Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The End of the Learning Debate

Author: Komputer Korner

Date: 13:01:00 03/22/98

Go up one level in this thread


What you are also forgetting is How is your big varied book made? If you
make it by dumping in PGN games and cutting it off at say 60 ply, then
you are also including a lot of bad lines. How do you weed these bad
lines out? Crafty has had this problem for the last couple of years, and
only now is finally resolving it because of it's learning feature.
Testing a book at blitz is not the same as testing at longer time
controls. So the bottom line is that making a huge book is no panacea
because your program will lose a lot of games from all the bad lines in
the book. making and adjusting a good opening book is somewhat the same
for a human. Either you do a lot of opening prep or you suffer a lot of
losses until you weed out th ebad stuff. Either way it is painful. There
is no easy answer. However my suggestion of letting programmers  submit
opening books at any time along with a sophisticated learning algorithm
along with random algorithm evaluation changes that Amir Ban has
suggested should fix the book cookers. Amir doesn't see book cooking as
any problem and I agree. It has been the programmers own fault that
their programs didn't vary enough in their evaluation of the best move.
Also Crafty has a special presearch of its opening book that lets it
decide what opening move to make which is also based on learning
results. So I guess as soon as all the programmers wise up with their
programs, you are right, the learning debate will be over soon.

On March 20, 1998 at 14:22:40, Bela Andrew Evans wrote:

>KK, I disagree with the idea that you can autoplay and handtune
>your opening book to gain a big advantage against a program with
>a sufficiently varied opening book.  For example, look at crafty's
>mega opening files, or Fritz's powerbook, etc.  How long would it
>take to hand tune opening against something which could play
>different opening lines for hundreds upon hundreds of games without
>repeating the same line.  Even with autoplay, this would take too
>long to make it practical.  On the other hand, hand tuning with
>autoplay against Rebel, which plays so many Caro Kann and Slav lines,
>could be quite effective.
>
>Having said that, I don't see any reason why the chess programmers
>shouldn't make periodic opening book updates available -- they could
>simply graft the latest games onto their already huge opening books
>and make the results available on their web sites. The buying public
>would be grateful, surely.
>
>Bela



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.