Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How do I gain depth in my book?

Author: Stuart Cracraft

Date: 13:53:21 12/15/05

Go up one level in this thread


On December 15, 2005 at 16:38:33, Billy Fuller wrote:

>Hello all ! , Ive been trying to solve a hidden secert, and hope someone will
>shead some light on this before I lose my mind! :)) Lets say I take 150,000
>really good games and import them into a empty .ctg book! Ok so far, then I use
>prority analysis for the selected lines I want this book to play. Now I let it
>play ,say for 200 games, go throuh and edit lines of losses ect.  but I find
>that its only in book up to move 7. Humm the games I have imported are all no
>fewer than 35 moves in lenght! Ok, so that tells me that particular game went
>into a variation of that line my book does'nt have. Now how can I add variations
>to those lines ( without entering them one by one ) seeing how this will take
>forever? Thaxs much!

Read more? :-)

What I like to do is take a huge collection of PGN plain-text
master games in a file and then have my program parse through
ALL of them. After each move in each game, it compares the hash
key with those in an on-disk database of hash keys (scores too, etc.)

If it's not there, it adds it.

A short-search is done (a few ply with quiescence) and that score is
stored.

Once that whole process is completed, you have a very useful book
and maintenance is next necessary. Just adjust the scores through
practical play by your program (wins and losses and draw) to go
towards the positions it wins in (bump up those scores artifically)
and away from positions it loses in (drop those down.)

Back the values up the tree (i.e. minimax the on-disk book) - that
will take a while.

If there's anything you don't want the program to play, just seriously
negate its score and perhaps put a do-no-change flag in there so it's
a permanent entry.

Book maintenance and Evaluation function creation are two chores I try
to avoid spending a lifetime on. I prefer twiddlings earch and learning.
Most of my personal efforts have tried to minimize the former and maximize
the latter.

One man's opinion.

Stuart



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.