It is night in the Apennine foothills, some time in 217 BC, and you are a Roman soldier holding a pass called Allifae. You have one job: this is the eastern door out of the Campanian plain, and the Carthaginian army is caged inside it.
Then the mountain lights up.
Thousands of torches are climbing toward you, spread across the slopes in a front far wider than anything Hannibal is supposed to have. And they are bellowing — a sound with nothing human in it, coming out of the dark, in every direction at once.
Greene's account is deadpan about what happens next:
Afraid that it would climb above them and surround them, the Romans fled from their garrison, abandoning the pass, too scared even to look behind them. And a few hours later, Hannibal's army came through, escaping from Fabius's cordon.1
The explanation arrives only in the Interpretation, and it is absurdly cheap:
At Allifae, Hannibal had bundles of kindling tied to the horns of oxen, then lit them and sent the terrified, bellowing animals up the slopes to the pass at night—creating an indecipherable image to the Roman sentries, literally in the dark, and a terrifying one.1
Cattle. Kindling. A match. An entire army walks out of a trap.
The setup matters, because Allifae is not a lucky escape — it is the fourth move in a sequence where every previous move had already trained the Romans to expect the improbable.
Rome declared war on Carthage in 219 BC and prepared to send an army to Spain. Before it could sail, the news came that Hannibal was already in northern Italy, having crossed the worst part of the Alps. Nobody garrisons a door nobody believes is a door. His march south was unimpeded.1
He had roughly 26,000 men left after the crossing. Rome and its allies could field close to 750,000.1 That ratio is the constraint every subsequent decision runs under.
At the Trebia he baited the consul Sempronius by having light cavalry cross the river and retreat, cross and retreat, until impatience did the rest. Sempronius brought his whole army across freezing water, which took hours and exhausted them, and then met elephants ridden by archers on one flank — animals the Gallic tribesmen in Roman service had never seen — while 2,000 Carthaginians hidden in riverside vegetation hit the rear.1
Legions then blocked the accessible Apennine passes. Hannibal crossed at the least accessible one, through four days of marsh no army had ever attempted.1 Then Lake Trasimene. Then Rome, in near-panic, appointed a dictator.
Fabius Maximus is the only Roman in this story who understood the problem. His answer was to stop playing: post the legions in mountainous country where Hannibal's cavalry was useless, harass, deny supplies, isolate an army very far from home, and never accept battle.1 It was correct, and it was politically fatal.
many Romans saw Fabius's strategy as disgraceful and unmanly. Worse, as Hannibal continued to raid the countryside, he hit none of Fabius's many properties, making it seem as if the two were in cahoots. Fabius became more and more unpopular.1
Sit with the economics of that. Hannibal declined to burn a set of farms, and the forbearance cost him nothing and bought him the removal of the one Roman commander whose strategy was working.
Greene names it precisely: an unorthodox use of politics and extraordinary means in war.1 The weapon was Roman suspicion, and Hannibal did not have to manufacture it — he only had to leave a shape in the burn pattern for it to attach to.
Cornered into acting or losing power, Fabius built the cage. Hannibal was in a fertile plain in Campania. Roman armies took every exit, each close enough to support the next.
And Fabius did the intelligent thing — he pattern-matched:
Hannibal had entered Campania through the eastern mountain pass of Allifae, and Fabius had noticed that he never left by the same route he entered. Although Fabius kept a sufficiently large Roman garrison at Allifae just in case, he reinforced the other passes in greater numbers.1
This is the hinge of the whole case, and it deserves to be stated plainly: Fabius was not being stupid. He was being an excellent analyst, and that is precisely what was used against him. He had a real behavioural regularity, drawn from real observation, and he weighted his forces by it. Hannibal's counter was to be regular in exactly one respect — never repeating himself — and then to break that.
Hannibal spent the intervening weeks reinforcing the read: cavalry sent north as if probing a breakout, the richest farms plundered to bait a battle. Fabius saw through the baiting and congratulated himself on it. Anyway, he knew Hannibal would try to break through to the east.1 He knew. He just also believed Hannibal wouldn't use the same door twice, so he thinned it.
Then the oxen came up the eastern pass, and by the following year Fabius was out of power.1
Terentius Varro took command burning to avenge Allifae, and met the Carthaginians on open ground near Cannae.
Greene's rendering of Varro's confidence is the best sentence in the chapter:
the terrain was clear, the enemy was in full view, there could be no hidden armies or last-minute tricks—and the Romans outnumbered the Carthaginians by two to one.1
Every clause is true. That is the point.
At Cannae, where the Romans were by this time expecting the unorthodox, Hannibal disguised his stratagem in broad daylight, lining up his army like any other army of the period.1
The centre gave ground because it was built to. The Romans pushed into it, crowded together, and the wings closed. Greene: it was a slaughter.1 The manoeuvre itself is held in full at Cannae: The Perfect Envelopment; what belongs here is only the reason it was available — Varro had been trained by Allifae to search for the trick, and searching for tricks is what makes an ordinary battle line invisible.
Greene's own compression:
The line of least expectation is the line of least resistance; people cannot defend themselves against what they cannot foresee.1
And the second-order effect, which is the thing actually worth stealing:
With less resistance in your path, the progress you make will inflate their impression of your power; Hannibal's small army seemed to the Romans much larger than it really was.1
Twenty-six thousand men read as an irresistible force because nothing they did could be blocked in advance. And once the reputation exists, it does the work by itself:
knowing to anticipate the unexpected is not the same thing as knowing what the unexpected will be. Before long your opponents will give way to your reputation alone.1
Someone competent is watching you, and they have found a pattern. That is the situation this page is about — not being unknown, but being known well.
Start by finding what they think they know. It will almost always be a behavioural rule rather than a plan: she never accepts the first number. He always escalates on Fridays. They never re-use a channel. Fabius's rule was one sentence long and it was correct.
Then check what that rule causes them to spend less on. This is the whole trick. A read is only exploitable where it has moved resources, and Fabius's read moved troops off the eastern pass. If their belief about you hasn't changed how they've allocated anything, it isn't yet a weakness — it's just an opinion.
Now reinforce the read cheaply for as long as you can afford to. Hannibal spent weeks looking north. Nothing you do in this stretch has to succeed; it only has to be consistent with what they already believe.
Then take the door they thinned, and take it with the least expensive theatre that will do the job. Notice how little the oxen cost. The temptation is to reach for something impressive; what you actually want is something indecipherable — an input the other side cannot resolve into a known category fast enough to act on. Sentries in the dark do not need to be beaten, only prevented from classifying what they are looking at.
And then, immediately, expect them to over-correct. The Romans came out of Allifae hunting for tricks, which is exactly the state in which a plain frontal engagement becomes lethal. Your next move after a spectacular one should almost always be a boring one.
The campaign narrative — Alps, Trebia, Trasimene, Fabius's delaying strategy, Cannae — is standard and well-attested. The oxen ruse at Allifae is reported by Livy and Polybius and is among the more famous stratagems of the Second Punic War. [POPULAR SOURCE]
The chapter's exemplar of unconventional genius lost his war. Greene concedes it in a single paragraph — no reinforcements from Carthage, Rome absorbing defeat after defeat, the war dragging on for years1 — and then draws the lesson from the tactics as though the outcome were a detail. This is the book-wide retrospection problem arriving inverted: usually Greene reads backward from victory and resolves luck into design; here he reads backward from a defeat and extracts design anyway, by narrowing the frame to the battles.
The unresolved question that leaves is a real one and the chapter never asks it: if line-of-least-expectation warfare wins every engagement and still loses the war, what exactly is it good for? The honest answer available from Greene's own evidence is that it substitutes for mass and cannot replace it — 26,000 men can be made to feel like an irresistible force, and feeling like one does not resupply you.
The Fabius manoeuvre is the page's real moral fault-line, and Greene files it as cleverness. Sparing one man's estates in order to have him destroyed by his own side's suspicion is not a tactic against an army; it is the deliberate destruction of a competent opponent through the credulity of his colleagues. It worked, and its working removed the only Roman strategy that was actually containing Hannibal. Greene's word for it is unorthodox.
Against Hannibal's Tactical Principles and Strategic Asymmetry this page is the mechanism the other page's principles run on. That page holds the asymmetry doctrine as doctrine; the Allifae case shows what it costs to sustain — a fresh invention per engagement, forever, because each one burns on use.
Against Hannibal in Italy: Consolidation and Stalemate there is a genuine friction worth keeping open. That page holds the strategic picture — an army in Italy that could not convert tactical superiority into political outcome. Greene's chapter holds the same years as a masterclass. Neither account is wrong and they do not fit together, and the gap between them is the most interesting thing either produces: the same fifteen years read as sustained brilliance and as sustained futility, with no disagreement about a single fact.
Against Hannibal and Indirect Motivation, the Gisgo joke before Cannae and the oxen at Allifae are the same man managing two different audiences with the same instinct — give people an input they cannot process in the register they expected. One breaks an enemy's classification; the other breaks his own men's fear.
Psychology → The competent analyst as the preferred target. Fabius lost the pass because he had built a correct model of Hannibal and allocated by it. The vault holds the general form of this in Strategy 13: The Intelligence Strategy — read your opponent, and act on the read.
The insight the pair produces that neither gives alone: a model is an attack surface exactly in proportion to how much it has moved. Someone with no read on you cannot be exploited through it; someone with a vague read has nothing to lever; someone with a sharp, correct, resource-moving read has handed you a map of where their forces aren't. Which inverts the standard security intuition. The dangerous opponent is not the one who has misjudged you — misjudgement produces random behaviour, which is hard to exploit. It is the one who has judged you correctly, because correct judgement is what produces predictable, targetable allocation. Fabius's error was not analysis. It was having only one rule and letting it carry weight it could not bear alone.
Creative practice → The cost of an effect is not its force. The oxen cost a herd, some kindling and a night. The effect was an army's escape from an encirclement. The vault's Constraint-Based Creativity holds the same relationship from the maker's side: limits do not degrade the work, they force the search into regions abundance never explores.
Together they produce a claim neither makes: the effect is generated by the receiver's interpretive machinery, not by the input, so the budget that matters is not what you spend but what the receiver has to spend to resolve it. A sentry with no category for "burning cattle at night" does the expensive work of imagining an army; Hannibal only paid for the torches. This is also why abundance so reliably produces flat work — a large budget lets you supply the effect directly, and an effect that arrives fully supplied leaves the receiver nothing to do.
Being read correctly is more dangerous than being read wrongly, because a correct read is the only kind that reliably moves someone's resources — and the thinned door is always where the accurate belief about you is pointing.