A great deal of violence in civil wars is informed by the logic of terrorism: violence tends to be used by political actors against civilians in order to shape their political behavior. I focus on indiscriminate violence in the context of civil war: this is a type of violence that selects its victims on the basis of their membership in some group and irrespective of their individual actions. Extensive empirical evidence suggests that indiscriminate violence in civil war is informed by the logic of terrorism. I argue that under certain conditions, that tend to be quite common, such violence is counter productive. I specify these conditions and address the following paradox: why do we sometimes observe instances of indiscriminate violence evenunder conditions that make this strategy counterproductive? I review four possible reasons: truncated data, ignorance, cost, and institutional constraints. I argue that indiscriminate violence emerges because it is much cheaper than its main alternative - selective violence. It is more likely under a steep imbalance of power between the competing actors, and where and when resources and information are low; however, most political actors eventually switch to selective violence. Thus, given a balance of power between competing actors, indiscriminate violence is more likely at early rather than late stages of the conflict. Overall, the paper suggests that even extreme forms of violence are used strategically.