[Computer-go] How many probes down the tree are necessary for a"good" bot?

Peter Drake drake at lclark.edu
Thu Nov 14 14:16:33 PST 2013


On Wed, Nov 13, 2013 at 8:12 PM, Hiroshi Yamashita <yss at bd.mbn.or.jp> wrote:

> Hi,
>
>
>  You say Aya "maybe 300 heavy playouts will get 5k".  Do you mean 300k
>> playouts?  Or perhaps mean 300 playouts examined per valid move?
>>
>
> It means 300 playouts examined per valid move.
>
>
>  300 wouldn't even examine every legal move in the early game?
>>
>
> Yes. 300 playouts on 19x19 initial board is like this,
>
>   move  num   winrate
>  1:d16   64,  0.437
>  2:q16   76,  0.447
>  3:d4    45,  0.444
>  4:q4   112,  0.508
>  5:d17    1,  1.000
>  6:r16    1,  1.000
>  7:q3     1,  1.000
>

These two statements don't seem consistent. This looks like 300 playouts
total, not per option.

In either case, my program (Orego) clearly needs MUCH heavier playouts. :-)

-- 
Peter Drake
https://sites.google.com/a/lclark.edu/drake/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20131114/39f47694/attachment.html>


More information about the Computer-go mailing list