‘Structural Variance’ or ‘Non-Monotonicity’ effects in combat models: A review

‘Structural variance’ is a phenomenon common to many deterministic combat models, in which output measures of effectiveness may appear as irregular, non-monotonic functions of the values of selected input variables. It can, of course, complicate any decision-making process based on model results. This paper is a review of this topic: its nature; the experimental evidence that has accrued to date; and the ‘remedies’ that have been suggested to ameliorate its effects. The causes of this phenomenon appear to lie mainly in the non-linear features incorporated in most modern models in order to represent the military decision-making process and the allocation of resources. It is suggested that many of these features do in fact feature in historical battles. The experimental evidence suggests that the incorporation of random features within the model, which themselves are typically a feature of real combat, can do much to ameliorate the effects of this phenomenon. The penalty lies in the increased variability of results, and so this puts a premium on experimental design and modelling strategies that will help reveal unambiguously the effects of conditions subject to modelling test. Potential candidate strategies are therefore discussed. Most of the knowledge concerning this topic is unsystematic and anecdotal, and so the paper outlines steps which would help to put it on a more rigorous footing.