Current NDRPDR tests, when run repeatedly, can lead to high stdev of results. That is because they assume deterministic DUT behavior. We need new kind of test, which assumes probabilistic behavior, thus leading to more reliable results, at the cost of longer running time.