is to be solved with a stochastic optimization algorithm (SOA) and that there
is only a fixed budget of function evaluations for finding the best solution
in each period. The SOA is run several times and the best solution found is
used. How many function evaluations should be allocated to each run?
Should we make many short runs or should we prefer few long runs?
This question has a (known) analytic solution provided the distribution of
the first hitting time is known. We discuss how to approach the optimal
solution if this distribution is not or only partially known.