Teaching a Neural Network to Play Cards