[Computer-go] How well does pure monte carlo play?

[Matthew Woodcraft]

Eric Baum wrote:
> If you don't do rave, or use patterns to predict likely moves, or any of
> that, but just do pure monte carlo, parallelized on a cluster,how well would
> that play Go? Anyone know?

I think it will be very weak.

At 1e6 playouts per move, pure light playouts (with the 'gobble' eye
rule but nothing more) in a tree with UCT but no RAVE is extremely weak
(well under 20k), and is gaining little with each doubling of effort.

I suppose a cluster might manage 1e8 playouts per move or so. I doubt it
would reach 10k, and I wouldn't be surprised at 20k.

-M-