[Computer-go] Criticality Revisited

[Stefan Kaitschick]

 I always thought that Rémi Coulom's criticality measure had to good 
for something.
Intuitively, it provides a remedy against mindless play, where the only 
guiding principles are the shapes and the focus of the last move.
But the question remains what to do about critical regions. Especially 
what to do about multiple critical regions.
Because the focus heuristic can already handle a single critical region 
quite well.
So criticality can only justify the expense of the added calculations if 
it can improve the situation with multiple unstable regions.
Here is my suggestion: on the border between the mc tree and the linear 
playout, tidy up the situation by trading an equal number of moves for 
black and white in each of the critical regions. This would be akin to 
quiescence search in chess, with the main difference that in mc search 
the playout moves still follow.

Stefan
_______________________________________________
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go