Most of agronomically important characters are biometric traits. An improvement of these traits in cultivated plants by deriving segregants superior to parents, which could be developed as cultivars, is a main goal in breeding of self-pollinated crops. Two problems need to be solved: when will the progeny be better than its parents and how can a genetic potential of a given pair of parental genotypes be predicted? In this paper, transgressive segregation in homozygous barley populations is shortly reviewed. Various approaches to choosing parental forms are shown, and a theoretical method for predicting the frequency of transgressive segregants in a homozygous population is presented. Additionally, relationships between parental diversity estimated with molecular markers and the progeny performance are discussed. Although the prediction of transgressive segregation is still a problem, it seems promising to apply an approach measuring the performance of the parental genotypes and estimating their genetic distance by molecular markers.