[computer-go] UCT finds the right answer, but...

[Christoph Birk]

On Tue, 7 Nov 2006, Peter Drake wrote:
> Consider this position:
>
> ...wBw...
> .w.wBwwww
> ...wBBBBB
> wwwwB....
> BBBBBBBBB
> wwwwBwwww
> w..wBw...
> w..wBw.w.
> w.wwBw...
>
> Orego (now using UCT) quickly finds the correct answer, but the estimates of 
> the probability of winning are strange. Here's a graph:
>
>
> The probability of winning by starting at b2 is greater than the probability 
> starting elsewhere, but shouldn't it approach 1.0, since b2 is a winning 
> move? Do others get this same behavior? Does anyone have an explanation?

b2 only wins if B plays a proper follow up move. Ie. after B(b2), W(b3)
Black has to play c2 otherwise White lives at c2.

> For what it's worth, here are the probabilities and <black wins> / <total 
> runs> through each move:
> ...

Black can win by playing any (including H1) if White fails to protect
at c2 and allows to Black to play there in the next move.

Christoph