The Game of 31 itself is a game for two players, and is played with 24 cards (1-6 four of each). At each turn a player plays a card which is added to the total, the first player to go over 31 loses. The reinforcement learner is supposed to decide which card to play based on information which is given to it. For more information about our project you can view our final presentation in powerpoint (no graphics 55K or with graphics 2.4 megs)
I found some errors on the page and fixed them right now. I finally fixed up the program, the human player works fine and you can watch a game between ais. This is probably the final modification of it, although there are some graphical errors with the watch function, but I don't really care enough to fix it.Later on August 20, 2000
Well I wasn't planning to change it, but when I noticed that it crashed when I tried to run it I decided to do something. So I recompiled it and it works fine. I also cleaned up two buttons which weren't being used in the program and packaged it into a nice self extracting zip.August 20, 2000
Posted the Game of 31 and project related stuff, everything was completed earlier and hasn't been changed in ages. I still plan to clean up everything and post it nicely but that will come MUCH later.
The Game of 31 was designed in Delphi and is posted with source code here (220K). Funny that the presentation is bigger than the program ...
It's finally fully functional, still has errors with graphics and I'm sure you could mess it up if you really tried. But I've done a reasonable amount of testing and it shouldn't crash or complain unless you give it reason to.