The
program internally deals out cards that are consistent with the bidding to
the opponents. It then plays out the hand to the end to see which card play
leads to the best result (most tricks won). It repeats this process roughly
100 times, each time with different cards for the opponent. After enough
simulations, it becomes clear which card play, on average, leads to the
best result. The program does not understand well-known bridge concepts
such as finesse or squeeze; everything is done using uninformed search.