Moral decision-making is always marred by uncertainty; in the ethics of killing, that uncertainty is especially acute and pervasive. The stakes couldn't be higher, and yet it is often difficult, even impossible to know whether the conditions for killing to be fact-relative permissible are satisfied. Any successful theory of the ethics of killing should offer an account of when killing is evidence-relative permissible. And yet few do. In this paper, I consider three possibilities: two that focus on thresholds (killing is evidence-relative permissible just in case it is sufficiently likely to be fact-relative permissible), and one that fits the ethics of killing into a broader decision-theoretic framework. I show the flaws of the first two approaches, and endorse the third, before exploring some of its implications and answering some obvious objections.