Assessing the predictability of racing performance of Thoroughbreds using mixed-effects model

The inheritance of racing performance in Thoroughbreds is of interest to breeders and geneticists. Therefore, the genetic parameters of racing performance have been investigated in various populations of Thoroughbreds. However, the predictability of the racing performance of a racehorse has not been assessed well. In this study, we built mixed-effects models for Japanese Thoroughbreds and assessed their predictability of racing performance. We used the average velocity as an index of racing performance and treated each category of racecourse and distance as different traits. Model selection using the deviance information criterion showed that explanatory variables, such as race, age and jockey effects are important for racing performance. Using the selected models, the phenotypic values of horses born after 2009, adjusted using the entire dataset, were predicted with the breeding values estimated from a partial dataset until 2010. The correlation coefficients ranged from 0.000 to 0.235 (average of 0.084 ± 0.066) and were higher for longer distances. When predicting the graded race winners born after 2009 from the partial dataset until 2010, the area under the curve values ranged from 0.516 to 0.776 (average of 0.613 ± 0.073) and were also higher for longer distances. Although these results indicate the predictability of racing performance, further efforts, including exploring more suitable racing performance indices and refining statistical modelling, are required for improvement.