By means of a population genetical model, we study the evolution of segregation distortion. Most models of segregation distortion focus on a single distorter allele. In contrast, we consider the competition between a large number of distorters. Motivated by systems as the t complex of the house mouse or the Sd complex of Drosophila melanogaster, we assume that there is some "complementation" between distorter alleles, i.e. that the fitness of individuals heterozygous for two distorter alleles is higher than the fitness of homozygous individuals. In the presence of complementation, the most efficient distorter allele with the highest segregation ratio often does not outcompete less efficient distorters. In fact, our results show that coexistence of a large number of distorter alleles is more typical than the competitive exclusion of less efficient distorters by a single superior allele. We first consider the analytically tractable system where all distorters show the same amount of complementation. In this case, all distorters with a segregation ratio higher than a certain critical value will persist, resulting in a polymorphic population where the average segregation ratio is only slightly larger than 0.5. If the degree of complementation varies, there may be more than one stable equilibrium, and the outcome of competition may depend on the initial conditions. Motivated by empirical examples, we also consider the case that the distorting ability of an allele is negatively related to its effects on individual fitness. Interestingly, the outcome of competition depends crucially on details of such a trade-off. We conclude that verbal arguments are insufficient to predict the evolution of segregation distortion.