dbacl project homepage: Can a Bayesian spam filter play chess?

interesting...

wonder if, instead of basing answers on what the previous people's moves were after the same sequence of play, if the moves were based on what previous people's moves were when the board was in the same configuration? That would take into account the piece values...