In the context of an introductory AI course, we would naturally show students that
the structure of the game tree consists of choice nodes that one maximizes alternating
with chance nodes that one averages. The basic relevant algorithm would be (depth-limited) expectimax. Given the branching factor of the problem, a depth-limited
expectimax without optimizations in modeling (e.g. fast bit representation and operations)
and search (e.g. caching), one would need to terminate at a shallow depth-limit in order
to achieve rapid play.