Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty

Author: Christophe Theron

Date: 22:24:59 08/22/00

Go up one level in this thread


On August 22, 2000 at 19:20:56, Robert Hyatt wrote:

>On August 22, 2000 at 17:56:23, Chris Carson wrote:
>
>>On August 22, 2000 at 17:09:24, Wayne Lowrance wrote:
>>
>>>I picked Crafty, looks like I was wrong on this. Hind sight should have told me
>>>otherwise. Dr Bob Hyatt does not tune his program against comps. His interest is
>>>in playing Gm's, so I think that is how he tunes his programs, not a good way
>>>for comps competition.
>>>
>>>Wayne
>>
>>This is not a knock on Crafty or Dr. Hyatt.  Crafty is a good program and Dr.
>>Hyatt has done a good job with it.  :)
>>
>>I wonder how Crafty could be tuned to play against GM's at 40/2 when almost all
>>the games that I know about for v17.x are from the chess servers against a
>>variety of people (or programs) at fast time controls (< 30 min/game).  If there
>>is a way to tune against GM's at 40/2 (let alone other programs), I am very
>>interested to know how to do it.
>
>
>
>Perhaps I tune vs games not on the servers?  IE I use the servers to test the
>code (you will notice that Crafty _never_ crashes or hangs in WMCCC events,
>it just plays chess) and I occasionally see things that are a "trend".  But I
>pay the most attention to long games against GMs (yes, it plays them.  You
>might see it playing a guest at times and be suspicious if you want. :) )
>
>>
>>If it is tuned to play against GM's, I would love to see some matches like the
>>one's Rebel, Junior, Fritz have done recently.
>>
>>Again, I think Crafty is a good program.  Is is possible it is tuned to play at
>>faster time controls against players on the server that have not been "noplay"?
>>I had read a lot recently that it was tuned for longer time controls, I just do
>>not understand how.
>
>
>You can tune to win fast games, and you can tune to win slow games.  It is
>_very_ hard to tune to win both, with a computer.  If you ask computer
>operators on ICC, you will find out who thinks they are better than Crafty
>at 3 0, and who thinks they are better at 30 30.  The answer will be pretty
>revealing, as most operators have figured out where they do best vs most
>all programs.
>
>For short games, you use knowledge, scaled pretty high, to offset a lack of
>depth.  At long games, you can't over-compensate like that or you get
>shredded




I tend to agree with you, but I also think it is possible to find a compromise
and to build a chess program that would do well at any time control (in relation
to other programs tuned for specific time controls).

That's what I'm trying to do with Chess Tiger, and I admit it's a very difficult
task.



    Christophe






> (hint:  at the 1997 WMCCC crafty was at an all-time high on ICC,
>killing every GM that came along.  It played horribly at the WMCCC event.
>Further study showed that at longer time controls it was dying...
>
>
>
>Blitz and standard are two completely different games.  And chess vs a
>computer vs chess vs a GM are also two completely different games.
>
>
>
>>
>>Just a question, no hidden meaning here at all.
>>
>>Best Regards,
>>Chris Carson



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.