Author: Terry Ripple
Date: 23:27:22 05/31/02
Go up one level in this thread
On May 31, 2002 at 23:00:58, K. Burcham wrote:
>
>
>
>Dan I dont feel the programs are ready for this. I wish they were. I do feel
>that the top ten programs can beat any GM including Kramnik and Kasparov in some
>positions, and in some games. If a GM can close the position without either side
>losing any pawns, then the GM can let the clock run out and collect his money.
>In these positions, I know that some GM understand what it takes to get a
>program to move a certain piece behind these walls. Some GM if they understand
>programs, can close files and draw when they want to. I also feel that no
>programmer is ready to take this on, long term, as a project to improve his
>program to beat humans. For now there is money for programmers to win against
>other programs in SSDF and computer world tournaments. These are the two big
>motivaters for selling ones own program.
--------------
Well, if you are going to feed the Grandmasters with a steady income, then make
them work for it. Have the Grandmasters play from certain Nunn positions to
avoid closing up the positon. After all, checker championships are done in
similar fashion to avoid consistant draws! This could be a very interesting test
for man vrs machine!
Regards,
Terry
---------------
>
>I know Robert said his goal is to create a version of crafty to be better
>against human GM.
>
>Example:
>1. A new version of a program is released titled "GM Slayer".
>2. Sarah gets beta, plays against top 10 commercial programs, 4000 games.
>3. "GM Slayer" places 4th in Sarah's tournament.
>4. SSDF releases new list, "GM Slayer is 4th on SSDF list.
>5. Chessbase sales of new program "GM Slayer" are below average.
>
>
>I think that most programmers want their program to be at top of SSDF and to win
>world computer chess championship, because of this, I think that programmers are
>testing against other programs, and test positions.
>
>I do not think that any programmer can afford to "hire" the GM, to improve his
>program against humans. I also do not think that any programmer is presently
>working on his program with the intentions of his program being better against
>human GM. He may want his program to play better against humans, and he may
>think it will when released, but he is not trying to prove this in testing
>during developement.
>
>It would be interesting to see this list as you describe it.
>It would be interesting to anaylize games from these matches.
>But I dont think programs are ready to "cross pollinate".
>
>The true strength of todays top programs against humans cannot be shown until
>someone takes this on a personal project, long term.
>
>Chessbase could do this, but there would be not be a return on investment.
>
>kburcham
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.