Abstract: A selection scheme for jumping sport horses is modelled with four stages of selection for males and one stage for females. The selection objective included three traits: conformation and gaits (CG, weighted 20%), competition jumping (CJ, weighted 60%) and a third trait (TT, weighted 20%) such as sperm quality or orthopaedic status. The first selection stage is based on knowledge of the pedigree with the aim of selecting horses suitable for CG test (at 3 years old) and CJ test (at 5 years old). The second stage includes the horse's own performance with respect to CG and CJ with the aim of selecting horses suitable for the TT test. The third stage is the selection of a limited number of males who are allowed to reproduce. The fourth stage (at 12 years old) takes into account the results of the horse's progeny. Females are selected in one step, whatever the number of performances measured at 5 years old. The annual genetic response was 9.4% genetic standard deviation of the objective, 2.6% for CG, 9.0% for CJ and 1.5% for TT. Results showed that selection by progeny testing did not contribute much to genetic response (12% of progeny issued from proven sires), the female pathway represented 26% of genetic response, TT was difficult to improve when the genetic correlation was unfavourable (- 0.6% genetic standard deviation for - 0.20 genetic correlation), and should consequently be directed towards the use of molecular markers. When compared with a selection scheme involving a station test, genetic response was the same if the breeding values used for selection before entering the station test took into account the results of the relatives for CJ and CG. This revealed the importance of an extensive performance test (like for competition performance) when designing breeding schemes for sport horses.