The role of balancing selection in maintaining diversity during the evolution of sexual populations to novel environments is poorly understood. To address this issue, we studied the impact of two mating systems, androdioecy and dioecy, on genotype distributions during the experimental evolution of Caenorhabditis elegans. We analyzed the temporal trajectories of 334 single nucleotide polymorphisms, covering 1/3 of the genome, and found extensive allele frequency changes and little loss of heterozygosities after 100 generations. As modeled with numerical simulations, SNP differentiation was consistent with genetic drift and average fitness effects of 2%, assuming that selection acted independently at each locus. Remarkably, inbreeding by self-fertilization was of little consequence to SNP differentiation. Modeling selection on deleterious recessive alleles suggests that the initial evolutionary dynamics can be explained by associative overdominance, but not the later stages because much lower heterozygosities would be maintained during experimental evolution. By contrast, models with selection on true overdominant loci can explain the heterozygote excess observed at all periods, particularly when negative epistasis or independent fitness effects were considered. Overall, these findings indicate that selection at single loci, including purging of recessive alleles, underlies most of the genetic differentiation accomplished during the experiment. Nonetheless, they also imply that maintenance of genetic diversity may in large part be due to balancing selection at multiple loci.