[computer-go] Lazy Evaluation in Monte Carlo/Alpha Beta Search for Viking4

[Don Dailey]

On Sun, 2006-07-23 at 15:26 +0200, Rémi Coulom wrote:
> Anyway, I am certain that neither of the two approaches will produce
> a 
> super-strong player. Global tree search simply does not work. We have
> to 
> find something else.

I think it's very likely that  highly selective global tree searchers
will someday become very strong players. 

I'll cite some initial empirical evidence of this: CrazyStone

 1. CrazyStone is indeed a highly selective global tree searcher.

 2. It has almost no explicit GO knowledge.

 3. It's possible that it's the best 9x9 Go program in the world. 

Hey, this look promising!  Nahhh we need to find something else.

- Don