Re: [computer-go] Computer Go Server

[Ben Shoemaker <planetb@xxxxxxxxxxxxxx>]

--- Gunnar Farnebäck <gunnar@xxxxxxxxxxxxxxxxx> wrote:

> Ben wrote:
> > A random player should maintain a stable ranking indefinitely. It
> > could be used as the bottom end of the ranking scale.
> 
> It might have a stable strength but it is utterly useless as a basis
> for ranks. The techniques to win against random play are very
> different from real go. For an example I refer you to
> http://www.lysator.liu.se/~gunnar/random_go_challenge.html
> 
> The top program there, Yellow, can give the random player an average
> of 24 stones. GNU Go only manages an average of 18 stones. However,
> GNU Go can give about 12 stones to Yellow. And this is on 9x9.
> 
> /Gunnar

Yes, it appears to be somewhat complex.  I suspected that random play would
not be the "worst" possible play.  However, since perfect play is not
known, it is hard to definitively say what aspects of random play are
"real" go.  I suppose what is most important is the relative ranking of any
two programs, which is pretty easy to determine with enough games between
them.  
"A" defeats "B" and "B" defeats "C" yet "C" defeats "A" relationships are
always boggling.

Ben.

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/