Le lundi 11 décembre 2006 02:11, Carter Cheng a écrit : > I still have difficulties understanding certain > aspects of the game of go. One of the things I fail to > grasp is the concept of shape. It is difficult to grasp :) David Fotland explaination is very clear.
> I noticed most go > computer programs probably dont seem understand it > that well either- but it appears to be a long term > desirable feature of a position and something to aim > for. > > Are there good ways already to define something > quality of the position like this? > GNU Go has a shape pattern database, but the documentation says: "tuning the shape database is a kind of art" Korean as a nice concept "Haengma" (litteraly = move of the horse) for describing correct "dynamics of the stones in a local situation", this often include shapes, tesujis and life&death. As MC program often find good shapes, i believe shape is a kind of "average knowledge" of a local problem, like fluid dynamics laws are average knowledge of random move. So including shapes in MC should spare computational effort (and it seems it does for some bots). No simple metric seems possible, except "it works (better) for what we can get" and there we fall back to evaluation of local/global situation: ....c.. ..Oab.. .XXO... choice for O in a, b, c or d ? (all well known shapes) .XOd... Alain. _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
