Feigned retreat

Why can an army that turns and runs away destroy the force chasing it?

A body of troops breaks contact and rides away. The enemy follows. Some distance on, the fugitives turn about and cut the pursuers to pieces. The manoeuvre recurs across very different armies — steppe horse-archers, Seljuk cavalry at Manzikert, and, by William of Poitiers' account, the Normans at Hastings.

The natural reading is that it is an ambush: the runners lead the chasers onto hidden troops. Sometimes there are. But that cannot be the mechanism, because the manoeuvre works with no ambush at all — the same retreating body turns and destroys the same pursuers who outnumbered it a mile earlier. Nothing was added. So what was subtracted?

Start with what makes a formed body of troops dangerous, which is not the sum of the individuals in it. It is the arrangement: men or horses in mutual support, flanks covered by neighbouring units, a reserve held back, and a chain of command through which a single decision reaches everybody at once. Take those away and you still have the same number of men, the same weapons, the same courage — and a mob.

Now ask what a pursuit does to that arrangement. Pursuit is, structurally, a race. The fastest horses and the boldest riders go first; the tired, the cautious and the heavily equipped fall behind. Units that began as distinct bodies intermingle as they overtake one another. Officers can no longer see, and could not be heard if they could. The reserve, seeing a rout, joins in rather than waiting. Within a few minutes the pursuing force has converted itself from a line into a stream — strung out along the line of flight, unequal in speed, mixed in unit, and out of anyone's control.

That conversion is the whole mechanism, and it happens whatever the pursued force does. All the retreating side has to add is the one thing the pursuers have lost: the ability to act on a signal. If the retreating body knows in advance where it will rally and on what cue, it can stop being a stream and become a line again in the time it takes to hear a horn. Then the encounter is no longer between two armies. It is between a formed line and the leading fragment of a crowd — and behind that fragment, arriving in ones and twos, more fragments, each defeated separately as it comes up.

Two further things compound it. The chase is exhausting, so cavalry that has galloped a mile meets cavalry that has trotted; and a force in pursuit believes it has already won, which is a poor state in which to receive a charge.

Then there is the arithmetic of casualties, which in pre-modern battle is the sharpest part of the picture. A body that keeps its shape loses relatively few men, because each is covered and can face the threat. Killing happens overwhelmingly once a formation breaks, when men are struck from behind while fleeing and cannot resist. So the feigned retreat is not a way to win a fight; it is a way to manufacture, on the enemy, the specific condition under which armies are destroyed rather than merely beaten — and to do so while your own force is still ordered.

Which explains, finally, why the manoeuvre is rare rather than universal. It requires the retreating troops to be more disciplined than the pursuers, at the moment when everything looks like a defeat, and to hold that discipline while their own rear ranks are being ridden down. It is far easier to attempt than to survive, and a feigned retreat that fails is simply a rout.

The first refinement is that the ambush and the feigned retreat are different things that often travel together. An ambush works by adding force at a chosen place. A feigned retreat works by subtracting cohesion from the enemy, and any hidden reserve is an amplifier, not the cause. This matters practically: a commander guarding against ambush watches the terrain, while a commander guarding against a feigned retreat watches his own units, and only the second precaution addresses the actual danger.

The second is that the countermeasure follows straight from the mechanism. It is not to refuse the pursuit but to pursue with only part of the force, keeping the rest formed and stationary. That converts a possible disaster into a modest loss — and it is a hard order to give, because it means declining most of the reward at the moment of apparent victory.

The historical claims deserve a caveat. That the tactic existed and was used deliberately is not in doubt for the steppe armies, whose whole style of war was built on it. Hastings is genuinely contested: several historians argue that the first Norman withdrawal was a real break that happened to succeed, with later ones repeated deliberately — and that the chroniclers, writing for the victor, had reason to present an accident as a plan. The mechanism does not depend on that reading, since a genuine flight produces the same disintegration in the pursuers; the deliberate version merely arranges for someone to be ready at the far end.

• Why casualties in pre-gunpowder battles cluster so heavily in the rout, and what that implies about the purpose of pursuit.
• How the same cohesion logic reappears in the walk-through on the defender's advantage, from the opposite side.

Every term the collection defines is gathered in the glossary.