[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: calculating best position



> Best win position is the one with the greatest probability that black
> has more than half the points at the en of the game.

A simple model would be to consider all probabilities independent.
For each point on board calculate p, which is a probability of black
getting the point (otherwise white gets it (Chinese counting)).
Variance of each point is p*(1-p) and standard deviation for whole score
(constant ignored) is sqrt(sum(p*(1-p))), with which you can divide the
estimated score before comparison.

This underestimates the variance, because for example probabilities of
points in a string of stones are in fact very dependent. But I think you
get the idea.

-- 
 Tapani Raiko, <tapani.raiko@xxxxxxxxxxxxxxxxx>, +358 50 5225 750,
 http://www.hut.fi/u/praiko (in Finnish)