[computer-go] Computer Go tournaments - various

[Petr Baudis]

On Tue, Nov 27, 2007 at 10:16:56AM -0800, terry mcintyre wrote:
> From: Don Dailey <drdailey at cox.net>
> 
> > On CGOS you will notice that a lot of the really good programs have to
> > "throttle down."   Even though they are capable of achieving 2500+ ELO 
> > they are isolated at the top so they often choose to play on CGOS at
> > crippled levels. 
> 
> > Just look at Greenpeep0.5.1 on CGOS to see what I mean - after 142
>  games
> > losing only 3 games it apparently stopped playing.    It's difficult to
> > get an accurate ELO estimate losing 3 out of 142 games. 
> 
> If there are "a lot of really good programs", wouldn't there be a benefit to matching them against each other at full strength? 
> 
> If it were possible to offer handicap stones, the top programs could offer two or three stones ( or even more ) to the weaker programs, and the odds of winning would approximate 50:50

That might not be so simple for some of the algorithms. In my
experimental montecarlo-based program I've found that giving it handicap
stones is of little help - its winning averages for all the moves simply
rise mostly uniformly and in the end it plays even sillier moves, being
confident about the win until five moves after white turned the game
around, it finally notices that it is somehow not winning anymore.

-- 
				Petr "Pasky" Baudis
We don't know who it was that discovered water, but we're pretty sure
that it wasn't a fish.		-- Marshall McLuhan