Computerized combat models are known occasionally to exhibit non-monotonic behavior in the sense that adding capabilities to one of the combatants degrades rather than improves the battle outcome for that side. Using a simple combat model, this report shows that for an important class of (nonlinear) combat phenomena--reinforcement decisions based on the state of the battle--such non-monotonicities can be widespread and are related to mathematically chaotic behavior caused by the nonlinearities. Unlike other common sources of non-monotonicity, the latter makes ameliorating the non-monotonicities difficult. This report also demonstrates that larger, more complex models are not immune to these effects. The emerging understanding of these effects has serious implications for the verification of models that are to be used for comparative purposes.
Dewar, James A., James J. Gillogly, and M. L. Juncosa, Non-Monotonicity, Chaos and Combat Models. Santa Monica, CA: RAND Corporation, 1991. https://www.rand.org/pubs/reports/R3995.html. Also available in print form.
Dewar, James A., James J. Gillogly, and M. L. Juncosa, Non-Monotonicity, Chaos and Combat Models, Santa Monica, Calif.: RAND Corporation, R-3995-RC, 1991. As of October 06, 2021: https://www.rand.org/pubs/reports/R3995.html