Discussion in this paper focuses on how strongly we should prefer non-punishment of persons guilty of serious crimes to punishment of persons innocent of them. William Blackstone's version of that preference, expressed as a ten to one ratio, is first shown to be untenable on standard accounts of legal punishment's justifying aims. Somewhat weaker versions of that ratio also appear suspect. More to the point, Blackstone's adage obscures the crucial way in which there are risks to be assessed in setting up a criminal justice system – the risk that it will not be eff ective enough at apprehending and punishing serious off enders, as well as the risk that it will expose innocent persons to harsh punishment. I urge a balancing of such risks, and argue that the salient features of many contemporary criminal justice systems can be plausibly interpreted as attempting to achieve and maintain such a balance. We do not grant criminal defendants a presumption of innocence that the state must overcome with proof of guilt beyond a reasonable doubt in order to institutionally express a strong degree of preference for non-punishment of the guilty to punishment of the innocent. Instead, we do so in order to ensure that institutions of legal punishment have suffi cient scope while preserving their ability to accurately sort the guilty from the innocent.