Wargaming

How can a war fought on a map teach real lessons while quietly hiding others?

Between the wars, the US Naval War College at Newport played the Pacific war over and over on a linoleum floor, with wooden counters and rulebooks. Afterwards Admiral Nimitz said that nothing which happened in that ocean had come as a surprise — absolutely nothing, he added, except the kamikaze.

Read that sentence twice, because it contains both halves of the puzzle. A game played on a floor genuinely anticipated a war fought across a third of the planet. And the same game was completely silent about one of the campaign's defining tactics. What kind of instrument is accurate and blind in that particular combination?

Begin with the obvious observation and then push on it. A wargame is a model, and a model is a set of choices about what to represent. Someone decided that fuel matters, so fuel is in the rules. Someone decided that a scout aircraft has a search radius, so search radius is in the rules. The game can then teach you things about fuel and search that you did not know before you sat down — real things, because the relationships between the represented quantities are worked out honestly by play rather than assumed by the designer.

So far so good. Now ask the question the other way round. What happens to something the designer did not represent?

It does not appear as a gap. It does not appear at all. That is the whole difficulty. An omitted factor produces no error message, no anomaly, no awkward result — the game simply runs smoothly, as though the factor did not exist, and the players' confidence rises exactly as it would if the model were complete. Absence is invisible in a way that a wrong number is not: a bad fuel-consumption figure eventually produces a strange outcome someone argues about, but nothing at all argues on behalf of a rule that was never written.

Sit with what that implies about the Newport games. They represented sortie rates, fuel, steaming distances, gunnery, the tyranny of the Pacific's dimensions — and by playing those honestly for two decades, the fleet learned things about carrier operations and underway replenishment that survived contact with reality. They did not represent a pilot deliberately not coming back, because in 1935 nobody had put that in a rulebook. The silence was not a mistake in the model. It was the boundary of the model, and boundaries do not announce themselves from inside.

There is a second effect worth noticing, and it is about people rather than rules. What do players optimise for? Whatever the game scores. If the rules reward concentrated fleet action and say nothing about commerce raiding, then over hundreds of games the officer corps gets very good at fleet action and never develops an instinct for the other thing. The game shapes the players, so a representation choice made once by a designer becomes, twenty years later, a professional habit that nobody remembers choosing.

And what happens when a game does start to break its own assumptions? Here the honest case is Millennium Challenge 2002, the large US joint exercise in which the retired general commanding the opposing force used low-technology methods — motorcycle couriers, small boats, a pre-emptive salvo — and inflicted heavy notional losses on the blue fleet in the opening move. The exercise was then reset and constrained, which its critics called scripting and its defenders called protecting the training objectives that the rest of the schedule existed to test. Both readings can be true at once, and the episode is disputed to this day. What it illustrates cleanly is the tension: a game exists to test something, and a result that destroys the test also destroys the schedule.

The correct conclusion is not that wargames mislead. It is that they are trustworthy about their own contents and silent everywhere else, and the practical skill is knowing which is which. That is why serious game designers keep the assumption list as a document in its own right, and why the most useful question after a game is not who won but which rule decided it.

Two further refinements. First, a game's value often lies less in its outcome than in what it forces its players to specify — writing a rule for something means arguing about how it works, and the argument frequently teaches more than the play. Second, one adjudicated game is a single sample, not evidence: run it once and you have learned what happened that afternoon, which is why analytically serious use involves repetition, varied assumptions, and deliberately hostile red teams.

The misconception worth naming directly is the idea that a more detailed game is a more truthful one. Detail deepens fidelity inside the represented space without widening it at all. A game with a thousand rules about gunnery and none about morale is more precise and no less blind, and the precision makes the blindness harder to see.

• Red teaming, and the structural problem of paying someone to break the exercise everyone else needs to finish.
• How the same blindness appears in economic and epidemic models, where the unrepresented factor is behaviour.
• Whether a game's real product is its result or the assumption list its designers were forced to write.

Every term the collection defines is gathered in the glossary.