[computer-go] Is MC-UCT really scalable against humans?

[Eric Boesch]

On Jan 22, 2008 1:43 AM, Petri Pitkanen <petri.t.pitkanen at gmail.com> wrote:
> Even top MC programs fail to see  that a group with 3 liberties with
> no eyes is dead.

A 3-liberty group with no eyes has a 100% chance to die during
playouts unless a surrounding group dies first. 100% chance to die is
as good a job of "seeing" deadness as a generic MC playout can ever
do.

MC playouts explicitly understand eyes at the "don't fill a 1-eye in
playouts" level and implicitly have a crude understanding of the 2-eye
rule. They essentially give you Benson's algorithm -- clearly an
eye-counting algorithm -- for free. Benson's-alive chains almost never
die in playouts while chains that lack eyespace and are surrounded by
Benson's-alive chains almost always do. Under-the-stones chains can
kill Benson's-alive chains during playouts,

 --------------   --------------   --------------
| . O O # . . .| | . O O # . . .| | . O O # # . .|
| O . O # # # #| | O . O # # # #| | O . O # # # #|
| O O O # # . .| | O O O # # . .| | O O O # # # .|
| # # # O # # #| | # # # . # # #| | # # # O # # #|
| . # . O O # .| | . # # . . # .| | a # # O . # .|
 --------------   --------------   --------------

but I think this scenario is so unlikely that it's not worth a second thought.