The objective of this study was to develop a new multivariate statistical model for genetic estimation of distance-dependent racing performances in German Thoroughbreds. Analysed performance traits were "square root of distance to first placed horse in races over sprint distances (until 1 400 m)", "square root of distance to first placed horse in races over mile distances (from 1 401 m to 1 900m)" and "square root of distance to first placed horse in races over long distances (over 1 900 m)". These traits were found to be influenced by the carried weight, which was determined by the horses' earlier performance. Therefore, new traits were developed based on random regression models, which were independent from the carried weights. Genetic parameters were first evaluated for these new created traits "new distance to first placed horse in races over sprint distances" (h(2)=0.088), "new distance to first placed horse in races over mile distances" (h(2)=0.081) and "new distance to first placed horse in races over long distances" (h(2)=0.137) using a multivariate animal model. Genetic correlations between these traits were high, but differed from r(g)=1. A further heritability was estimated for the distance-independent trait "new distance to first placed horse in races over all distances" (h(2)=0.101) applying a univariate animal model with a fixed distance effect. These two different models were compared by two criteria. The ranking of breeding values for the distance-independent trait (estimated with the univariate model) was first correlated with each of the rankings of breeding values for the three distance-dependent traits (estimated with the multivariate model). Correlations varied from r=0.668 to r=0.813. The second criterion for comparison was the percentage of incorrectly selected raced stallions by breeding values estimated with the univariate model. Between 47.4% and 69.7% of stallions were incorrectly selected. The use of a total selection index including breeding values of the three distance-dependent traits with suitable weightings was recommended as a possible future selection criterion.

In a genetic analysis of German trotters, the performance trait racing time per km was analysed by using a random regression model on six different age classes (2-, 3-, 4-, 5- and 6-year-old and older trotters; the age class of 3-year-old trotters was additionally divided by birth months of horses into two seasons). The best-fitting random regression model for the trait racing time per km on six age classes included as fixed effects sex, race track, condition of race track (fitted as second-order polynomial on age), distance of race and each driver (fitted as first-order polynomial on age) as well as the year-season (fitted independent of age). The random additive genetic and permanent environmental effects were fitted as second-order polynomials on age. Data consisted of 138,620 performance observations from 2,373 trotters and the pedigree data contained 9,952 horses from a four-generation pedigree. Heritabilities for racing time per km increased from 0.01 to 0.18 at age classes from 2- to 4-year-old trotters, then slightly decreased for 5 year and substantially decreased for 6-year-old horses. Genetic correlations of racing time per km among the six age classes were very high (rg = 0.82-0.99). Heritability was h2 = 0.13 when using a repeatability animal model for racing time per km considering the six age classes as fixed effect. Breeding values using repeatability analysis over all and within age classes resulted in slightly different ranking of trotters than those using random regression analysis. When using random regression analysis almost no reranking of trotters over time took place. Generally, the analyses showed that using a random regression model improved the accuracy of selection of trotters over age classes.

The genetic associations between racing performance and preselection of horses considered as the binary trait racing status (trotters without or with at least one racing performance in life were classified as 0 and 1, respectively) as well as disqualified races (disqualified and non-disqualified trotters were classified as 1 and 0, respectively) were analysed in German trotters. Variance components for racing performance traits square root of rank at finish, racing time per km, and log of earnings with racing status were estimated based on an animal model using REML. Heritabilities of racing status, racing time and rank at finish were 0.30, 0.21, and 0.06, respectively. The genetic correlations between racing status and racing time or rank at finish were -0.74 and -0.32, indicating that horses started at least once showed a higher genetic potential in racing time or finishing ability than never started horses. This showed the high preselection of German trotters especially based on racing time. To account for this preselection, it was recommended for additional use of racing status in the German evaluation system. Breeding values of the three racing performance traits were estimated by two distinct models, in- or excluding racing status and compared by using three criteria. Racing time per km showed the highest correlation (r = 0.98) between breeding values evaluated by these two distinct models. Therefore, incorrect selection rate of horses using breeding values from the model without racing status, was lowest for racing time per km (9.7%). Selection response increased about 1% for this trait after including racing status in the model. For the estimation of rank at finish, inclusion of racing status in the multiple trait model was much more important as indicated by a low correlation between breeding values (r = 0.29) and high percentage of incorrectly selected stallions (97.5%). The trait disqualified races was first analysed using an univariate threshold model. Heritability of this trait was low (h(2) = 0.12) and repeatability (r = 0.43) showed a moderate magnitude. Using a linear multiple trait animal model, disqualified races showed a low heritability (h(2) = 0.05) and a moderate favourable genetic correlation (r(g) = 0.43) with racing time per km. Consequently, selection on racing time per km is expected to improve indirectly the reliability of racing performance. Combined selection of reduction in disqualified races and racing time may even further improve the reliability of racing trotters.

The objective of this study was to develop new statistical models for genetic estimation of racing performances in German thoroughbreds. Analysed performance traits were "square root of rank at finish", "square root of OF distance to first placed horse in a race" and "log of earnings". These traits were found to be influenced by the carried weight, which was determined by the horse's earlier performance. Therefore, new traits were developed based on random regression models, which were independent from the carried weights. Heritabilities were first estimated for these created traits "new rank at finish" (h(2) = 0.101) and "new distance to first placed horse in a race" (h(2) = 0.142) by using two univariate animal models. When considering a linear regression of carried weights as fixed effect in the statistical model, heritabilities for "square root of rank at finish" (h(2) = 0.086) and; square root of distance to first placed horse in a race" (h(2) = 0.124) decreased. Breeding values of "new rank at finish" and "new distance to first placed horse in a race" were compared with breeding values of "square root of rank at finish" and "square root of distance to first placed horse in a race", in which carried weight was considered as fixed regression in the model. These two different models were compared by two criteria. Breeding values were overestimated for low performing thoroughbreds and underestimated for high performing horses when considering a linear regression of carried weights as fixed effect in the model. Statistical models considering new created traits ("new rank at finish" and "new distance to first placed horse in a race") which were independent of carried weights, showed better suitability for genetic estimation. Due to high genetic correlation with other traits and showing highest genetic variance a univariate animal model for the trail "new distance to first placed horse in a race" was recommended for genetic estimation.

Two different statistical models considering racetrack or individual race as fixed effect were compared, regarding genetic parameters and by using cross validation. Data for variance component estimation consisted of 48,942 performance observations from 4249 trotters. Variance components for the traits square root of rank at finish, racing time per km, and log of earnings per race were estimated by REML using two multiple trait animal models involving different racetracks or individual races. When including each individual race instead of racetracks in the statistical model, heritabilities increased from 0.05 to 0.07, 0.19 to 0.23, and 0.08 to 0.09 for square root of rank at finish, racing time per km, and log of earnings per race, respectively. Genetic and phenotypic correlations among traits increased also after consideration of individual races. Square root of rank at finish, as well as racing time per km and log of earnings per race, was highly genetically correlated with −0.99 and −0.88. The two statistical models were compared on the basis of their predictive ability by using cross validation. Data for these analyses consisted of 706,082 observations from 21,363 trotters. Randomly eliminated performance observations were predicted by cumulation of fixed and random effects obtained from estimation of breeding values for both models. Estimates for racing time showed lower bias and mean square error (MSE) when considering individual races instead of racetracks. Also, the correlation between predicted and true phenotypic value increased from 0.85 to 0.92. Estimates for square root of rank at finish were unbiased, but with a higher MSE when considering individual race effect. A similar high bias and MSE with both models were obtained for log of earnings. In order to avoid bias in estimation of genetic parameters and breeding values for racing time and square root of rank at finish, inclusion of each individual race in the statistical model was recommended.