Unfortunately, the probabilities that arise in numerical forensic methods are not correct simply by fiat, as the probabilities in risk-limiting audits are. That is because the probabilities do not arise from the auditor selecting a random sample in a controlled, deliberate way, or from any other mechanism known to or controlled by the investigator. Rather, the calculated probabilities derive from assumptions about the probability distribution of clean election results: how the numbers “should” behave absent error, fraud, or manipulation, and how that behaviour does or does not vary with other factors.