1st let me start off by saying the Poki has several heads up versions. Some of those versions are harder than others. I think the original version of Poki tried to play heads up "optimally". In game theory "Optimally" means "cannot be beaten". Playing paper/rock/scissors optimally would be like playing rock, paper, and scissors each with probability 1/3. Notice that anyone who plays this strategy can't be beaten by any other strategy in the long run. But, this strategy won't beat a moron who ALWAYS ALWAYS ALWAYS plays rock either so take the word optimal with a grain of salt.
The problem is that heads up holdem is much more complicated than paper/rock/scissors. So actually figureing out an "optimal strategy" is very very difficult. The Alberta team made a valient effort to reduce the complexity of the problem as best they could without effecting the playing ability of the resulting Poki bot.
Which brings us to this problem.
[ QUOTE ]
However it frequently would call to the river with 1 overcard to the board and no draw.
[/ QUOTE ]
Believe it or not this is actually correct if the bot is going to be "optimal". This way if you are trying to beat the bot you can't just bet,bet,bet,bet your way to victory. The bot will every now and then call you down with something like Q high. So, if the bot didn't call you down like this sometimes you could exploit this weakness and beat the bot.
Later on the bot was evolved to exploit weakness of its opponents. I have not played that version of the bot. But another poster said that it was at least competent
For a detailed explination of the older bot look here
academic paper about heads up holdem