Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Book design

Author: Sune Fischer

Date: 07:14:23 11/28/02

Go up one level in this thread


On November 28, 2002 at 09:59:00, scott farrell wrote:

>one other thing I play with time to time, is to let my program choose from the
>book moves.
>
>The idea is to pick the first few moves of the game semi random based on
>weighted stats and etc. Then afer that you pick the best few moves from the book
>based on stats, and use that as you move list from root, as if those were the
>only legal moves. I give my program a smaller allotment of time, as theorically,
>it cant make a bad choice. This way, you wont walk into any openings your
>program cant understand in the given time limits, or walk into any blunders that
>are in your book (it does happen). I found when my program "didnt understand"
>the opening, when it was out of book the score was artifically low, becuase it
>didnt understand what the the original GM was trying to do, so its initial move
>(or 2) is to deconstruct the opening. ie. move you bishop back from say g6 to
>somewhere safer, or more subtle, it finds itself in the middle of a gambit, and
>tries to save the piece instead, or does not capitalize on the temp etc etc etc
>.... not good.

I am doing this precalculated, right now in fact.
I have 2.5 GHz crunching on my old book, it's sorting good from bad lines, so I
don't have to waste valuable time during a game :)

The lines may not be bad objectively, but if the program considers them bad,
then I think they are better off not being played.

Need a good book for CCT5, hehe.

-S.

>Scott



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.