[Pachi] [Computer-go] MCTS and perfect endgame

Jean-loup Gailly jloup at gailly.net
Sun Jul 3 20:22:11 CEST 2011


> Of course I am always willing to be proven wrong.

I want to prove you right, not wrong :-) I like your idea. I am just careful
because I had so many excellent ideas that turned out to weaken pachi...

Jean-loup


2011/7/3 Leon Matoh <leon.matoh at guest.arnes.si>

> **
> On 07/03/2011 06:45 PM, Jean-loup Gailly wrote:
>
> > Suppose you have 5 valid point which lead to same result r.
> > No matter how you play your winning percentage is  (r-a)/ (b-a)
> > where a lower bound of random komi, b upper bound of random komi, r is
> fixed result
> > And (b-r)/(b-a) for opponent of course.
>
> This is true only if a <= r <= b. You select moves with higher r, which is
> also
> what the parameter val_scale does. The cutoff b is similar to the
> current parameter val_points. The question is which heuristic is better,
> and this can only be determined by experiments.
>
>  Jean-loup
>
> Those are two completly different things.
>
> First a<= r <= b is always true so there is no cutoff and almost no noise
> Second result is win/loss ratio and goes to the root. result is always in
> certain bounds.
> Third winrate is accurate description of point value.
> Fourth solution is clean, short and fast, no more than 20 lines of code
> most.
> Experiments would confirm something if you can play lost game till the end
> easier
> and won game perfectly.
>
> I am by definition a Lazy Programmer. Never write more code than needed to
> get optimum.
> By heart I am mathematician and exact proof and elegant solution are my
> satisfaction.
> And of course I am doer. It does the job.
>
> Of course I am always willing to be proven wrong. Som I can learn
> something.
> There is plenty mathematicians in go circles. Maybe somebody can help
>
>
> Leon.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rover.ms.mff.cuni.cz/pipermail/pachi/attachments/20110703/fc65f2fd/attachment.html>


More information about the Pachi mailing list