The above graphs plot the fitness levels of the 20 agents over the entire simulation. Low mutation leads to diverse, and stagnant strategies. Thus the racers do not learn from eachother, and their fitness levels diverge. High mutation forces the racers to interact and learn, causing their fitness levels and strategies to converge. The lower plots illustrate the "just right" mutation rate. Low mutation causes fitness levels to diverge over time, while high mutation causes them to cycle. The "just right" mutation rate would result in lower, but not zero, variance over time.