Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: What Makes a Chess Engine Better Vs Humans?

Author: Serge Desmarais

Date: 20:32:33 09/02/98

Go up one level in this thread


On September 02, 1998 at 12:20:23, Robert Henry Durrett wrote:

>On September 02, 1998 at 03:22:31, Ed Schröder wrote:
>
><snip>
>
>>>If specific answers can be obtained, then what do these answers suggest for
>>>future design guidelines for chess engines?
>>
>>For playing against (todays!) computers: add much more tactical stuff.
>>
>>For playing against (strong) humans: different playing style.
>>
>>- Ed -
>
>(1) Ed Schröder's idea above sounds like it may be the key:  Select the best
>"anti-human" playing style.


   There is no such things as "anti-human" playing style. The greatest strenght
is that humans REALLY learn from their mistakes, while for computers, it is a
SIMULATION of the learning process : that is why I usually put learning in
brackets when applied to computers. Remember when I reported a kind of "position
learning" for Crafty to you? For the learning to apply, the VERY SAME position
must occur again! While for humans, a SIMILAR position is enough... Exactly like
when in dos you type "copRy fileX C:", the computer would just report that it is
a wrong/unknown command, while ANY human would IMMEDIATELY know that you
intended to COPY a file to the C-drive. Not so long ago, humans had success with
"anti-computer chess". E.G. In an ordinary opening you just had to insert a
a2-a3 move as White and then resume with the thematic moves for that opening and
the program, out of book, would start playing stupid moves! Of course, that
doesn't work anymore, except in very rare case. Still Kasparov tried crawling
openings against Deep Blue to get it out of book and not give it anything
concrete to calculate. That was supposed (maybe) to win VERY easily... We know
it did not turn that way. Computers are also known as being notoriously dumb and
weak in CLOSED positions (as Fritz 5 has a negative score when playing the --or
against the-- French against other programs). But now, not all programs have
this flaw. Of course, your "anti-human style" could work in one game or 2
against a GM (would it?) and then turns against the computer! For me, the best
computer and/or program would be the one that could adapt itself to its
opponent's style and the evolution of that style!


Serge Desmarais



>(2) Now, what remains to be said is what, specifically, we mean by a "playing
>style," and to identify, define, and itemize the specific elements of the best
>"anti-human playing styles."  To be complete, it also should be said:  how and
>why those specific elements contribute to the strength of the engine against
>humans.



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.