University of Chicago law prof Cass Sunstein has a useful post defending prediction markets like Intrade against critics who bashed them for failing to foresee Hillary's win in New Hampshire. As he correctly notes, long shots sometimes do win:
Intrade had Clinton at about 8% likely to win, and 8% of the time, 8% chances should come through. (As in horse races, so on Intrade.) It would be a great surprise if 8% chances come though 0% or even 2% of the time. But on prediction markets, prices have been close to probabilities. Events that are priced to be 90% likely to happen occur about 90% of the time, events that are priced to be 70% likely to happen occur about 70% of the time, and so on.
The problem is that (a) people don't understand probability and (b) prediction markets have been oversold. There's an ongoing scholarly debate about the predictive value of prediction markets, but for the moment, we should all be able to agree that what happened in New Hampshire (a single event) proves little either way.