The above graph shows the relationship between the mutation and crossover operators. As the probability of mutation increase, the average frequency of crossover rises.
The above graphs highlight mutation's effect on learning and "efficiency." First, consider the low mutation case. The far left point represents the racer currently in the lead. She has been riding at a relatively fast pace, with no push from behind and no help in front. Since mutation is so low, she will probably never adopt a strategy that places behind a rider. This is a pity, because if she did, she would adopt a strategy that increases her fitness. The moderate mutation case keeps riders closer together. They gain from riding in packs, and frequently update their strategies to imitate the high fitness riders in the near proximity.