[computer-go] scalability study

[Łukasz Lew]

On 1/29/07, Don Dailey <drd at mit.edu> wrote:
> I don't understand what you are saying here.
>
> Here is what I THINK you are saying ...
>
> simple MC with all as first beats "standard" UCT at
> 19x19 go.
>
> Is that what you mean?

Yes, That is what I meant.
I.e. Noise is so high on 19x19, that You need much more playouts
to have the same confidence about move.

I.e. just look at the numbers - how many playouts each move received
(ni UCT algorithm).

Best Regards
Łukasz

>
> My experience with simple MC is that it does beat
> UCT at really fast time controls in 9x9 and I believe,
> although I haven't tested it, that simple MC
> (with all-as-first)  will be superior at 19x19 until
> you reach some fairly high number of play-outs.
>
> In the 24 hour tournament I didn't use my MC program
> for a couple of reasons,  one of them was my concern
> about the memory requirements and the other was that
> it wasn't clear to me that it would play stronger.
> I felt that simple MC might be in it's sweet spot
> even though it's not scalable beyond a certain point.
>
> However, I have no hard evidence - it was just a
> "hunch" and I did no studies as it was a last minute
> decision to enter the tournament.
>
> - Don
>
>
>
>
>
>
> On Sun, 2007-01-28 at 23:49 +0100, Łukasz Lew wrote:
> > Try simple MC with all as first :)
> > I guess it beat any UCT totally.
> > ( one playout here will be as 200-300 in UCT)
>
>