Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: You don't know that

Author: Uri Blass

Date: 11:37:57 08/15/00

Go up one level in this thread


On August 15, 2000 at 13:40:04, Severi Salminen wrote:

>Hi!
>
>>In my humble opinion, I think Chessmaster 6K or 7K is "stronger" than Crafty
>>17.13, so I'm not surprised.
>
>How on earth can you say that (well, of course I am not to say what you can say
>or can't, but... ;)? Crafty 17.13 isn't even released yet. And besides, this is
>not a matter of opinions. What do we mean when we say that a program is stronger
>than another? I think we all agree that the program which gets more points
>against another program is stronger, when number of games --> infinity, period.


I  disagree.

By this definition chessmaster is weaker than most of the programs because
chessmaster has no learning function and after enough games it may start to lose
the same game again and again.

Suppose someone does a program that learn from experience but it needs
1000000000000000 games in order to start to play better than kasparov.

By your definition this  program is the strongest program in the world but the
practical test will show you that this program lose again and again because you
will practically never get enough games.


>And if people knew more about statistics they wouldn't make so many posts like
>"Crafty won Rebel 1-0!! Something wrong with Rebel????" When two programs are
>very equal you need many (100+ or even 1000+) games to see the true relation. 1
>game means allmost nothing. One game means _a little_, but can't draw any
>general conclusions on just one game...IMHO... :)
>
>Severi


If a program wins one game against another program then it will be more logical
to guess that the winner is better than the opposite so it is logical to have
the opinion that the winner is better.

You can change your opinion when you get more information.

Uri



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.