Don Dailey wrote: > How do the classic programs handle these sequences of ko threats that > must be precisely calculated to extreme depths?
Usually with big difficulty and crude heuristics. GNU Go determines that it should play a ko threat if the top move turns out to be an illegal ko capture. In that case the top move is removed from consideration and other moves are reevaluated by adding the value of followup moves. After that the new top move is chosen, which may be a ko threat if it's big enough and otherwise the originally second move. The difficulty is of course to determine the followup value, which tends not to be very exact at all, and far from all threats are found in the first place. Still GNU Go can play ko decently, at least compared to its general level of play. /Gunnar _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
