Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Yup more games should be played when there's a big gap in elo

Author: James T. Walker

Date: 03:46:20 02/25/06

Go up one level in this thread


On February 25, 2006 at 01:59:00, Dann Corbit wrote:

>On February 24, 2006 at 22:06:44, Graham Banks wrote:
>
>>On February 24, 2006 at 21:56:43, emerson tan wrote:
>>
>>>On February 24, 2006 at 15:07:22, James T. Walker wrote:
>>>
>>>>Obviously more games are needed! :-)
>>>
>>>
>>>More games should be play when there's a big agap in elo so that when the strong
>>>engine loses, it has many games to recover also
>>
>>
>>In a knockout tournament that I ran with 32 engines playing 30 game knockout
>>matches against each other, in ALL 31 matches, the engine that was winning after
>>20 games was the eventual winner.
>>
>>I think that the 40 game matches that SSDF play are more than sufficient,
>>especially when you consider the time taken to run the matches at 40 moves in 2
>>hours repeating.
>>
>>I mean, in PRACTICAL terms, where does one draw the line?
>>
>>Regards, Graham.
>
>I think 40 games tell us what we want to know:
>
>A >> B
>
>(where >> is 'Much greater than' rather than shift operator)

In the case of Fruit vs Fritz 5.32 there is no doubt that Fruit is much better.
My only point was if you want a good comparison of the two then more games are
needed.  In the case of the SSDF this is accomplished by playing more games but
not necessarily between these two engines.  The standing on the SSDF list
provides the "how much" one is better than the other and not just the 42 games.
I remember Fritz 5.32 as a very strong engine.  Now it can't compete with old
hardware.  Might be more interesting to see what it does on equal hardware.  A
40.5-1.5 match is not very interesting.  I also believe it might serve to
overate the stronger engine.
Jim



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.