According to the favorite-longshot bias, the expected return on an outcome tends to increase in the fraction of bets laid on that outcome. We derive testable implications for the direction and extent of the bias depending on the ratio of private information to noise present in the market. We link this ratio to observables such as the number of bettors, the number of outcomes, the amount of private information, the level of participation generated by recreational interest in the event, the divisibility of bets, the presence of ex post noise, as well as ex ante asymmetries across outcomes.