"A Bayesian Dominoes Player"

Supervised by Professor PD Green

Abstract

Since the beginning of their widespread use, computers have been used toemulate human opponents in classic card and board games as well as computer games.

This project tackles the problem of creating a computer player for the game of 5's and 3's dominoes. In this game, players need to make estimates as to the location of certain dominoes in order to play at an advanced level. The aim of this project is to create an expert 5's and 3's dominoes player that deals with this problem of uncertainty and estimation.

Bayesian Networks are an ideal tool for use in this problem, and this report will explain the decision, design and implementation of using Bayesian Networks to create an expert computer dominoes player.