Better selection of the best

We present a new method of selecting the best of several competing system designs on the basis of expected steady-state performance. The method uses a new form of time-series bootstrap in a sequential hypothesis testing framework. We illustrate the efficiency of the new method by comparing its results against the current state of the art in a series of simulation experiments. The new method achieves equal or greater probability of correct selection of the best system with substantial savings in the number of simulation observations required for the decision.