Although the past few years have seen significant, empirically-based activity in modelling First Person Shooter (FPS) game traffic, there has been only limited work published that explores the underlying reasons for observed FPS traffic characteristics. As a result the generality and scalability of these empirical models is open to question. In this paper we demonstrate how a useful predictive model for FPS traffic patterns can be derived from a number of generic multiplayer game-play requirements. We begin by describing a possible theoretical basis for modelling the message rates and message sizes between server and clients in FPS games. Based on this model we make predictions as to the characteristics of FPS game traffic and then compare the predictions with traffic statistics collected from Quake 3 and Unreal Tournament game trials. Agreement between the predictions from our theoretical model and the statistics from the game trials is very good.