Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Are programs good enough to play at postal GM level?

Author: Uri Blass

Date: 01:12:27 06/19/03

Go up one level in this thread


On June 19, 2003 at 03:33:57, georges alain wrote:

>On June 18, 2003 at 08:34:51, Uri Blass wrote:
>
>>On June 18, 2003 at 08:24:43, georges alain wrote:
>>
>>>On June 18, 2003 at 06:19:26, James T. Walker wrote:
>>>
>>>>On June 18, 2003 at 05:51:38, georges alain wrote:
>>>>
>>>>>On June 18, 2003 at 04:26:06, Uri Blass wrote:
>>>>>
>>>>>>On June 18, 2003 at 03:47:38, georges alain wrote:
>>>>>>
>>>>>>>On June 18, 2003 at 02:06:35, Uri Blass wrote:
>>>>>>>
>>>>>>>>On June 18, 2003 at 01:46:51, Peter Hegger wrote:
>>>>>>>>
>>>>>>>>>Hello,
>>>>>>>>>On very fast hardware with todays best programs, how would those programs fare
>>>>>>>>>in a round robin correspondence tournament playing exclusively against postal
>>>>>>>>>GMs?
>>>>>>>>>Even if they couldn't yet compete at this level, how far off is the day when
>>>>>>>>>they are bona fide postal GM strength?
>>>>>>>>>Opinions?
>>>>>>>>>Regards,
>>>>>>>>>Peter
>>>>>>>>
>>>>>>>>I believe that they can compete at this level.
>>>>>>>>GM's in correspondence chess are players who played well in the past relative to
>>>>>>>>their opponents.
>>>>>>>>It tells me nothing about their level relative to computers.
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>>GM's who got their rating not in the last years may be even weaker than
>>>>>>>>computers because they did not use fast hardware to get their rating.
>>>>>>>>
>>>>>>>>Uri
>>>>>>>
>>>>>>>Not sure !
>>>>>>>Christophe Léotard, better French ELO by correspondence, pulverized on 4 parts
>>>>>>>the softwares  Hiarcs 7 and Chess Tiger 14 (+3=1-0).
>>>>>>
>>>>>>The hardware was not fast hardware and I think that at least Hiarcs chose bad
>>>>>>opening because of book.
>>>>>>
>>>>>>It is better if programs trust less the open library in that time control and
>>>>>>leave the opening book earlier.
>>>>>>
>>>>>>
>>>>>>>" The more time of reflexion is reduced, the less the man can compete.  By
>>>>>>>correspondence, it is not rare to reflect 15 days on a position, to analyze
>>>>>>>alternatives which go from the opening to the finale.  In addition, the human
>>>>>>>ones have a great advantage on the machines in the sense that their libraries of
>>>>>>>opening are largely higher, as well qualitatively as quantitatively.  It is far
>>>>>>>from being the case with the clock.  The world n°1 by correspondence, Timmerman,
>>>>>>>is classified 2734.  It is established that the best machines do not exceed 2100
>>>>>>>at rate correspondence, and I am perhaps still too generous."
>>>>>>
>>>>>>No
>>>>>>
>>>>>>Based on my experience it is not the case and I won a lot of 2500+ or 2400+
>>>>>>players based on mainly computer moves.
>>>>>>
>>>>>>Steve Ham played against computers and lost 2.5-1.5 and he also did not play
>>>>>>against the best software and the best hardware of today.
>>>>>>
>>>>>>I expect 2600 player to beat 2100 player 4-0 in most of the cases so even the
>>>>>>3.5-.5 suggests that the programs are more than 2100.
>>>>>>
>>>>>>I do not understand french so I am not going to respond to the last comments.
>>>>>>
>>>>>>Uri
>>>>>
>>>>>Hello
>>>>>
>>>>>You are right if you speak about "chess with clock" but coresspondance chess are
>>>>>very different: no time pressure, no tactics tricks, possibility to check your
>>>>>opening line in an encyclopedia. or in database.
>>>>>
>>>>>Look for exemple a the last game of the match Junior-Kasparov.
>>>>>Kasparov was in a véry good position beut hard to win so he prefer to drawn the
>>>>>game and the match. Do you really think that kasparov would have draw this game
>>>>>in correspondance chess ? certainly not.
>>>>>
>>>>>please excuse my poor english
>>>>>
>>>>>Phili
>>>>
>>>>But would this position arise in corrrespondence chess?  You talk of how
>>>>Kasparov would do better with more time but what about the computer?  It's not
>>>>going to do better also?  Another thing.  You talk about the best chess player
>>>>in the world not just an ordinary GM.  Why is it when people try to claim
>>>>computers are not GM level they always want to compare the the #1 rated player
>>>>when there are hundreds of GM players that computers beat easily.
>>>>Jim
>>>
>>>hello
>>>Computers don't do much better with more time , have you tried to replay the
>>>games of this match on your computer ? I do, with a computer wich is a least 10
>>>x less faster, moves are often the same! .
>>
>>
>>This is wrong.
>>
>>Computers do much better with more time.
>>
>>The ssdf list always have better rating with better hardware.
>>
>>I will start to believe that computers do not do much better with more time when
>>Movei200 on P1000000 is going to get smaller ssdf rating than Movei200 on
>>P500000 inspite of hardware that is 2 times faster thanks to a statistical
>>error.
>>
>>Uri
>I am very suspicious with ssdf list andd for testing program really I prefer
>match with humans. More than 2700 Elo for shredder (or others)! well I have seen
>moves made by shredder (or others) that even a GM with only 2500 Elo and two
>liters of beer in stomac never play;
>If you don't trust me please  go to this :
>http://ajec-echecs.org/articles/leotard2.php  ( in this page click on
>"Hiarcs-Léotard" )
>
>and you will see what a coresspondence player which is not world champion and
>only 2600 Elo  can do to the poor Hiarcs 7

Hiarcs7 is a weak program relative to the programs of today and it was not
optimized for long time control(maximal depth of 30 or 32 plies is a joke for
correspondence games).

I read that page in the past and it did not convince me.

I have Hiarcs7.32 and I almost never used it for my correspondence games because
it also had the problem of a branching factor of 6 or 7 at long time control and
not only the problem of not search more than 30 or 32 plies even in the longest
lines.

Uri



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.