Let me add a little more. When calculate 1/f, it's the ratio of bias of the chosen move to the total number of playout for level 1. For other levels it's the ratio of the bias of the move chosen to the bias of the chosen move in the parent level.
A less effective search policy will spread the playout over a larger area. A more effective policy concentrates the playout around the more relavent threads. Thus, has smaller f. For a 'perfect' policy f is appraoching (but larger than) 1. DL
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
