What type of revolutions are most vulnerable to counterrevolutions? I argue that violent revolutions are less likely than nonviolent ones to be reversed because they produce regimes with strong and loyal armies that are able to defeat counterrevolutionary threats. I leverage an original dataset of counterrevolutions from 1900 to 2015, which allows us for the first time to document counterrevolutionary emergence and success worldwide. These data reveal that revolutions involving more violence are less at risk of counterrevolution and that this relationship exists primarily because violence lowers the likelihood of counterrevolutionary success—but not counterrevolutionary emergence. I demonstrate mechanisms by comparing Cuba’s nonviolent 1933 uprising (which succumbed to a counterrevolution) and its 1959 revolutionary insurgency (which defeated multiple counterrevolutions). Though nonviolence may be superior to violence when it comes to toppling autocrats, it is less effective in bringing about lasting change and guaranteeing that these autocrats never return.