General Game Playing
Assignment 4
General
Artificial
Intelligence

Lessons Notes Videos Readings Resources Arena

Read Chapter 8 in the Notes. Your job this week is to build two players - one that uses Monte Carlo Search and one that uses Monte Carlo Tree Search. You should take special care in your implementations, as you will likely want to use one of these players in the final competition.

Note that players are not expected to get 100 points on all games. See below for target scores associated with each game. Your players might get lower scores in some matches due to the stochastic nature of MCS and MCTS, and that's okay. Likewise, they may score significantly higher! (They might even score consistently higher, if you really nailed your implementation.)

  1. Implement a player that uses the Monte Carlo Search technique. Once your player is ready to go, click on the links below to test it out.

  2. Hunter (target: 75+)
    Kono (target: 60+)
    Connect Four (target: 100)
    Checkers on a Barrel No Kings (target: 60+)
    Nine Board Tic Tac Toe (target: 100)

  3. Implement a player that uses the Monte Carlo Tree Search technique. Once your player is ready to go, click on the links below to test it out.

  4. Hunter (target: 75+)
    Kono (target: 60+)
    Connect Four (target: 100)
    Checkers on a Barrel No Kings (target: 60+)
    Nine Board Tic Tac Toe (target: 100)

  5. Comparison. Create one process running your MCS player and a separate process running your MCTS player and run them against each other on the games listed above. To do this, go to Arena, click on Games, click on the desired game, click on Manage a Match, enter the details of your two players, click Setup, and then click Start. How do your players compare to each other in terms of results? Check the number of nodes explored by each of the two players. Is there a significant difference?

Comments and complaints to genesereth@stanford.edu.