Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty NewBook vs Crafty CurrentBook (incl.pgn)

Author: Peter Berger

Date: 09:59:08 09/14/01

Go up one level in this thread


On September 14, 2001 at 11:24:58, Slater Wold wrote:

>On September 14, 2001 at 11:12:47, pete wrote:
>
>>On September 14, 2001 at 10:56:56, Slater Wold wrote:
>>
>>>Playing a program against itself is pretty much pointless.
>>
>>I disagree - he wants to test the quality of his opening book here, not the one
>>of the engine : it's a nice and simple idea to play automatches to test the
>>quality of two different opening books for the same program as long as you don't
>>start tuning the books against each other.
>>
>>Everything else is left equal.
>>
>>pete
>
>"quality of an opening book" is relative.  I know openings that Crafty has a 90%
>winning winning record with, as opposed to Deep Fritz's 40% winning record.
>(With the same exact opening.)

Maybe - but this is not about the same exact opening but about the same exact
opening _book_.

I assumed one thing here I didn't post : book learning has to be disabled (as I
do that most of the time anyway).

In this case it's a rather perfect test of the general quality of the opening
book IMHO.

And I don't believe that if you play a few 100 game matches Fritz against itself
with two different books and Crafty against itself with that same books you'll
find one opening where Crafty scores 90% and Fritz only 40% assuming you have at
least 20 games played with that opening.

I'd like to suggest another test ( same conditions) ; some of the commercial
Chessbase engines with a Kure book ( like the one of Nimzo 8 or Deep Fritz )
against themselves with a default book ( like the one of Fritz 5.32,Tiger14 or
Shredder 5.32) . I am convinced all 100 game matches will be won by the side
that has the Kure book. So what would this test prove ? The quality of the Kure
book is higher q.e.d.

Btw I also don't believe that some of the top engines prefer the Queen's gambit
and some prefer the Ruy or sth. similar - there might be some extremely tiny
effect but in general the better quality of the book is much more important than
the opening chosen itself ( again: in _this_ kind of match !).

>
>He is basing the result of the games, to the quality of the opening books.

That sounds very reasonable to me - if everything else is equal and the number
of games is big enough.

Regards,
pete



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.