The stochastic versus deterministic argument for combat simulations: Tales of when the average won't do

Analysts use combat models to provide information to decisionmakers who must make and justify decisions involving billions of dollars and impacting many lives. A longstanding debate is whether combat simulations should generally be deterministic or stochastic. This article argues that the nature of combat, along with fundamental mathematical principles, implies that most combat models should be stochastic. One decisive reason is that deterministic approximations to stochastic elements almost always generate biases in outcomes, which might foster poor decisionmaking. This fact is illustrated by examples that cover the critical combat elements of attrition, detection, timelines, tracking, data fusion, and queues.