Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Thanks SSDF :-))

Author: Thorsten Czub

Date: 11:42:12 05/24/98

Go up one level in this thread


On May 24, 1998 at 12:43:22, Mark Young wrote:

>All of the latest programs that I know of played other programs on the
>slower hardware.

This does not make the statistic BETTER, even weaker.
You cannot excuse ONE mistake by explaining that:
We have always done it this mistaken way !

> Fritz 5 is not the only program that has been tested in
>this way. There is nothing wrong under ratings theory in testing this
>way.

Of course there is !
The rating of Nimzo3.5 p90 that was generated against p90 enemies
and the rating of it against p200 Mhz machines is IMO not the same, but
is added into ONE rating.
It is not the same because the playing strength of a chess program does
not develop linear IMO.



> Since the programs running on slower hardware should have a lower
>rating. Making a win less important. And draws and losses costing Fritz
>5 more rating points.

But a slower program will not often win against a faster platform
running machine.

It will always win more than lose.
And not because the software is better, but because the hardware is
better.

>Is this the best way of testing chess program? Maybe not,

Aha !

> but since the
>other programs also have been tested in this way.

And this was also always wrong.
Hiarcs 486-33 vs. Genius 486-50/66 was not better.
And I have also complained against this thing.
Years have gone, and they are not one ELO point more intelligent wiser
in sweden.



>I don't see what your
>problem is. Other then you hate fritz 5.

I have no problem with HATING fritz.
I just want to complain about these silly way of testing-methods.

I do not complain because fritz is top. I have always complained. Long
before fritz was top. And all they do is: continue their measuring of
hardware.
They don't need to measure hardware. The price of the hardware would
show in a much better way about the elo-differences than their
time-wasting games would do.

As I said, I have a 386sx-16 Mhz somewhere. Why not put Fritz on this
and do some 300 games vs. Diep on Pentium2 / 400 Mhz.
Why not ?
If they have ALWAYS done this...





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.