Conversely, a skilled adversary can always fabricate fraudulent results in such a way that the numbers will pass any given suite of statistical tests. There is evidence that this occurred in a case where the mean of the second digits fits the predictions of Benford’s Law essentially perfectly, and the mean of the terminal digits fits a uniform model essentially perfectly (Kalinin & Mebane, 2017). Kalinin and Mebane conclude that this agreement is too good to have occurred naturally, and argue that it is a deliberate signature by fraudsters. More generally, there is no such thing as a statistical test that has high power against all alternatives (Freedman, 2010).