Sun Pin, commander of Chʻi and disciple of Sun-tzu, faces the armies of Wei, which outnumber him two to one.
*"Let us light a hundred thousand fires when our army enters Wei," suggested Sun Pin, "fifty thousand on the next day, and only thirty thousand on the third."*1
*On the third day the Wei general exclaimed, "I knew the men of Chʻi were cowards, and after only three days more than half of them have deserted!" So, leaving behind his slow-moving heavy infantry, the general decided to seize the moment and move swiftly on the Chʻi camp with a lightly armed force. Sun Pin's troops retreated, luring Wei's army into a narrow pass, where they ambushed and destroyed them.*2
⚠ I knew the men of Chʻi were cowards.
The general's own word is knew. ⚠ The fires did not persuade him of anything — they supplied evidence for a proposition he was already holding, and he says so in the moment of acting on it.
Which is why the deception required no skill at concealment. ⚠ Counting campfires is a crude and obvious intelligence method, and Sun Pin did not have to make the deception undetectable. He had to make it confirmatory.
⚠ A lie you would have to be talked into is expensive; a lie that completes a sentence somebody has already started is nearly free. The corpus recorded the same structure at Law 33's look for contrasts and at the thumbscrew: the operator supplies the last piece, not the whole picture.
⚠ One night of thirty thousand fires proves nothing.
A single low count could mean a small army, a cautious one, or a shortage of wood. ⚠ The information is entirely in the sequence — 100k, 50k, 30k — because a rate of decline is a story and a level is not.
And the story it tells is specific. ⚠ Halving, then halving again, over three days, is not attrition or disease — it is desertion, which is the only mechanism that produces losses at that shape.
⚠ So Sun Pin does not fake an army's size. He fakes a trend, and lets the general infer the size, the cause, and the moral condition of the enemy from it. Three inferences, none of them stated, all of them wrong.
⚠ Not believing the fires — the more consequential error is in the next sentence.
Leaving behind his slow-moving heavy infantry, the general decided to seize the moment and move swiftly.
He did not merely advance. ⚠ He restructured his force, discarding the component that would have made an ambush survivable, in order to exploit an opportunity that had a deadline he had invented himself.
⚠ And the deadline was the deception's real payload. A fleeing enemy must be caught before it escapes — so the fires did not just tell him the enemy was weak, they told him he had to hurry, and hurrying is what stripped him of his infantry.
The corpus recorded exactly this at Law 35's forced time: ⚠ the attack is on the schedule, and the schedule attack is what produces the structural error. Sun Pin's fires are a tempo weapon wearing an intelligence weapon's clothes.
⚠ 🚩 SECONDARY WITHOUT PRIMARY · 🚩 SINGLE SOURCE — no date, no battle named, no source, no historian, for an episode that is one of the most famous in Chinese military literature.
🚩 ⚠ The episode is the battle of Maling (c. 342 BCE) in the standard accounts, and the Wei general is Pang Juan — Sun Pin's former fellow student, who had earlier had him mutilated. ⚠ Greene omits the personal history entirely, and it is not incidental: the general's contempt for Sun Pin's army is inseparable from his contempt for Sun Pin, which is precisely the chink in their vanity the law is about. The chapter's own thesis is stronger with the omitted material. [PLAUSIBLE — needs corroboration] on the details as Greene gives them.
🚩 ⚠ Sun Pin, commander of the armies of Chʻi and loyal disciple of Sun-tzu — the relationship is traditionally described as descent rather than discipleship (Sun Bin is held to be a descendant of Sun Wu, writing roughly a century later). [CONTESTED].
🚩 The two-to-one odds, the exact fire counts and the general's quoted exclamation are all unsourced. ⚠ A quoted line from a fourth-century BCE battlefield should be read as literary.
You want an opponent to act prematurely and you are wondering what to show them.
Find the belief they already hold about you. ⚠ Do not construct a deception — construct a confirmation. The Wei general said I knew, and that word is the whole mechanism. A deception aimed at a blank slate has to carry the entire argument; one aimed at a standing prejudice has to carry a single data point.
Fake the trend, not the level. ⚠ A state is ambiguous and a rate is a story. Three readings in a sequence let the target derive the cause themselves — and the corpus recorded at Law 37 that a conclusion the audience generates is one they defend. You are not telling them your army is deserting; you are letting them work it out.
Build the deadline into the picture. ⚠ This is the part that produces the error. A weak enemy invites an attack; a fleeing enemy compels a fast one — and speed is what makes people leave their heavy infantry behind. Ask what your opponent will have to discard in order to arrive in time.
⚠ And know that this is a plan, not a trick. Sun Pin had the pass selected and the ambush prepared before the first fire was lit — the same finding as Selassie's church, bribes and leaflets. The provocation is the last step in both.
Strongest evidence. A three-step deception with a specified gradient, an explicit statement of the target's prior belief in his own words, and a described structural consequence — the abandoned infantry — that follows from the induced urgency rather than from the induced estimate.
Tension — the chapter presents this as fooling an enemy about strength, and the operative effect is on his schedule.
Tension — the omitted personal history between Sun Pin and Pang Juan is the chapter's own chink in their vanity, and it is not mentioned.
🚩 [POPULAR SOURCE] · 🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY — no date, battle, source or historian for a famous episode. 🚩 [CONTESTED] on "loyal disciple of Sun-tzu". ⚠ The quoted exclamation should be read as literary.
Open questions. Would a rising count have worked as well? ⚠ The chapter's logic implies the direction matters — desertion, weakness, urgency — but a rising count would also have produced a decision, namely caution and delay. Which suggests the technique is more general than the case: you are not always making them advance. The corpus has one direction demonstrated and the other is untested here.
⚠ This is the clearest instance in the chapter of the law's offensive half, and it confirms the corpus's finding on the law page: every offensive case in Law 39 involves a provoker holding overwhelming prepared advantage. Sun Pin had the pass; Selassie had the church; Houdini had the skill. ⚠ The provocation is the trigger, never the weapon.
Against Law 35's forced time the identity is exact and unremarked: ⚠ Mehmed held an emissary in place to protect a flank; Sun Pin made a general hurry into a pass. Both are schedule attacks and the corpus can now count at least five across the build, filed under five different laws.
⚠ And against Houdini and Kleppini at Law 30 — which the Keys of this chapter explicitly invoke — the mechanism is the same: reveal an apparent weakness to lure the opponent into a commitment. Houdini let Kleppini steal a combination; Sun Pin let Wei count his fires. ⚠ Two laws apart, and Greene names the connection here — one of the few places in the build where he cross-references his own cases.
Intelligence and deception — the confirmation rather than the lie. Information Asymmetry and Intelligence Dominance treats the control of what an opponent can observe, and through which channels, as the operative variable.
Campfires are an observation channel Sun Pin did not have to compromise — he had only to feed it.
The insight neither produces alone: the deception frame usually concerns inserting false information into a channel, which requires defeating whatever verification exists. ⚠ Sun Pin's case shows the cheaper version: the channel was legitimate, the count was accurate, and every fact the general received was true. He lit exactly thirty thousand fires. ⚠ The falsehood was entirely in an inference the target performed himself — which means there was nothing to detect, because nothing untrue was transmitted. The general's own reasoning was the vulnerability, and it had been armed by a prejudice he arrived with. The general rule: the most robust deception is one in which every observable is genuine and only the conclusion is wrong.
Behavioural science — the prior that does the work. Cognitive Dissonance describes the pressure toward consistency between a held position and incoming evidence, and the reinterpretation of evidence to preserve the position.
I knew the men of Chʻi were cowards.
What the pairing produces: the chapter treats the fires as a clever trick and Sun Pin as the author of the outcome. ⚠ The consistency frame reallocates the credit: the general had a standing commitment to a proposition about the enemy's character, and the fires were simply the first evidence that let him act on it. He was not deceived so much as released. ⚠ Which explains why the deception needed no sophistication — the work of believing was already done, and Sun Pin supplied only the occasion. And it gives the operator's actual search: not what would fool them but what are they already waiting to have confirmed — which is a question about the target's history rather than about your own cleverness, and is exactly what Law 33 is for.
Sharpest implication. ⚠ I knew the men of Chʻi were cowards — the general says knew, and the fires only supplied evidence for a proposition he was already holding. A lie that completes somebody's existing sentence is nearly free. And every observable was genuine: Sun Pin lit exactly thirty thousand fires, ⚠ so the falsehood lived entirely in an inference the target performed himself, and there was nothing to detect. The payload was the deadline: a fleeing enemy must be caught, and hurrying is what made him leave his heavy infantry behind.
Generative questions.