An Application of the Gibbs Sampling to the Battleship Game

Abstract

Battleship is a classic two player game where the goal is to sink the opponent’s ships. Programming a winning strategy for this game is difficult because the state space representing the possible coordinates for the opponent’s ships is huge. To solve this issue, we implemented an algorithm based on the Gibbs sampling to estimate the probability of each coordinate to contain a ship. Simulation results regarding the number of guesses to sink each ship and to complete a game are presented along with strategy insights.

Faculty Mentor: Cristina Anton

Department: Mathematics and Statistics