Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF, Fritz5 games

Author: Thorsten Czub

Date: 18:18:43 02/28/98

Go up one level in this thread


>Without using my feelings but mostly my games statistics,

!!!! :-))))


> I can say that
>giving similar programs different hardwares emphasizes the score
>differences.


AHA ! Thanks for your point of view !
I confirm 100 %.


>By giving prog A on machine A twice the time you give to prog A on
>machine B, it is not uncommon to get a 15-5 result. This doesn't fit
>with the ELO rules.

Exactly !!
I could not have said it any better.



>The same problem arises when you self-play very similar versions of the
>same program. A little change in the program can give a 60% score, but
>often you would not notice any difference if you tried against different
>programs.

Confirm 200 % :-))) !!

>It could be that some of the top programs are so similar that
>differences between them are overestimated if hardware is that much
>different.

Right. The incest problem. Suddenly the program that is NOT similar
makes the run...
Or the fastest one kills the similar programs with a too high
elo-advance.
THIS CANNOT HAPPEN IN HUMAN CHESS DUE TO THE FACT THAT never ever 2
chess players are that kind of similar like 2 versions of ONE program on
different hardware is !


>That's why I would prefer to see the results of Fritz5 P200MMX against
>other programs on P200MMX. With the results posted, the only thing I can
>say is: "Buy a P200MMX, not a P90".

AH ! Well said !!



>
>
>    Christophe

Thanks...



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.