On Wed, Nov 25, 2009 at 15:49, Alain Baeckeroot <[email protected]> wrote: >> If using a more generic approach, >> the strategy can be parametrized and optimized (both offline and >> online), hopefully resulting in a better gameplay.
> I don't understand how anything could be better than the expectation, > exept if you have additional information. > For example mogo1 does not know bulky-five-dead-shape and has lots of > problem to find it is dead (before being killed). So when playing against > mogo1 you want to bias the estimator in the branches were bulky five > appears ? Or did i totally misunderstood you ? I wasn't going to go that far. The kind of detailed biasing you mention belongs to a different layer, IMHO. I was only considering the amount of risk one would like to take, so that there is a difference between a move with outcomes ((+10, 10%),(-20, 90%)) and ((+9, 20%),(-23.5, 80%)) which with a simple linear weighting both yield -17. Depending on the situation, I might even rank higher ((+3, 50%), (-42, 50%)) that gives -20 with the simple linear approach. The strategy can be dynamic, both in terms of opponent (if it's stronger, play more carefully) and as the game proceeds (my previous invasion failed, be more careful) (we go into endgame, I'm ahead, play conservatvely). best regards, Vlad _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
