[computer-go] MC and tactics?

[Darren Cook]

> That's why MC by itself is not so good.  If you have a group of stones
> subject to capture but can be defended with only 1 specific move ...

My understanding of how monte carlo programs work improved last week
when I installed gogui and crazy stone, and looked at the score
histogram that crazy stone can produce, for a middle-game position.

There were a few games of e.g. -81 and +79, almost impossible to imagine
how they could happen. But then typically they were two humps, one each
side of zero, i.e. depending on who wins the main semeai/L&D problem.
The mean score tended to be quite a reasonable estimate. It was very
interesting for me to see how this understanding can appear out of
random games.

Darren

P.S. I also though CrazyStone played an atrocious opening and early mid
game, but then got stronger as the game got more tactical and closer to
the ending. It desperately needs an opening book.