I have a primitive program  called "Occam" playing on the chess server now. I don't have any kind  of [aspiration search](Aspiration_Windows "Aspiration Windows") in it, just pure [alpha/beta](Alpha-Beta "Alpha-Beta"), no [PVS](Principal_Variation_Search "Principal Variation Search"), no [MTD](MTD(f) "MTD(f)") or anything. I will implement PVS first so that I  can do comparisions. 

Soon after the challenge was announced, several members of the game AI-research community attempted to tackle the challenge. The late Don Dailey (developer of the [Komodo](Komodo "Komodo") chess engine) was the first to create an Arimaa engine. Within a couple months Don had a surprisingly strong program offering games in the Arimaa gameroom. In designing Arimaa, I had used the [Zillions-of-Games](Zillions_of_Games "Zillions of Games") [general game-playing](General_Game_Playing "General Game Playing") engine, which allows specifying the rules of the game in a [Lisp](index.php?title=Lisp&action=edit&redlink=1 "Lisp (page does not exist)") like language, and then immediately being able to play the game against the Zillions engine. Using only look ahead and no game-specific knowledge, this engine was able to play at a strong level in most games one could dream of. However, it was incredibly lousy at Arimaa due to the high branching factor. This helped build my confidence that Arimaa would be a difficult game for computers if only a [brute-force](Brute-Force "Brute-Force") search was used. However, Don’s program, called OCCAM, surprised me in how well it played when the search engine was much faster and included some game knowledge. When Don realized that there was very little expert knowledge available for Arimaa, he felt that he could not make much progress on his program and went back to developing his chess engine which now has gone on to become the highest rated in the world.