Variance component estimations for race performance of thoroughbred horses in Brazil by random regression model

A total of 306,698 racing performance data recorded between 1992 and 2002 from 25495 Brazilian Thoroughbred horses was analyzed to estimate the variance components using the Random Regression Model (RRM) compared to the Classical Repeatability Animal Model (CRAM). The performance was evaluated using the race time (in seconds) to run distances of 1000, 1100, 1200, 1300, 1400, 1500 or 1600 meters. The pedigree of each horse was extended as far as possible, with a total of 36,659 animals. The simple correlation between Breeding Value (BV) for the race time estimated by both procedures was high (r = 0.963) when the data were expressed at the mean level of all distances. However, with the RRM it was possible to estimate the genetic parameters and the BV of all animals at the trajectory of each one of the seven distances. Also, results clearly showed that a single BV estimated by CRAM, is neither an adequate nor a sufficient indicator for the selection of the best horses throughout the trajectory of distance performance. We concluded that the RRM procedure is highly recommended for the evaluation of racehorse performance.