Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strength Implications of an Opening Book

Author: Johan Hutting

Date: 08:46:07 08/02/01

Go up one level in this thread


On August 02, 2001 at 09:47:56, Adam Oellermann wrote:

>>>PS: I know that a better plan would be to build a "hand-tuned" opening book, but
>>>there are a few problems:

Make sure you can still finetune it, there will still be plenty of crappy lines
in your book.

>>>- under the terms of the bet I would have to do it *myself*
>>>- I am a rather weak player (maybe 1500)
>>>- I don't really have the time or inclination anyway
>>
>>Something which can help is if you want a small openings book use Kasparov's
>>games most of the time computers do agree with these moves.
>>But i also have a database with 35.000 games which is prety good for an
>>openingsbook
>>containes games from 2600 rated and above and analyses
>
>I have a similar database, as well as larger databases from 2450+ players etc.
>I'm trying to find out what experience other programmers have in terms of the
>strength improvement of no opening book vs. automatic-generated opening book.

The issue you should be looking at isn't the book making your engine better, it
will take alot of time to tune the book towards your engine. However, the
benefit of a book is that your engine will play more 'random' games instead of
repeating lines over and over. Even a book with just the ECO keys would be
sufficient for this.

Celes has a book made from TWIC326 to (currently) TWIC351, by using a util from
another engine author which translates the moves to lines (f.i. e2e4e7e5g1f3
yadda yadda).
The implementation I use is quite simple and eats up alot of memory while in
book, but it works for the moment (until I spend some time coding a 'hashed'
book implementation). Now it's plain string comparison, I just dump all lines
that are not equal to the currently played line, then I'll randomly play one of
the next moves.
>
>This will give me an idea as to whether it's worth the effort now, or if I
>should concentrate on other areas and come back to opening books when the engine
>is stronger.

Seeing the games my engine has played vs yours, I'd suggest working on the
engine and adding a 'small' openingbook.

I've got a number of games vs 31337/celes, in which blikskottel usually loses 3
0 games, has a negative score in 5 0 and plays about equal in 25 0... if you're
interested in them let me know, I'll mail them to you.

Regards,
Johan



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.