An attempt to predict the earning status of a thoroughbred in France by genealogical data

Abstract

The objective of this study was to create indicators to predict which horses had more chances to be placed in races than others. The study was limited to genealogical independent variables (i.e. concerning sires, dams, paternal and maternal half sibs of the horse). It concerned the racing career of 60 851 thoroughbreds born in France between 1980 and 1995 (i.e. annual earnings in flat races of 2 to 5 year-olds and in jumping races of 3 to 5 year-olds). The earning status is the dependent variable; since it is bi-modal, a logistic regression was chosen as the appropriate statistical method. A stepwise procedure with a threshold of 0.001 allows to select adequate independent variables. The out coming figure of this purely phenotypical approach leads to the following conclusions: it is not possible to predict the probability of being placed as a thoroughbred in France only from genealogical data; the most valuable information available to predict this probability, is the previous performance of the horse itself; there is a great dissymetry between the information given by the dams' performances and that given by the sires' performances, which can be ignored; it is also noticeable that the best predictors of a given earning status are those obtained for the family for the same age and discipline. It can be inferred that genetic correlations between age and discipline are different from 1; these correlations are clearly negative between flat and jumping races. This may also be the case in flat races for the correlations between performances at 2 years and performances at 4 and 5 years. These negative relations never appeared in earlier studies since they were based on selected data of placed horses.