Following Ali (1977), horses are grouped by `favorites’ . The horse with the
lowest odds Oi to win a race is known as the ® rst favorite; the horse with the
second lowest odds Oi to win the race is known as the second favorite, and so on.
On the basis of information of win odds on the horses in a race, the likelihood
estimates of the parameters a i of each model are obtained. Following the analysis
of this data by Ali (1977) and Lo (1992), the parameter d is set to 1.16 for the
races at Saratoga, and to 1.13 for the races at both Roosevelt and Yonkers. We
then use the estimated model to estimate the probability that favorite i ® nishes
second ( p(i, 2)) or third ( p(i, 3)). These are model-based probabilities. These
probabilities are obtained by numerical integration, utilizing the following formulae: