Wild crop relatives are an important source of genetic variation for improving domesticated species. Given limited resources, methods for maximizing the genetic diversity of collections of wild relatives are needed to help spread protection over a larger number of populations and species. Simulations were conducted to investigate the optimal strategy of sampling materials from populations of wild relatives, with the objective of maximizing the number of alleles (allelic richness) in collections of fixed size. Two methods, based on assessing populations for variation at marker loci (e.g., allozymes, restriction fragment length polymorphisms), were developed and compared with several methods that are not dependent on markers. Marker-assisted methods yielded higher overall allelic richness in the simulated collections, and they were particularly effective in conserving geographically localized alleles, the class of alleles that is most subject to loss.