I disagree, poker robots aren't all that impressive. Against players who consistently win against humans they've been shown to be pretty bad. All it takes is for someone to realise they're playing against a machine.

Once the realisation that an opponent is a bot is made it's a matter of finding the patterns. The vast majority of the bots will play percentages and NOTHING else, it will bet certain amounts based on the odds of winning etc. With this in mind a person can learn to read a bot perfectly 100% of the time because it has no random factor. Even the bots that do have been proven not random, programming can only create pseudo random unless they use some sort of environmental noise function as part of the algorithm (like wind).

Image of random bitmap generated by normal program: http://www.boallen.com/assets/images...p_computer.png

Image of random bitmap generated with true random use of environmental noise: http://www.boallen.com/assets/images...itmap_true.png

Implementing true random into a bot is pretty difficult and due to the time sensitive nature of turns it's almost impossible without extreme processing power.

A player can milk a bot entirely dry by reading him consistently. Good players don't lose to bots, bad players do. The people who make their livings playing poker don't have much trouble with them.

Bots do however turn a profit despite losses such as this because there are far more bad players than there are good players.

Used to play a bit once upon a time, figured I could make something of it, turns out I'm not quite good enough, others have proven the bots to be worthless against good players though.