Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF Fritz 5.32 MMX - CT 2004 A1200 Ended: 13-31

Author: Dann Corbit

Date: 12:41:00 12/03/04

Go up one level in this thread


On December 03, 2004 at 13:31:33, Dann Corbit wrote:

>On December 03, 2004 at 10:10:08, Eduard Nemeth wrote:
>
>>Plese let play CT against Junior 8, Fritz 7 or Shredder 6 on Athlon 1200 or
>>HIGHER!! - and not more against an P2 200 MHz. No interesting for the
>>computerchess community I belive!!
>>
>>##################################
>>[Event "SSDF 120'/40+60'/20+0'0. "]
>>[Site "palp.gbg"]
>>[Date "2004.11.17"]
>>[Round "1"]
>>[White "Fritz 5.32 P200MMX"]
>>[Black "Chess Tiger 2004 A1200"]
>>[Result "0-1"]
>>[ECO "D85"]
>>##################################
>>
>> No non non no................
>
>This is a fundamental misunderstanding of mathematics.  Fritz 5.32 is one of the
>best possible opponents if we want to know how strong CT is.  The more games the
>program has played, the more clearly we understand its strength.  We know the
>exact strength of Fritz 5.32 MMX very, very well.  And to play against opponents
>of dissimilar strength is also very helpful in understanding the true ability
>(unless the difference is grotesque).
>
>Think of the opponent chess programs as measuring devices.  Some of them will be
>like a rope with knots tied in it, some like a yardstick and others like a
>micrometer.  The more games a program has played, the closer to a micrometer it
>is.  The fewer games it has played, the more like a rope with knots tied in it.
>The best programs to find the true strength are the ones with the most games.
>
>And (I believe) that random fluxuations are far more troubling in programs of
>nearly equal strength.  In such a situation, the "bad runs" of a coin toss
>experiment can come into play.  But if you play an opponent that is of higher or
>lower strength (at least 100 Elo) such an effect must necessarily be of lesser
>importance.
>
>So I think the choice of opponents was an excellent one.

If you examine this page:
http://w1.859.telia.com/~u85924109/ssdf/rlwww042.txt

And come to this entry:
                                           Rating   +     -  Games   Won  Oppo
                                           ------  ---   --- -----   ---  ----
  63 Fritz 5.32  64MB P200 MMX               2493   14   -14  2409   44%  2539

You will see that the error bar is only +/-14 Elo [And with 2 standard
deviations at that!].  In fact, that particular engine is the finest micrometer
in the entire bunch.  Ironic (truly) that you should complain of this one.  It
tells us more information about games played against it than any other engine in
the bunch.



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.