Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Games 8-11. Chess Tiger 12.0 - CM 7000 Total: 5.5-5.5 !!!

Author: James T. Walker

Date: 04:36:56 10/21/99

Go up one level in this thread


On October 20, 1999 at 16:34:21, blass uri wrote:

>On October 20, 1999 at 15:29:31, James T. Walker wrote:
>
>>On October 20, 1999 at 09:45:32, Didzis Cirulis wrote:
>>
>>>OK, I will replay Game 8
>>>
>>>Didzis
>>
>>Hello Didzis,
>>I think you should count the game.  I think Christophe Theron is just being
>>overly "Fair".  I believe a program should lose if it continues to make the same
>>mistakes over and over.  This is part of improving a program and making it
>>better.  I personally hate to repeat games vs a computer and I think the program
>>should be "penalized" for not learning.
>>Jim Walker
>
>There are different opinions in this subject
>My opinion is that most of the customers use programs to analyze and not to play
>so learning is not important for most of the customers
>unless the program can really learn positions and not only to avoid opening or
>to repeat winning opening.
>
>
>
>Hiarcs7.32,crafty16.x and some other program can learn.
>Junior6  have the ability to learn.
>Junior5 and Fritz5 cannot learn by this definition(I do not know if Fritz6 can
>learn by this definition).
>
>Uri

Hello Uri,
I agree with what you are saying.  I dislike "Book Learning" because it has a
couple of drawbacks in my opinion.  1. It eventually narrows down your book to
where it wants to play only a few openings.  2.  Just because a program loses a
couple of games in an opening line vs Hiarcs 7.32 does not mean it is a bad
line.  It may take the same line and win vs Fritz.  If the line is really bad it
should not be in book to begin with.  Book learning seems to be a "Lazy" way of
pruning out some bad lines but it also prunes out some lines that are not so bad
because no program can win all the time.
The "Position" learning is a different story it seems.  If you make a mistake in
a certain position it would be good to avoid that position again or to find a
better move in that position.  I don't know why some programs limit the amount
of positions learned.  Maybe it takes up too much time to search for them in
later games.
Jim Walker



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.