Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: New SSDF-list

Author: Albert Silver

Date: 06:08:31 02/23/98

Go up one level in this thread


On February 22, 1998 at 15:52:29, Paul Petersson wrote:

>
>      THE SSDF RATING LIST 1998-02-22   59362 games played by  179
>computers
>                                    Rating   +     -  Games   Won  Oppo
>                                    ------  ---   --- -----   ---  ----
>  1 Fritz 5.0 Pentium MMX 200 MHz     2589   54   -50   198   67%  2463
>  2 Nimzo '98 Pentium MMX 200 MHz     2534   43   -40   297   63%  2436
>  3 Hiarcs 6.0 Pentium MMX 200 MHz    2533   36   -34   414   63%  2441
>  4 Rebel 9.0 Pentium MMX 200 MHz     2528   34   -32   482   67%  2404
>  5 MChess 7.1 Pentium MMX 200 MHz    2523   30   -29   575   60%  2448
>  6 MChess 6.0 Pentium MMX 200 MHz    2519   36   -34   416   63%  2422
>  7 Rebel 8.0 Pentium MMX 200 MHz     2511   43   -41   293   66%  2398
>  8 Genius 5.0 Pentium MMX 200 MHz    2504   33   -31   496   63%  2407
>  9 Shredder 2.0 Pentium MMX 200 MHz  2500   55   -53   169   56%  2456
> 10 Rebel 9.0 Pentium 90 MHz          2455   32   -32   472   53%  2435
> 11 Hiarcs 6.0 Pentium 90 MHz         2450   23   -22   973   57%  2400
> 12 Rebel 8.0 Pentium 90 MHz          2449   21   -20  1173   59%  2386
> 13 Hiarcs 5.0 Pentium 90 MHz         2436   41   -39   323   63%  2342
> 14 Genius 5.0 Pentium 90 MHz         2429   22   -22   981   54%  2403
> 15 MChess Pro 6.0 Pentium 90 MHz     2422   21   -20  1150   53%  2402
> 16 MChess Pro 5.0 Pentium 90 MHz     2407   26   -25   759   62%  2318
> 16 Genius 3.0 Pentium 90 MHz         2407   27   -26   719   62%  2320
> 18 Genius 4.0 Pentium 90 MHz         2405   25   -24   849   62%  2320
> 19 Rebel 7.0 Pentium 90 MHz          2401   26   -26   740   61%  2325
> 20 Junior 4.0 Pentium 90 MHz         2396   26   -27   687   46%  2422
> 21 Hiarcs 4.0 Pentium 90 MHz         2394   25   -25   766   55%  2360
> 22 Shredder 1.0 Pentium 90 MHz       2390   59   -58   145   53%  2370
>
> 1 Fritz 5.0 Pentium MMX 200 MHz, 2589
>Nimzo98 P200X    9.5-10.5  Rebel 9 P200X     14-6     MCP 7 P200MMX
>26-19
>Genius5 P200X      9-9     Shred 2 P200X    6.5-3.5   Rebel 8.0 P90
>28-7
>Junior 4 P90    28.5-6.5   SPARC 20 MHz    11.5-3.5
>
>
>
>We have a new leader of the rating list! Fritz 5.0
>P200 MMX has got 2589 after 198 games!! That's
>the highest rating ever and 55 points more than
>any other entrant has!
>
>Fritz 5 is about 160-170 points stronger than Frans
>Morschs earlier program versions! It's an enormous
>improvement! Congratulations to Frans Morsch and the
>ChessBase team!!
>
>It has been claimed earlier that the results for
>Fritz 3 and 4 would have been better if SSDF had had
>PCs with larger RAM. For Fritz 5 we have used at least
>64 MB RAM. The auto232-version from ChessBase doesn't
>work with less RAM... For most games the PowerBook has
>been used and for the rest Fritz 5 has played from
>the ordinary opening book. In both cases the book has
>been on the hard drive in order to enable book learning.

I find this last paragraph somewhat disturbing as it seems to me to lack
scientific consistency. Most games (most???) benefit from the PowerBook,
and others used the ordinary. How precise can that possibly be? Sounds
like a friend of mine who has the nasty habit of analyzing positions
with conclusions that go like: "e5, f6, bishop takes rook and then
this..." When I ask for precise moves he stops cold. Then there is the
question of the RAM being used. Just about everyone in this forum knows
by now that Fritz is a huge RAM consumer and benefits enormously from a
generous amount of it. Fine, but I don't recall anyone going to such
lengths to benefit other programs being tested. Did the other programs
also benefit from at least 64 MB RAM? What about the aspect of the book
learning? I assume the other programs with book learning had the same
priviledges. The way the testing is presented it sounds awfully biased
if you ask me. I am not questioning the possibility that Fritz 5
deserves the no.1 spot, but rather the procedures used to award it that
spot.

>
>The margin of error for the rating of Fritz 5 is still
>about 50 points, so large changes could occur when more
>games are played. But since the gap to the other programs
>is 55 points and more, it is most likely that Fritz 5 will
>keep it's first place for some time. The competitors have
>a slight disadvantage since they have to play manually
>against Fritz 5 as long as the auto232-version isn't generally
>available.
>
>The rating changes for the other programs are rather small,
>seven points or less. But a few points can easily change
>the order of the programs. Nimzo '98 is now the second
>strongest program, followed by Hiarcs 6 and Rebel 9.
>MCP 7.1 is now four points ahead of it's predecessor.
>
>We will continue to play with the best programs. Hopefully
>new programs will arrive during the next months.
>
>Next rating list will be made mid April.
>
>Thoralf Karlsson



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.