Putting the art before the force

Abstract

We use a dataset from the Battle of Kursk to test three estimators of attrition: linear, quadratic and log dependence on the number of soldiers in each force. Data giving force numbers per day show significant collinearity, so we use force and loss ratios for our tests. We demonstrate that the strongest correlate in the dataset for a sides attrition is its own force strength. This supports the log estimator, and we evaluate the proposition that this counterintuitive connection is a product of the pre-battle art of war, where commanders attempt to balance their forces to their expectations of threat. Thus expected losses generate actual force numbers whereas we seek information on the ways that force numbers generate actual losses, and both processes are based on the same correlation information. We argue that the dataset must still contain information on the mechanisms of attrition, so we widen our search criteria and uncover some remarkable facts. © Austral. Mathematical Soc. 2010.