Adverse Selection without Private Information: Evidence from Sports Prediction Markets