Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: DF to have unfair advantage against Kramnik?

Author: Mike S.

Date: 17:16:44 06/02/02

Go up one level in this thread


On June 02, 2002 at 19:34:28, Dante Rosati wrote:

>but an opening book is nothing else than the results of playing out the lines
>and seeing who comes out better: shouldn't a computer's algorithms be able to
>compute this firsthand? (...)

Not what mankind took hundreds of years (and decades of modern experts analysis)
to produce. You probably overestimate search depth, and calculation. Openings
theory, and practise, at high chess levels depends very much on evaluation (of
course), and the "vision" of an experienced master, reliable and deep positional
judgements and the like. That goes into practical games (also theoretically
important correspondence games), and/or openings analysis. Engines can't replace
that easily. What I mentioned, were interesting exceptions (which prove that
engines have good opening capabilities), but not more.

>(...) If a computer has access to a book and database during the game, while the human
>has to rely on what is in his memory, then its like a human/human game where one
>human has book and database access while the other doesn't: is this fair?

Yes, it is perfectly fair. Both need to have opening knowledge, and by that,
they do, both. (It has been discussed often.)

If you remove the opening book from a program, it would have to re-invent the
wheel in every game, again and again. I don't expect you would consider that
being fair?

And if you have the program *generate* it's own book, by calculating, say 3
years on the fastest multi-cpu computer - what would be the result? An *openings
database* again, like it has one now. Just from a different source. You could
crosscheck it then, against existing theory...

(It hopefully doesn't matter, if it is stored in RAM or on HD :o).

So it would end just where we are already.

It could be interesting though, as a big AI experiment, because the differences
of that book to the known theory would probably be interesting to explore. But
that is pretty much away from practical computer chess and what users want from
a program and it's components.

Also, part of that idea is reality already, whenever openings and (new) variants
are analysed with engines, checked for tactical soundness etc., but just not as
one single experiment, but on thousands of computers by thousands of chess
players. I'm sure some of the results of that constantly make their way into
opening theory...

Regards,
M.Scheidl



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.