Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: CM 9000 test so far

Author: Harald Faber

Date: 14:23:33 10/12/02

Go up one level in this thread


On October 12, 2002 at 10:44:41, Terry Ripple wrote:

>On October 12, 2002 at 10:39:38, Harald Faber wrote:
>
>>On October 12, 2002 at 10:35:28, Kurt Utzinger wrote:
>>
>>>On October 12, 2002 at 10:17:09, Harald Faber wrote:
>>>
>>>>On October 12, 2002 at 09:27:25, Kurt Utzinger wrote:
>>>>
>>>>>There are three differences:
>>>>>
>>>>>1.)
>>>>>Harald Faber is using the autoplayer and two PC's
>>>>>
>>>>>2.)
>>>>>Harald Faber is playing with ponder=on
>>>>>
>>>>>3.)
>>>>>Harald Faber used the kostick.btk (under Shredder-GUI) for CM9 and the
>>>>>commercial books for the other programs, whereas we chiefly let all programs
>>>>>play with the remis.ctg (write protected, thus no learning). We wanted to test
>>>>>playing strenght of engines only and not good/bad books.
>>>>>
>>>>>In my opinion the last point may have the greatest influence on the outcome of
>>>>>matches, and furthermore the Kostick book may be too old fashioned to keep up
>>>>>with other modern books. And eventually, there is always a statistical margin of
>>>>>error in all matches played.
>>>>>
>>>>>Kurt
>>>>
>>>>
>>>>IMO the Kostick book is even better for testing than some remis/draw opening
>>>>book. Just my 2 cents.
>>>>And you forgot to mention
>>>>
>>>>4) I am testing 40/120+60 whereas you test 40/40.
>>>>( 5) The opponents use the 5-man tbs)
>>>>
>>>>4 SIGNIFICANT differences.
>>>
>>>With respect to books you are surely wrong Harald. The program (CM_9000) using
>>>the Kostick.btk (converted Kostik.opk) has a disadvantage vs the top modern
>>>commercial books of Deep Fritz, Hiarcs8 and so on. That was the reason why we
>>>gave all program the same remis.ctg book. This book does not at all include
>>>lines that lead to drawish games, but the programs are at the same moment out of
>>>book in balanced (positional/dynamical) situations.
>>>Kurt
>>
>>May I remind you on the CLEAR loss CM 9000 plus Shredder opening book vs.
>>DeepFritz 7?
>>For me it does not look like a big dependency on the book.
>>Anyway, sure, for further testing I will modify the Kostick book, there are some
>>strange King's Indian and A11 lines contained which a program should not play.
>--------
>
>Will it help to make each program play the black and white side in each opening
>line to alleviate weak lines as both must play it?
>
>Terry


Maybe, but so far I haven't found out how to do 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.