Opponent Modeling in Stratego

Stratego is a game of imperfect information, where observations of the opponent’s behaviour are crucial for determining the best move. This paper describes how one can model the opponent in the game of Stratego, using a Bayesian approach. By observing the moves of the opponent, a probability distribution can be derived to help determine the identity of unknown pieces of the opponent. Experiments show that there is a significant increase in the percentage of correctly guessed unknown pieces. Moreover, the average probability assigned to the real identity of an unknown piece, shows an increase as well. These results eventually translate into an improved win rate.