Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Proof for the CSS Claim that WM-Test Elo Numbers meant Strength!

Author: Bryan Hofmann

Date: 08:13:14 06/15/04

Go up one level in this thread


On June 15, 2004 at 11:01:50, Rolf Tueschen wrote:

>On June 15, 2004 at 10:45:28, Bryan Hofmann wrote:
>
>>On June 15, 2004 at 08:43:58, Steve Glanzfeld wrote:
>>
>>>On June 15, 2004 at 08:01:03, Bryan Hofmann wrote:
>>>
>>>>I give you a prime example as to why this just does not work. A while back I ran
>>>>Tao 5.6 through several EPD tests using different seetings to see which one got
>>>>the best results. I then ran a few other engines through the same set of tests
>>>>with there default settings;
>>>>
>>>>
>>>>
>>>>                ECM-GCP IQ2 BK Aemis Aemis2 MATS Totals
>>>>Tao 5.6           147   175 17  66    46     13   464
>>>>Aristartch 4.41   116   163 16  63    51     11   420
>>>>
>>>>Looking at the above results you would say the Tao is stronger then Aristarch
>>>>and this is not the case as Tao would only win 35% of the games against
>>>>Aristartch.
>>>
>>>I guess these tests didn't cover all elements of chess strength. Maybe Tao is
>>>the faster tactician, but Aristartch is better in the endgame or positionally?
>>>Nevertheless, when the strength difference between two programs is BIG (bigger
>>>than in you example probably), then the stronger program will also be tactially
>>>faster in most of the cases. When the two programs compared have a similar
>>>strength, then it is much more likely that the faster tactician is not the one
>>>which is stronger in general (as long as the other program meets a certain speed
>>>requirement which is necessary to avoid blunders).
>>>
>>>Steve
>>
>>You can belive what ever you would like, I just showed one example I have
>>several others. The point is a EPD test will only show how an engine will make a
>>move in that position only. The only EPD test that would show the strength of a
>>engine would be one that contains 2^160+ positions.
>
>
>You are a sadist! That would afford another two hours of testing!
>
>Rolf

What can I say...... I spent hundreds of hours messing with EPD tests and play
engine vs engine matches... I just want to shre the pain....


>
>:)



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.