Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: GM Smirin vs 4 Comps - Match Predictions

Author: Allen Lake

Date: 10:09:48 04/17/02

Go up one level in this thread


>It is ok for smirin to use the chessbase opening book to prepare but it is also
>okay for the programmers to change the opening book.

Agreed, since there appears to have been no agreement between the parties
preventing them from changing the opening books.  However, I would personally
like to see a programming team take on a match where their opening book was
freely available to their opponent.

>If giving a GM access to the opening book of a program is fair then it is also
>fair to give the humans who play against smirin all the information about the
>opening preperations of smirin before he play against them.

Don't want to belabor the obvious, but human-human and human-computer matches
are subject to completely different rules.  Is it fair that a substantial
portion of brain and nervous system activity in a human is devoted to regulating
and maintaining the life of that human -- breathing, heartbeat, movement, etc --
so that not all of the potential brain power could be focused on calculating
variations?  Wouldn't it be more fair to the GM if we forced the chess engine to
spend an equivalent percentage of its thinking time to maintaining the health of
the local computer system -- scanning the system for viruses, perhaps?

Seriously, though, because of the nature of the two opponents, only a portion of
the accepted OTB rules of the game are applicable (after all, was it really
possible for Shredder or Hiarcs to commit a touch-move error?), so your
comparison above was, in my view, a "straw man".


>I think that chess is a game that nobody should know the opening preperation of
>the opponent.

Once again, presumed in human-human matches, but negotiable in human-computer
matches.

>You can also ask the following question:
>If humans need unfair means like getting the opening book and the opening
>preperation of the opponents(something that they never get against humans),
>does not this say enough about how far GM's are from actually equalling the
>ability of computers?

To borrow a phrase, "unfair" lies in the eyes of the beholder.  I think a case
could be made that opening books are an "unfair" advantage for computer programs
against humans.  Why?  Because they allow the program to minimize its usage of
calculation time (a limited resource) in the early stages of the game, when the
size of the tree of potential moves is rapidly expanding.  The opening book
prunes the search tree at precisely the time when it is most advantageous to the
computer to reduce the size of the search tree.  The time saved at this point
can then be used with greater effect later in the game, when the search tree is
much narrower (leading to greater search depth).  Combine this with endgame
tablebases, which provide perfect lines of play with small numbers of pieces,
and you may find that the first 25 percent(or more) of the game and the last 10
percent (or more) of the game have been reduced to database lookups of
precalculated data rather than calculation of new data, precisely at the points
where it is most advantageous for the program to have them.  Is that really
"playing" chess?

For the sake of argument, I propose the following match conditions:  the
computer program of your choice on the hardware platform of your choice, but
without opening books or endgame tablebases.  The opponent:  any GM of your
choice, playing blindfolded.

Who do _you_ place your money on?  And why?








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.