Date: Tue, 14 Jan 1997 21:49:40 GMT Server: NCSA/1.5 Content-type: text/html Last-modified: Mon, 04 Mar 1996 17:32:13 GMT Content-length: 4232 Learning in Games: Othello, Checkers, and Hearts

Learning in Games: Othello, Checkers, and Hearts

Our interest in these games is mainly from the perspective of machine learning and AI search. That is, we try out different search strategies, time management and learning approaches to see what works best. For this purpose, Paul Utgoff describes simple protocols, implemented for all three games, that allow programs to play each other on request, directly over an internet connection. The protocols are described in a technical report. Source code implementing the protocols, as well as basic player programs, is also available for anonymous ftp.

If you are interested in machine learning for games, you may also want to check out Jay Scott's page.

Othello (or Reversi)

Reversi, better known as Othello™ (a registered trademark of Tsukuda Original, licensed by Anjar Co., © 1973, 1990 by Pressman Toy Corporation), is an ancient game where two players alternate turns flipping two-colored disks on an 8 by 8 grid, according to certain rules; the player with the most disks on the board at the end of the game is the winner.

There are a number of good sources of information on the WWW:

To implement your own player, the simplest approach is to get the source code in C for o-admin (the game administrator), o-listen (the listening program to start a player remotely), and the basic player program (that does random legal moves) and substitute your own strategies. See the technical report for further details on the protocol, the programs, and some related issues.

A special administrator for ladder tournaments is in the works, as well as a WWW interface to show games and play human players against the programs.

Checkers

A similar protocol has now also been developed for checkers, described in the same technical report.

Hearts

A third protocol has been developed for Hearts--a multiplayer card-game. Details of the protocol and the implementation are described in the same technical report.

Other Games

Some of the above games have variant rule sets (e.g., Russian Checkers) that are not currently implemented in the protocols; it should be fairly simple to modify the code to implement other rule sets, simply by changing the function that generates legal moves. If you do, please consider making the code available by sending mail to Paul Utgoff at utgoff@cs.umass.edu such that the code (and appropriate credit) can be included in the next release.

Similarly, using the internet communication and principles described in the technical report, other games can be implemented. Please let us know should you decide to do so, such that we can provide the appropriate links to the code on this page.


Last updated July 21 1995.
Paul Utgoff - utgoff@cs.umass.edu
URL: http://www.cs.umass.edu/~lrn/othello.html