For some, but not all, performance measures, there is an additional question of whether one is interested in peak performance, or average performance, considered over all these experiments. In evolutionary computing it is typical to suggest that algorithm A is better than algorithm B if its average performance is better. In many applications, however, one is often interested in the best solution found in X runs or within Y hours/days/weeks (peak performance), and the average performance is not that relevant. This is typical, for instance, in design problems as discussed in Section 9.1. In general, if there is time for more runs on the given problem and the final solution can be selected from the best solutions of these runs, then peak performance is more relevant than average performance.