Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Just learning capability?

Author: Mogens Larsen

Date: 12:00:13 06/13/00

Go up one level in this thread


On June 13, 2000 at 14:39:21, pete wrote:

>Let's throw another argument not yet mentioned into the ring :
>
>The idea about opening books discussed up to now seems too narrow to me .
>
>It mostly concentrates on opening books created by pure collection of human
>masters' games so the idea that this is some sort of cheating seems natural at
>first sight .
>
>Two aspects I think are not taken into account :
>
>a.) It is _not_ trivial to build a good opening book this way ; you have to
>implement learning eventually ; sometimes players play bad openings but still
>win etc. ( you won't want those in your book for long )
>
>In no way does this make up with the way strong humans build their opening
>repertoire , it can only _repeat_ what others _played_ before ; and something
>like the REBEL EOC is a non-trivial additional feature that probably took some
>time to implement too .
>
>Human players tend to know and understand far more about their pet lines than
>actually played games .

I've been trying to build a book for the program Amy for some time, which hasn't
been a picnic I can assure you. But the level difficulty in building a good book
isn't really an argument. It might be an argument for not using opening books,
since a program plays a different kind of chess. Letting it discover its own
strength and weaknesses could maybe ensure that the program plays an opening it
"understands".

>b.) How do you want to deal with the programs who have a hand-build opening book
>containing lots of original analysis done by program and program author ( or
>mostly independent master player or opening expert member of the team ? )
>
>I remember finding lots of completely original analysis in the MChess books
>build by Sandro Necchi or in the Hiarcs books ( can't remember the author at the
>moment ) ; same probably true for the specially tuned books of Alexander Kure
>or Jeroon Noomen .
>
>Is it really "fair" to decline this ? It seems to me it requires much knowledge
>to find opening lines which are on the one hand good and on the other hand suit
>your program best as possible.

I wouldn't have a problem with original analysis by the author and/or the
program itself. However, some might argue that it's impossible to control. From
a theoretical point of view it's okay IMO.

Best wishes...
Mogens



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.