Bayesian estimation of genetic parameters and comparison of statistical models for race performance traits in the Turkmen horse

Racing performance is an important component of genetic evaluation and breeding decisions in racehorses, but it is a complex trait influenced by genetic and multiple non-genetic factors, including race conditions, management, handicap, and jockey effects. Different specifications of these effects may produce different estimates of genetic parameters and breeding values, highlighting the need to compare alternative model structures. This study aimed to estimate genetic parameters, compare alternative statistical models, and evaluate their suitability for genetic evaluation of racing performance traits in the Turkmen horses bred in Iran. Data comprised 10 830 official race records from 1 667 horses, including 899 stallions and 768 mares, collected from 2003 to 2023. Three racing performance traits were evaluated using eight hierarchical models that differed in their random-effect structures. Model 8 (M8) was selected as the best-supported model for all three race performance traits, including race completion time (RCT), rank at the end of competition (REC), and average speed (AS). Heritability estimates were 0.006 ± 0.001 for RCT, 0.06 ± 0.01 for REC, and 0.004 ± 0.001 for AS. Prediction bias ranged from 0.12 to 0.20, and prediction accuracy ranged from 0.60 to 0.78. The low heritability estimates indicate that additive genetic variance accounted for a relatively small proportion of phenotypic variation in this population under the models considered. Although the parameter estimates and model rankings are population-specific, the findings illustrate a broader methodological consideration for genetic evaluation of racing performance: model specification should account for relevant race-level and jockey-related effects when these factors contribute substantially to observed performance. This consideration may inform genetic evaluation strategies in other racing horse populations with similarly complex performance environments.