The objective of this work was to determine the sufficient number of replicates for estimation of dissimilarity measures among maize cultivars. Data of five variables were used, which were evaluated in an experiment with 15 maize cultivars, in randomized block design with nine replicates. A number of 511 data files were formed, being 9, 36, 84, 126, 126, 84, 36, 9, and 1 obtained, respectively from 1, 2, 3, 4, 5, 6, 7, 8, and 9 replicates. Three dissimilarity matrices were determined between i and i’ cultivars containing, respectively, Euclidean, Manhattan, and Chebyshev distances. For each of the 105 distances between cultivars, in each dissimilarity measure, the power function was adjusted for the coefficient of variation of the (dependent variable) as a function of the number of replicates (independent variable), totaling 315 equations. For each equation, the abscissa axis value (Xs, sufficient number of replicates) was determined, corresponding to the maximum curvature point. With the increase of the number of replicates, there is an improvement in the accuracy of the estimates of dissimilarity measures among maize cultivars, however, the gains in precision decrease gradually. Six replicates are sufficient to estimate the dissimilarity measures among maize cultivars expressed by the Euclidean, Manhattan, and Chebyshev distances.