You're right, I should have been less specific. If the harm of false positives is significant you may not want need to have random or worse than random results to feel obligated to stop the project.