It is interesting to note that academic experimental EC research falls in the first category – there is always time to perform more runs on any given set of test problems. In this light, it is strange that the huge majority of experimental EC research is comparing average performances of algorithms.