Re: computer-go: Perfect play

[Mike Gherrity <gherrity@xxxxxxxx>]

> At 2:15 PM -0700 9/28/00, birk@xxxxxxxxxxxxxxxxx wrote:
> >  > This is also an issue for the game of RoShamBo (paper-rock-scissors).
> >  >	http://www.cs.ualberta.ca/~darse/rsbpc.html
> >
> >It's a little off topic, but anyway:
> >I don't see how any program can win against "Random". The only
> >way a program could get a score above "Random" is by having "bad"
> >programs in the tournament that give away games, or (as you can
> >read in the tournament description) having programs in the
> >tournament whose strategies are known so you can prepare
> >against them.

The winning players did not do this.  You can look at their code to verify
this.


> RoShamBoGod plays "random".   She knows that random play is 
> game-theoretic optimal.

I thought RoShamBoGod would be better than any mere mortal player.  According
to the tournament results, "random" finished 41st in a field of 64 players
with a rating of 1800.  The best player had a rating of 2307.  Shouldn't
RoShamBoGod have the highest rating?


> RoShamBoDevil does statistical analysis of *your* supposedly random 
> moves to find hidden correlations in your random number generator. 
> And then exploits them, of course.

The best RoShamBo players do not play randomly, so what would RoShamBoDevil
do in that case?


Off Topic?!  If you can't explain perfect play in a simple game like RoShamBo,
why are you so confident you know what you are talking about in a more
complicated game like Go?


	mike