Computer Chess Club Archives


Search

Terms

Messages

Subject: It is a bit misleading to say "Fritz 5.32 leads new SSDF Rating List"

Author: Dann Corbit

Date: 18:29:25 06/02/99

Go up one level in this thread


On June 02, 1999 at 02:19:10, Brett Clark wrote:

>The SSDF Rating List - June 01, 1999
>66289 games played by 191 computers
>                                           Rating   +     -  Games   Won  Oppo
>                                           ------  ---   --- -----   ---  ----
>   1 Fritz 5.32  64MB P200 MMX               2577   29   -28   627   66%  2458
Fritz 5.32 has a rating between 2606 and 2549 with 95% confidence.  The swath of
possibilities would be even larger for 97% confidence, etc.

>   2 Chessmaster 6000  64MB P200 MMX         2576   61   -53   182   76%  2378
CM 6000 has a rating between 2637 and 2523 with 95% confidence.

>   3 Hiarcs 7.0  64MB P200 MMX               2570   27   -26   715   63%  2475
H7 has a rating between 2597 and 2544 with 95% confidence.

>   4 Fritz 5.0 PB29%  67MB P200 MMX          2565   24   -23   946   67%  2436
Fritz5 has a rating between 2589 and 2542 with 95% confidence

>   5 Nimzo 99  64MB P200 MMX                 2563   32   -31   513   63%  2471
Nimzo99 has a rating between 2595 and 2532 with 95% confidence

>   6 Junior 5.0  64MB P200 MMX               2543   30   -29   576   63%  2448
Junior 5 has a rating between 2573 and 2514 with 95% confidence

>   7 Rebel 9.0  47MB P200 MMX                2525   24   -24   883   62%  2442
Rebel 9 has a rating between 2549 and 2501 with 95% confidence.

etc. for the rest.
Notice with this group, there is no way to know for sure if any of them is
really better than the others.  The "odds are" that the top programs are
somewhat better than those below them.  But notice that within the same
confidence band, Rebel 9 might be 2549 and Fritz 5.32 could be also.

The highest mean ELO estimate goes to Fritz, but I think that CM or Hiarcs are
equally likely to be of the same strength or higher.  In any case, all of these
programs are on a par in ability and they would have to play each other
thousands of times to find out which one is really the strongest.  By which
time, of course, we would all have upgraded and nobody would care.
;-)




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.