Despite having gained widespread acceptance, the decline of war thesis rests on a surprisingly thin evidentiary basis. The data that have been marshaled in its favor often do not reflect quantities of theoretical interest, a fact that can produce very misleading results. The near-absence of formal statistical tests makes it impossible to distinguish between real changes in conflict behavior and random noise—a problem that is especially acute when dealing with substantially non-Normal data distributions. I analyze the system-wide rate of international conflict initiation and demonstrate that, contra the decline of war theorists, no systematic decrease in countries’ propensity to use war against one another is in evidence prior to the end of the Cold War. I also outline a hypothesis that the complexities of international order provide a much better explanation for variation in the data than a decrease in humanity’s propensity for violence. I conclude with some thoughts on the implications for peace.