Une étude d'histoire militaire instruite par la théorie des jeux et quelques amplifications méthododologiques

The paper suggests that military history should be subjected to modelling based on the mathematical theories of rational choice, and especially the theory of games. This suggestion primarily concerns military events that have already been reported in narratives focusing on the agents' instrumental rationality, such as Clausewitz's narrative of the Waterloo campaign. By selecting relevant facts and hypotheses relative to this campaign and Napoléon's final collapse, we will investigate the decision he made on 17 June 1815 to detach part of his army against the Prussians he had defeated on the day before at Ligny. Military historians disagree on the rational sense of this crucial decision, but game theory is in a position both to illuminate it and decide between their interpretations. The model of this paper is a zero-sum, two-person game between Blücher and Napoléon and Grouchy, who are treated here as a single player. The calculated solution makes it clear that it could have been a cautious strategy on the Emperor's part to divide his army as he did, a conclusion which goes some way towards clearing him from the charges of factual mistake and even irrationality that have been laid against him by a majority of commentators since Clausewitz. Once this result is obtained, the paper moves up to the reflective and methodological plane. It compares the main objections that can be raised against its formalization with those already levelled against the recent, so-called analytic narrative school. On this score, we conclude that military campaigns - especially, though not exclusively, from the 19th century - provide a better opportunity for successful application of rational choice theories that the political examples thus far favoured by the North American school. We generalize the argument by investigating the conflict between mathematical models of rational choice and narratives - that is, the historians' canonical mode of expression. As the paper eventually suggests, the two may be reconciled. This conclusion is argued on the basis of Clausewitz's methodological precedent and an analysis of the concepts of a narrative (in general), a historical narative and an event.