> The idea of using f(score) instead of sign(score) is interesting. Long 
> ago, I tried tanh(K*score) on 9x9 (that was before the 2006 Olympiad, so 
> it may be worth trying again), and I found that the higher K, the stronger 
> the program. Still, I believe that other f may be worth trying.

In fact, that's because I had seen that suggestion (maybe not from you),
that I further looked into this particular form. And by particular,
I mean f(score).
My point, though, is that you would probably get weaker, unless you set
K of the order the number of playouts, in which case the effects are
cursory except in very imbalanced situations.

I think we cannot beat 0-1 (except with the problems of nakade) by such
easy means, and not at all if the number of simulations at one node is
high.

Jonas


_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to