COMP343 - Tic-Tac-Toe Agent Study
SP25 Github Classroom Assignment
A Game Playing Agent Program Study via Tic-Tac-Toe
Tic-tac-toe is a simple enough game that we can implement and run agents that play perfectly via the standard minimax algorithm. Games last no more than 9 moves and the branching factor decreases as the game progresses. You don’t even need to account for symmetry. Further more, the state-space is so small that we can compute a complete table of the true minimax utility of each state and use that as the basis for a perfect tic-tac-toe agent. So, while it’s not a game that really requires modern AI techniques, it is a game where we can, using PC-level hardware, explore these techniques and benchmark them against AI with provably perfect play.
You’ve been given all the game logic for tic-tac-toe as well as code that can run and report on complete games between two agents. The module tictactoe provides the class TTTState which encapsulates the representation of a game state along with all the necessary game logic. The module tttengine provides a single class, TTTEngine which is used to play games.
The module agent provides the basis for building tic-tac-toe agents and contains two pre-implemented agents, Human for human players, and Random for a completely random agent.