[computer-go] Re: Amsterdam 2007 paper

[Sylvain Gelly]

Hello John,

Thank you for your interest.

> Figure 3 in your UCT paper shows the accuracy of different simulation policies.
> Could you repeat these experiments for accuracy of win/loss determination only?
Actually the labelled positions are rather end game positions, and are
labelled as 0/1 (loss/win). So we already are in the case you propose.

BTW, experiments in actual play using the "best" simulation policy
with RLGO (by "best" I mean giving the best MSE) has also been done
(given in one table, I don't remember which) and showed the relevance
of the MSE measure. (winning rate vs gnugo was around 9% against 8%
with random simulation policy).

Sylvain

> So for each test position, you determine if it's won or lost under perfect play,
> and then see how close each policy gets to either 0% or 100% wins.
> One might think that this measure of accuracy is what actually matters to UCT,
> since it is not concerned with margin of victory.
> Is the advantage of Mogo's policy still as pronounced?
>
> regards,
> -John
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>