Forecasting horse race outcomes: New evidence on odds bias in UK betting markets

Using bookmaker odds data derived from ten seasons of Flat Racing in the UK, we confirm the existence of a favourite-longshot bias; i.e., the expected return to level stake bets placed at shorter odds exceeds that to level stake bets placed at longer odds. We find, however, that the degree of bias declines significantly over the period studied, with a lower bias being evident in the five years after 2000 than in the five preceding seasons. These findings have important implications for methods of forecasting race outcomes that include odds as an endogenous variable. In particular, we conclude that the parameters of the bias need to be re-estimated through time in order to generate the most efficient forecasts based on the prevailing odds distributions.