This paper investigates the multivariate Polya distribution (MPD) for application in combat models for which enemy targets must be randomly assigned to fighting units. Six sets of actual combat data motivate the allocation model. Scenarios include air-to-air combat, tank, and submarine warfare. Goodness of fit tests are derived which verify the validity of the MPD, and reinforce the notion that multinomial allocation is too simplistic for such combat models. Simulations to determine optimal allocations for minimal combat cost are applied in order to illustrate the role of the MPD in combat modeling. © 2001 John Wiley & Sons,Inc. Naval Research Logistics 48: 1–17, 2001