Empirical Evidence of Error in Pricing of Favorites and Longshots in Greyhound Racing
Gulati, Anil, Shetty, Shekar, Journal of Economics and Finance
Sports bettors' success depends on the ability to accurately assess the true probability of outcomes. Successful racetrack bettors can realize returns better than the track take out. Historical empirical evidence shows the presence of favorite-longshot bias (FLB) in horse racing where bettors underbet favorites. Conversely, bettors overbet longshots. We tested for FLB bias in racing data from three greyhound racetracks. Our results show opposite behavior. We show bettors apparently underestimated for longshots, and overestimated for favorites, the true probability of winning. In 10 out of 14 grades bettors significantly overbet favorites, and underbet longshots in 8 out of 14.
(JEL G13, D84)
Efficient market hypothesis states that the price of a risky asset is an unbiased estimate of its fundamental value (Fama, 1991). Generally speaking, traditional financial instruments pose difficulties in testing this hypothesis as exogenous measures of fundamental value in those contexts are complex and harder to estimate. Wagering markets in general, and racetracks in particular, offer researchers rich experimental settings for testing asset pricing theories of uncertain contingent claims, of which efficiency of price discovery under risk conditions is an example. Unlike equity markets, races have a short and finite life span. Bettors' decisions are validated upon conclusion of the race at which point the true value of the asset is revealed. Economic theory assumes that decision makers are rational in their asset allocation among risky outcomes. Decision situations with imbalanccd risk/reward ratios invite adjustments until those imbalances disappear and the market reverts to risk neutrality. In the long run, assets offering excess returns attract buyers which drive returns lower towards normality. Unlike racing markets, price adjustments for financial instruments are either perpetual or at least relatively long lived. In the racing context, the distribution of bets in a race reflects the market's subjective estimates of risk.
Worldwide, parimutuel form of wagering is the most frequently used type in racing markets. In this form of wagering, money is pooled by each bet type. From each pool, the market maker (racetrack operator) extracts a fixed portion called the track take (w1) to cover expenses, wages, taxes, and profits. Remainder of the pool is returned to winners in proportions to the amount wagered with zero returns to losing wagers.
Racing market participants respond to the risk/reward imbalances with their betting dollars. Assuming no additional dollars are wagered on competing entries in the race, every additional dollar wagered on an entry reduces its expected return as the winning pool is shared by more dollars. Conversely, return from an entry increases when additional dollars are wagered on one or more competing entries, even though the entry itself did not receive any additional bets. In the long run, under performing entries will attract fewer betting dollars relative to the field, thus bringing the risk/reward ratio into balance. Similarly, if the actual fraction of the pool wagered on an entry is consistently smaller than its success rate, those wagers will realize excess returns. The market will notice the winning potential of that contestant by increasing the bet fractions thus bringing the risk/reward ratio into balance. Opportunities to generate superior returns, therefore, are a function of the collective error the market makes in estimating the true probability of winning.
Review of Past Studies
It has been documented that the relative amounts of bets in parimutuel racing markets closely track empirically observed win frequencies thus supporting the conclusion that bettors are able to detect differences in the probabilities of successful outcomes in these type of risky situation and bet accordingly (Griffith, 1949). Additionally, several studies provide empirical evidence of presence of an anomaly called the favorite-longshot bias (FLB). …