For people interested in the use of neural nets to evaluate board position, it might be worthwhile reading up on TD-gammon which this work builds Off of