Re: [computer-go] future KGS Computer Go Tournaments - two sections?

[Don Dailey <drd@xxxxxxx>]

>  Ah, so you mean computer go tournaments aren't about programs winning, but
>  authors winning?

Yes.  That's what go tournaments should be about.

On the other hand, exhibition matches and long controlled test
sequences (publicly would be prefered) can be used to indicate which
programs are best because this can be done with much more statistical
certainty.  

Tournaments are about playing the odds, not who is best, and the
competitive guys all are very well aware of this.  

When I did computer chess with Larry Kaufman, we made decsions like
what level to set the program.  The decision was not which level made
the cmoputer play stronger, it was really which level gave us the best
winning chances.  Set the level too high and the operator has to be
extremely alert or you will lose on time.  It was the highest level
the operator could comfortably tolerate without getting fatigued and
screwing things up.

A few months back David Fotland very articulately explained how
tournaments don't really determine who the best player is with much
certainty.  Another way to put this is that "the winner is slightly
more likely to be the best player."

- Don
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/