0,6)	eng 11.01.02	256	2.698	70
3	  Deep Fritz 7	        eng 04.08.02	256	2.696	70
(...)
52	  MChess Pro 8	        exe 19.09.98	60	2.630	47
53	  Chessmaster 8000 	exe 12.02.01	256	2.630	47
54    >>> Chess System Tal 2.03	exe 24.05.99	128/64	2.629	46
55	  Aristarch 4.4  (UCI)	exe 02.08.02	256	2.628	47
56	  Shredder 5.32	        dll 26.05.01	256	2.627	46
57	  Li.Goliath 2000 v3.6	exe 07.05.02	256	2.627	45
58	  WBNimzo 2000b	        exe 05.11.99	256	2.624	47
59	  Nimzo 7.32	        dll 04.08.99	256	2.624	46
(...)

http://www.computerschach.de/test/index.htm

My experience with CST in my own testsuites was, that (unlike it's gameplay) it
is a quite "solid solver", but a bit slow(er), depending on what you compare it
to of course. In the Quicktest, which is much easier than the WM-Test, CSTal
2.03 is rated very similar to Aristarch 4.0 and Genius 5, on Athlon@1.2 GHz. See
XLS file:

http://meineseite.i-one.at/PermanentBrain/quick/quicke.htm


Regards,
M.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.