The main goal of plant breeders is to create and select genotypes that are well-adapted (i.e. that produce high yield) in future growing conditions. Plant adaptation depends on the genotype (particular DNA composition of an individual), the environment (defined by the meteorological, soil and management factors in the growing area of interest) and the genotypic sensitivity to the environmental conditions. For example, some genotypes might be more tolerant to diseases than others, or they might differ in their efficiency to use water or nutrients to produce yield. When genotypes show different sensitivities to the environment, genotype by environment interaction is observed. Genotype by environment interaction can cause that the best genotype for one condition might not be the best for another condition, complicating the selection of superior individuals. To increase the chances of creating and selecting well-adapted genotypes, breeders need to evaluate thousands of individuals across a number of environments (years and locations). Furthermore, multiple plant characteristics (traits) are often important for farmers and consumers and they need to be taken into account in the selection process. The availability of molecular markers allows breeders to identify the genetic basis and predict yield or other traits of interest, speeding up the selection process. The large number of genotypes, environments, traits and molecular markers, commonly leads to large data sets, making the use of powerful and flexible statistical models into a pre-requisite for the production of well-adapted varieties. In this thesis, we propose and discuss strategies based on statistical and/or crop growth models to assist breeders in the process of selecting of well-adapted genotypes. We show examples of wheat, barley, maize and rice.