This paper presents results on the use of problem generators to analyse the performance of a genetic algorithm, which combines reproduction, crossover, mutation and an inversion operator. The main conjecture is that inversion may be useful for particular classes of problems depending on the levels and type of epistasis present. Inversion works by reversing the order of genes between two randomly chosen positions within the chromosome. While other genetic operators search for good combinations of alleles, an inversion operator has the ability to search among good string arrangements. Interaction (also called epistasis) between genes means that the contribution of a gene to the fitness depends on the fitness of other genes in the chromosome. By using a number of problem generators to alter both the level and type of epistasis and by varying the rate of inversion, we develop an empirical methodology to analyse a classic inversion operator in a simple genetic algorithm and present results of such analysis.