Are legendary generals just lucky?

John Ohno
John Ohno
Aug 16, 2017 · 3 min read

There’s an infrastructure around legendary generals (and other figures charged with the strategies, tactics, logistics, and general management of war) — not just around teaching their techniques to new officers, but around applying cultural importance to them. We have statues of generals, universities aimed at producing effective military commanding officers, and until pretty recently history has largely been the story of the decisions of military officers rather than the story of bottom-up organization by civilians. But, it occurs to me that for all the effort we put into military colleges, the generals that get the most cultural reverence either didn’t attend them or did poorly. Are we focusing on underdog stories? Are we elevating particular generals in the popular consciousness for reasons unrelated to their effectiveness? Or, is performance in a school designed to produce effective military officers largely unrelated to the effectiveness of military officers?

There’s plenty of reason to suspect the last case, because there’s plenty of reason to suspect that military command falls into the category of something one can be bad at but not good at. Lots of things are like this, but I’d like to focus on one that I think has a lot of similairities: stock picking.

Like war, stock picking is an extremely complex adversarial system. Certainly, in both, one can pick strategies that are much worse than random: you can have all your soldiers shot for cowardice before their first sortie (losing by default) or spend all your money on the single highest-priced stock and sell as soon as its price drops precipitously. Yet, we’ve found that a completely random strategy in stock picking does better than professionals. Is the same true of war?

There are plenty of similarities. Both are a little like ‘quantum suicide’: a good recipe for creating the impression that some people are much better than others at something is to start with a large pool and drop everybody after their first big mistake; in a situation where outcomes are totally random, someone will rarely but reliably win everything or almost everything by chance, and that person will never be able to know whether or not skill played a part. We can’t estimate well whether or not somebody’s successes were fully attributed to chance in such a situation (although we would expect that some were, and we could guess by the distribution of successes). It’s difficult to determine, in that circumstance, the difference between a very difficult problem that only very few people can reliably solve and an impossible problem that nobody can reliably solve (though with a large enough pool we can find the difference between the actual success rate and the expected rate for a fully random one), and when a system is adversarial we have a good excuse for ‘skilled’ people ‘losing their edge’.

There’s a pretty big difference, however. We can test empirically how well a stock picker does: it just takes a big pool of money, and while money is rare, it’s not as rare as cannon fodder. Nobody’s willing to put a random number generator in charge of an army, even if we had pretty good evidence to suggest that a random number generator would reliably perform better than the best human general.

I’m neither a statistician nor a military historian, so I’m not really qualified to crunch the numbers on this. However, it’s a question that I’ve never seen asked, let alone answered. And there’s some import to it: if skill at war isn’t possible, our already-unreliable mechanisms for ranking military leaders are completely pointless; expensive systems for officer training would benefit from being overhauled for less of a focus on tactics and strategy and more of a focus on keeping up morale; war becomes even more of a numbers game with respect to resources; we have no more excuse to have statues of generals in our squares (and we should probably replace them with statues of common soldiers, who, in the absence of value in top-down strategy, at least contributed their lives).

John Ohno

Written by

John Ohno

Resident hypertext crank. Author of Big and Small Computing: Trajectories for the Future of Software. http://www.lord-enki.net