I am an avid enthusiast of the boardgame Arimaa, and am involved in a number of Arimaa-related projects.
Because Arimaa is not marketed and sold yet, I've made several handmade Arimaa sets for my own use and for holding tournaments. Most of the information is on a separate boards page
Friends convinced me to go to the Gateway 2007 gaming convention held by Strategicon, and I volunteered to hold an Arimaa tournament there. I donated one of my handmade Arimaa sets as a prize, and Omar contributed $30 for the prize as well. Eight people entered, six of whom had never heard of the game before that weekend. All in all, it was quite a success. I have a photo gallery of the event online.
Files for your downloading pleasure
In early 2006 I sponsored two competitions (with cash prizes) challenging players to beat the available menagerie of AI programs while dramatically constraining their own play options. I plan to restart this series of challenges the next time the Arimaa community is bored and without any group activities..
I sponsored two bots (gameplaying AI programs) for the 2007 Arimaa Computer Championship. My bot, Zombie, took second place in the computer championship, behind the perennial winner, Bomb, by David Fotland.
Faerie is an update with only minor modifications of bot_Fairy, written by unic (Ola Mikael Hansson). Unic apparently abandoned further development of Fairy in mid 2006, but released the source code as a gift to the Arimaa community. Several players wanted to see how Fairy would perform in the 2007 computer championship, so I adopted it as bot_Faerie and made the necessary changes to the code to allow it to qualify for the tournament. (The changes necessary were the implementation of time-control awareness and fixed-search-depth versions for the bot ladder.)
Zombie is my own first attempt to implement a bot. I have a large number of ideas for how to write a good static evaluator for an Arimaa bot, but I've been intimidated by the task of writing the search routines, since I have little experience in game programming and none in chess AI. After adopting Fairy, I decided to try using Ola Hansson's search code as a platform for beginning development on my own bot. I only had a week or so to prepare for the 2007 CC, so the current version of Zombie is very similar to Faerie. The primary difference for 2007 is a material evaluation scoring based on my research into Arimaa material evaluation.
There is a long ongoing debate over how to score the relative value of pieces in an Arimaa game. I've attempted to contribute to the debate via some more-or-less scientific studies of the historical game database. This has led to some evaluation algorithms and coefficients that differ remarkably from those formed through players' intuition. I may add more information later, but in the meantime here are links to the discussion threads relevant to my research.
The primary result of my research is that players and bots currently overvalue high pieces as compared to large numbers of small pieces.
I have generated this graph of the historical distribution of play ratings between bots and humans, showing how the performance gap opened dramatically between early 2004 and today.