That's spectacular.

Awesome recursion

hahaha!

Yes and no. Computers are actually phenomenal at playing poker, because they always know the exact odds, and can fairly easily build general models about their opponents through machine learning. What's missing, and what sometimes makes it look as though they stink, is that no program I'm aware of reads facial expressions and body language to fill into their opponent model, meaning that computers miss obvious tells. That's honestly more an issue at this point of someone taking the time to put all the pieces together, so I wouldn't really regard this as a major thing left.

A bunch of OpenCV code should solve this...

Yeah; detecting facial expressions shouldn't be too hard anymore, and given that computers can even trivially do things like detect pulses and blood pressure from video feeds, it'd be easy to program the computer to catch obvious fake facial expressions. It might not be as good as a human, but it'd be good enough.

Funny you should mention that -- the New York Times just posted a fascinating article about a computerised Texas Hold 'Em machine that seems very, very capable http://www.nytimes.com/2013/09/08/magazine/poker-computer.html

Fascinating piece. The difference is the bot in the NYT story plays limit hold 'em, which is a much more solvable problem, math-wise. No-limit (tournament style) hold 'em is much trickier for computers, because the range of possible moves is so much greater, and the human behavior is much harder to model.

Aha, thanks for the clarification.