I don't know much about probabilities and statistics, and I only know in summary form how game theory is applied as a means for prediction of human behavior. But I do think that any statistical, probabilistic, or game-theoretical means, any attempt to measure, quantitatively, in the social sciences is admirable. If ultimately it can't capture the extent of human behavior it would like, it is at least worth the effort. If it is a failure, it will just bear no fruit, and so much the worse for the theoretical underpinnings but not for our effort.
A new article called "How to Win at Forecasting" is about a government-funded project that is an attempt to put to the test the putative expertise of political scientists. It is a five-year research project.
So far, interestingly enough, the man who has been conducted the project has said that he has noticed in previous instances that the people who make the best kinds of predictions not only make more modest predictions but they are often people who believe that making these kinds of predictions is not possible or not a good idea! They are called in the article "foxes," as opposed to the hedgehogs who are more likely to think that human behavior works like clockwork and is quite predictable. The "foxes," on the other hand, often think that human behavior is too complex to make accurate predictions. And yet those people tend to make accurate predictions.
I'm a little sleepy, and maybe I haven't explained this as well as I'd lile. But anyway you can read the article here: http://www.edge.org/conversation/win-at-forecasting. See what you think.