In order to find early selection criteria to improve the longevity of show jumping horses in competition, a specific protocol was constructed. Before entering competition, young horses were measured for many traits. These horses were offspring of two groups of sires selected as having the highest and lowest estimated breeding values for functional longevity in jumping competition, as calculated from progeny. Functional longevity was defined as the time spent in competition corrected for the level of performance. The dataset included 952 horses (mainly French Saddlebred) and 77 blood parameters. Heritability was estimated using a mixed model including the effect of age, sex, place and date of collection, weight and animal random additive value with 10,280 horses in pedigree. Heritability of blood parameters was generally moderate to high: 21 values were higher than 0.5 and 39 were between 0.2 and 0.5. The most heritable traits were hematology and enzyme traits: mean corpuscular volume (0.90) and mean corpuscular hemoglobin concentration (0.92) but also traits as liver isozyme (0.72) or total alkaline phosphatase (0.68), small lymphocytes (0.67), superoxide dismutase (0.60). Principal Component Analysis reveals the low correlations between traits: 32 variables were required to achieve 90% of the variance explained. However, correlation patterns between variables in the same function group revealed functional relationships. Logistic regression to predict group of sires according to longevity revealed, in order of appearance, the effect of mean corpuscular hemoglobin concentration, leukocytes, liver isozyme of alkaline phosphatase, 2-globulin, monocytes, 1-globulin and aspartate transaminase. Biological explanations are underway.