New Scientist reports on the success of a new (Canadian!) poker program, Polaris, at beating some top poker players. As with Deep Blue, this success is not exactly decisive as it was restricted to one-on-one play of Limit Texas Hold 'Em. This is surely the easiest version of poker to master.
The article mentions both the lack of perfect information and the contextual nature of the best play as complications in programming poker programs: "One of the fundamental problems for any poker player is that the best strategy varies, depending on your opponent's style of play." While it would be easy to see how simple calculations would handle the lack of information, I am more interested in seeing how programmers can deal with the interactions between playing style and optimal no-limit betting!