[en] Selective breeding of Arabian horses in Algeria is not yet well developped. No accurate estimation of the breeding values of racing horses is carried out untill now. The estimation is based on one or several measurable traits, linked to race performance. To allow for an accurate estimation, it is important to determine the part of the phenotypic variability that is due to non-genetic factors (environmental). This first study estimates phenotypic correlations between the recorded traits in order to deduce to what extent the different measures bring different information on animal performance. Then it estimates the influence on performance of non-genetic factors such as sex, age, year of performance and the interactions between these various factors. From 1995 to 2007, the data related to flat racing was collected. The General Linear Model was used to identify and quantify the non-genetic factors affecting racing performances. Three traits were used: two earnings traits (the logarithm of annual virtual earnings: LAEV and the logarithm of average annual virtual earnings per start: LAEV/S), and one ranking trait (the ranking transformed and normalised: PERF). Significant high positive correlations was observed between the three traits (p<0.001), which thus give account of similar informative aptitudes. The effects of sex, age, year as well as the interactions between age and sex and between age and year turned out to be significant for the three traits (p<0.05). These results indicate the need to adjust the earnings and ranks traits in order to use them as criteria of selection of racing Arabian horses.
