[computer-go] New version of Crazy Stone

[Erik van der Werf]

First of all, congratulations to Rémi!


On 6/3/06, Darren Cook <darren at dcook.org> wrote:
> > only 3 or 4 stones away from perfect play is
> > a great stretch of the imagination ...

and a misleading statement...

>
> 3 stones to a top pro is an immense difference; it is my understanding
> that any "run of the mill" 9-dan pro if given even just a 2 stone
> handicap would have fairly much a 100% chance of winning against the
> world's best.

Right, 3 or 4 stones doesn't sound like much for an amateur, but one
has to keep in mind that it roughly corresponds to 50~60 points and as
strength increases the difference in the number points needed to win
tends to decrease drastically.

A better way to identify the strength difference is to set the
probability of winning between two players that are one grade apart to
a fixed value (e.g., 66%) and then see how many grades one needs to
get a balanced situation. A perfect player only needs to win by 0.5
point, so there may still be a huge number of grades between the top
pro and the perfect player.


On 6/3/06, Chrilly <c.donninger at wavenet.at> wrote:
> Crazy Stone seems to be also the first programm which is according to
> Peter Woitkes requirement for his next Go-programm: On an infinite fast
> computer the playing strength goes also to infinity.

Nope, my programs already met this requirement 4 years ago, and I'm
pretty sure others have done it before me. Now if only I could get my
hands on that infinitely fast computer...

Erik