There have been many attempts, theoretical and empirical, to explain the persistence of a favorite-longshot bias in various horse betting markets. Most recently, Snowberg and Wolfers (2010) have shown that the data for the US markets support a misperceptions of probability approach in line with prospect theory over a neoclassical approach of the Quandt (1986) type. However, their paper suffers from two basic difficulties which beset much of this literature. First, the theoretical model used fails to allow for the existence of horse betting markets which either display no such bias (or a reverse bias) as in Hong Kong and at least one large Australian market (Busche and Hall, 1988, Schnytzer, Shilony and Thorne, 2003 and Luppi and Schnytzer, 2008). Second, econometric testing and theoretical modeling are facilitated by the highly unrealistic assumption that the betting population is homogeneous with respect to either information or attitude to risk or (usually) both. Our purpose is to show that allowing for heterogeneous betting populations (in terms of both attitude to risk and access to information) permits the explanation for the different biases (or their absence) observed in different markets within a strictly neoclassical framework of rational bettors. We conclude with empirical support for our model.