Choosing plant cultivars based on the probability of outperforming a check.

A major consideration in most plant breeding programs is the development of cultivars that have high probabilities of outperforming the check cultivar in a broad range of environments. Methods are presented for estimating and testing hypotheses regarding these probabilities, which are termed reliabilities. Reliabilities are shown to be directly related to several commonly used stability parameters. Data from international maize yield trials are used to illustrate and evaluate the repeatability of the approach. Results indicate that reliabilities can be useful aids to plant breeders since they (1) are easy to understand and compute, (2) are indices that weigh the importance of the difference in performance relative to stability, and (3) are potentially useful as genetic parameters since they are generally repeatable across randomly sampled sets of environments.