Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is it too easy to say "programmed for humans"

Author: Kurt Utzinger

Date: 03:31:05 12/21/03

Go up one level in this thread


On December 21, 2003 at 03:54:24, Derek Paquette wrote:

>I've been reading a lot of articles lately on computer chess,  and a lot of
>posts here on this forum, and something that always jumps out at me is when
>people say that an engine is great vs humans but not so great vs computers.
>
>the Fritz team says that DF8 is geared towards fighting humnans, where as
>Shredder 7.04 is undoubtably the best vs programs but perhaps not vs humans, but
>who can prove this?  Its impossible at this point.
>
>If the author of BamBam were to say "my program is 2800 elo vs humans but only
>2400 vs programs" we as a collective community could not prove him wrong.
>
>No one has the access to a GM to continually play long games against using
>different methods.  I've also heard people say Fritz hasn't 'improved' their
>engines in awhile, just tweeking it for humans, I'm starting to see that line as
>a broken record,
>
>when Shredder 8 comes out, and (most likely) dominates the computer chess world
>(some people) will jump on the band wagen again and say "its good against
>computers, but how about humans?"
>
>
>Frankly, If these other programs (junior8, Shredder 7.04, Deep fritz 8) are
>playing all at the 2800 level in tournament time control games, and Shredder 8
>is smoking them all,  how could you not say it would not be the best vs humans?

     Difficult questions but I think you somewhat miss the point.
     The matches we have seen against GM's are [in my opinion] not
     the most important thing to make conclusions. Why not: the
     programs played on super fast hardware, they were seriously
     prepared in the opening just for the opponent and eventually
     much too less games were played. The main question for me is:
     against which programs encounter human players with 2000-2300 Elo
     the greatest difficulties in getting draws [please do not claim
     that this is not possible, I am with 1978 Elo still able to get
     a lot of draws with my boring, do-nothing-but-do-it-well-strategy
     as explained many times before] if humans do not play for a win,
     if humans try not to get in time trouble and if humans do not
     allow complications. Even in tournaments like Argentina where
     Chess Tiger 15 won the tournament and had to play vs different
     opponents, most of these opponents used their normal style and
     were then punished. The same things can be said about the
     tournament which was won by Shredder. From my own experience
     I would claim that Hiarcs, Rebel and Shredder [and even MChessPro
     on fast hardware] are the most dangerous and difficult opponents
     for humans, at least when trying to use the style described above.
     On the other hand I do not think that we can say and found a modern
     program that is programmed for humans only. And something else no-
     body has so far thought about: when humans try to use "my style" vs
     computer programs this is not at all good for their overall playing
     strength vs equal opponents in human tournaments and for this reason
     I would recommend all GM's to no longer play vs computer programs:
     either they loose their matches [what a shame] or they still get
     draws [fantastic] with a boring style people are not interested
     in and which does badly influence their skill for fights human
     vs human.
     Kurt



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.