Author: Melvin S. Schwartz
Date: 10:40:26 06/01/99
Go up one level in this thread
On May 31, 1999 at 21:47:17, Dann Corbit wrote:
>On May 31, 1999 at 20:50:59, Melvin S. Schwartz wrote:
>[snip]
>>It has been my experience with other people having the same software that with
>>faster pc's you will get better results from a program. And that means the
>>program will play better. It is printed in the manual for Hiarcs 7 in reference
>>to improving Hiarcs strength: "Alternatively, you can purchase a faster PC!" It
>>is also stated in the manual for Fritz 5.32 that the speed of the software the
>>program is running on has a great influence on the playing strength of Fritz. It
>>also states "Obviously a 200 MHz Pentium is going to give you a much better
>>performance than one running at 133 MHZ. There are other factors as well, but as
>>you can see, the hardware a program is running on is very important. And that is
>>why I criticized SSDF for running Hiarcs 6 on a P90 against Nimzo 98 on a P200.
>>Of course they claim that they are able to compensate for the different speed
>>and come up with accurate calculations; however, I am not totally convinced of
>>that. I still feel they're dealing in a hypothehtical situation. At Shep's site
>>there are tournaments held where all programs are run against each other on the
>>same hardware. That I believe is the absolute correct way to evaluate programs.
>You are wrong. Simple as that. The strongest, fastest, most secure and
>complete way to determine the strength of an opponent is to play against an
>opponent of known strength. The exactness of the measure of the known
>opponent's strength bears a direct relationship to the quality of the estimate
>derived from the contest. If I have a program of strength 2300 for which I have
>one thousand games played, a program of strength 2500 for which I have fifty
>games played, and a 2600 program for which I have 5 games played, I will learn
>much more mathematically by playing against the 2300 ELO program simply because
>of the greater certainty of the measurements. Your criticism is not correct.
>However, you should not feel badly that your intution is wrong. The SSDF gets
>criticized for this very thing at least once a month. Mathematically speaking,
>the SSDF is very correct and rigorous in their procedures. They even present
>the mathematical certainty along with the normal estimate.
Hello Dann!
I probably should just set-up Hiarcs, Fritz, or Rebel and take away their Queen
before the first move and have some fun instead of prolonging this topic;
however, I am a person who when having strong opinions feels he must express
them even when he knows nothing about what he's talking. :-)
If the SSDF could test all the programs against each other on the identical
computer, don't you think they would consider that a more accurate way to get
ratings than what they are doing now?
Regards,
Mel
>There is another point which you could bring up that I feel is far more valid.
>People would be more interested in games against equal opponents (despite the
>fact that they may be much less interesting from a mathematical standpoint).
>However, the SSDF runs on volunteer machines. They do not necessarily have the
>option of playing program x verses program y on PIII 500MHz machines.
>
>The solution, obviously, is to run your own tournaments and record the results
>for the pleasure of all {except for those companies who are terrified that they
>might get clobbered and have forbidden publication of contest results for that
>reason}.
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.