Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Kramer Vs Kramer (Crude estimate of the value of an amateur book )

Author: Mogens Larsen

Date: 02:31:44 08/27/02

Go up one level in this thread


On August 26, 2002 at 17:37:56, Arturo Ochoa wrote:

>Hello:
>
>No, it is not better and it doesn't have any sense. As you declared in the long
>thread below, you believe (but it is not demonstrated) that a engine with book
>is not better than a engine without it.

The problem with matching two identical engines with and without book is the
deterministic nature of the games due to identical evaluation. Some would call
that an advantage, ie. a more truthful picture of the situation after the book.
However, that doesn't simulate actual matchplay conditions. The ponder
percentages are lower (if played with ponder) and the hashtables less likely to
be filled with "correct" information. The other problem is identical or very
similar games. Here learning won't usually help the bookless engine.

Furthermore, it doesn't follow that the book advantage is tranferable to matches
against other engines. My conjecture is that if you use the same book (eg. same
PGN collection) in matchups between different sets of identical engines, some
will display a difference and some none at all. Maybe even do worse. I also
suspect that the ELO difference will decrease with extra time on the clock.

The problem with testing against another engine is that any ELO differencies may
exaggerate or diminish the actual gain by opening books. So to get a realistic
picture a selection of opponents is necessary. Here I agree with Uri's
suggestion. Some kind of gauntlet seems better if you want to test the case no
book vs. book.

Regards,
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.