Selective breeding is the deliberate choice of specific sires and dams based on heritable traits the breeder wants to increase — speed, conformation, temperament, coat color, disease resistance — with the effect of raising the frequency of favorable alleles across successive generations. It is the mechanism by which every established horse breed was created and the primary tool of modern breed improvement programs.
The effectiveness of selective breeding depends on three factors: the heritability of the target trait (how much of the variation in that trait is genetic rather than environmental), the selection differential (how far above or below average the selected parents are for that trait), and the generation interval (how often new offspring can be evaluated and selected). Traits with high heritability — bone density, body dimensions, some performance metrics — respond predictably to selection; traits with lower heritability require larger populations and more generations. The foal-to-breeding-age interval of roughly three years makes the horse a slower species to select in than, for example, cattle or pigs.
Long-term selective breeding can narrow the genetic base of a breed, increasing the risk of inbreeding depression — reduced fertility, immune function, and viability in the offspring crop. Breed registries manage this through outcrossing policies, careful pedigree analysis, and, in some breeds, controlled infusion of outside genetics. The relationship between a closed studbook and a performance-oriented open studbook reflects different trade-offs between trait concentration and genetic diversity.
Further Reading: Wikipedia’s article on selective breeding explains heritability, selection differentials, and the population-genetics basis of breed improvement across species. The Wikipedia entry on progeny testing covers how offspring records are used to infer breeding value more accurately than phenotyping the parent alone.