In any breeding program the decision of which crosses to make, among the thousands of possible combinations, is a crucial aspect that can impact future success. To select the best, breeders have limited information, which is often based largely on the performance and observations of the potential parents in field trials. Genomic resources and statistical tools allow us to include high density genetic information from the parents, estimating their breeding value, and using this information to predict the highest performing progenies from all possible crosses. Making this a precious aid for the selection process, gaining also in efficiency. The main objective of this study was to compare two cross selection strategies for grain yield: 1) traditional cross selection (breeders’ choice), and 2) parents breeding value based cross selection. We report the results from a field-based validation trial, where we compared 12 selected crosses, 6 best performing predicted crosses against 6 crosses selected by the breeder (traditional strategy), based on the performance in terms of grain yield of the progeny. From each cross a double haploid population of at least 52 lines was derived. The experiment consisted of 800 plots, divided into four individual but interconnected trials. In total, 624 progeny lines were evaluated, along with 12 parental lines. A partially replicated (p-rep) experimental design with two blocks was used, with 16% of the total experimental units was replicated. Each cross was represented with 52 DH lines. Individual trials consisted of 156 progeny lines (12 crosses, 13 DHs per cross), and 12 parents. Twenty progenies selected randomly, and the parental lines were repeated once, making a total of 200 plots per trial (156+12+20+12=200). Fungicide treatment was used to prevent interference of diseases in yield expression. A two-step statistical analysis was performed. First, phenotypic means were adjusted following field design and adding post-blocking factors (row and column). Second step was the evaluation of crosses and the cross selection strategies. Comparisons were made based on progeny mean, mean of the top 10% of the progeny, the variances of the predictions and the field data, and a quantile regression analysis. No differences in mean grain yield among crosses or selection strategy were detected. Indicating that selecting crosses based on breeding values for grain yield was at least as good as selecting crosses based on performance and field data.