Results obtained with stochastic dominance, linear risk-programming (MOTAD), regression line evaluation and expected value calculations on maize-cultivar selection under varying circumstances were compared. Maize cultivar trial results of four localities over 14 years were used. Two levels of management were included. Cultivar selections obtained by the different methods show a remarkable similarity. This is probably the result of the relative normal distribution in cultivar gross margins. The largest single disadvantage of cultivar selection by adaptation regressions and expected value calculations is that it considers only the value of the gross margin and largely ignores the variance. This can lead to selection of ill-adapted cultivars. Stochastic dominance and MOTAD take both the expected value and variance into consideration. MOTAD is particularly useful as it invariably selects cultivar combinations.