Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Eval tuning based on LCTII results

Author: Ratko V Tomic

Date: 12:35:09 10/25/99

Go up one level in this thread


> Has anyone written a kind of "here is the reason the solution is
> the best move" document about this test suite or any test suite?

There is a book "Test Your Positional Play" by Robert Bellini and Pietro
Ponzetto (1985 Macmillan Publishing Company, ISBN 0-02-028090-4) with 30
strategy test positions from real games.

The first part of the book (60 pages) describes techniques for analyzing
positions, creating and executing plans, positional evaluations, etc. The second
part of the book (about 130 pages) contains the 30 positions, with 4 pages of
systematic analysis per position. For each position the initial 2 pages propose
3 plans, each giving the best argument for the plans A, B or C. Then the 2 page
solution critiques the plans and explains why the one chosen is the best. (It
also shows the rest of the game so you can see how the plan unfolded.) Each of
the 3 plans carries the points and a chart at the end of the book gives you the
ratings equivalent of the obtained point score. Even though this book was meant
for human players, I think it would be an excellent positional tune-up tool for
the chess programs.

Although I didn't have time for a systematic test of the existent top programs,
they generally did poorly on the few random positions I tried. Perhaps there is
an EPD of this test suite on the web and someone with enough time to play with
it can obtain the positional ratings for the existent programs at few different
time levels.




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.