If prediction markets are efficient, an event with a contract trading at 4 out of 100 should occur 4% of the time. If it never occurred, the contract would be overpriced. A better test would be to take a bunch of contracts trading at low values today, and then check in the future to see how many actually happened.