Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Shredder Breaks 2800 Barrier !!

Author: Jens Kahlenberg

Date: 09:25:08 07/09/03

Go up one level in this thread


On July 09, 2003 at 12:14:49, Uri Blass wrote:

>On July 09, 2003 at 11:58:22, Jens Kahlenberg wrote:
>
>>On July 09, 2003 at 10:17:37, Uri Blass wrote:
>>
>>>On July 09, 2003 at 07:41:56, Terry McCracken wrote:
>>>
>>>>On July 09, 2003 at 07:15:47, Peter Stayne wrote:
>>>>
>>>>>Mick, see about 10 threads below.
>>>>>
>>>>>On July 09, 2003 at 05:59:20, Mick Turner wrote:
>>>>>
>>>>>>First Chess Program to ever to do on SSDF !
>>>>>>
>>>>>>http://w1.859.telia.com/~u85924109/ssdf/
>>>>>>
>>>>>>
>>>>>>http://www.chessbase.com/newsdetail.asp?newsid=1048
>>>>
>>>>It also means there is inflation with the ELO at SSDF.
>>>
>>>No
>>>
>>>I do not see inflation because inflation means that all programs get better
>>>rating.
>>>
>>>>
>>>>Many people confuse such numbers, thinking that means these programmes are
>>>>reaching 2800 FIDE which simply isn't so.
>>>
>>>1)The pools are different pools but we have not a proof that Shredder7.04(A1200)
>>>is not at the level of 2800 FIDE.
>>>
>>
>>That's true and we have not a proof that Shredder 7.04 has even reached the
>>level of 2700 FIDE-ELO.
>>
>>>2)It is also impossible to find fair Fide rating by games against humans because
>>>humans can learn the weakness of the program by playing against it at home when
>>>they cannot do the same against human opponents.
>>>
>>
>>Hmmm ... a specialist might tune the engine on games basis of the human
>>opponent, is that fair?! (BTW: Boris Alterman did more than a 24 hours job after
>>Deep Junior lost 1st game against Garry)
>
>These games were not used to calculate fide rating for a specific combination of
>software+hardware so they are not relevant for this discussion.
>
>>
>>
>>>3)I think that the only  test to get ssdf rating for humans is to play a
>>>tournament when no player know the opponents.
>>>
>>
>>... but in your ssdf-rating-tournament the humans should be allowed to look into
>>an opening book and of course have access to TBs (if you really think that
>>getting a ssdf rating for humans is a good idea)
>
>Humans in real tournament remember theory lines so you can look at the opening
>book of computers as part of the memory of it.
>
>Tablebases are also part of the memory of chess programs.
>
>Uri

So my conclusion is:
1. It is impossible to get a feasible FIDE-ELO-rating for programs
2. It is impossible to get a feasible ssdf-rating for humans

Jens



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.