Law 15's Reversal opens by conceding almost nothing and then conceding something real.
"This law should very rarely be ignored, but it does sometimes happen that it is better to let your enemies destroy themselves, if such a thing is possible, than to make them suffer by your hand."1
The military version follows: a good general knows that attacking a cornered army makes its soldiers fight much harder. Better to leave them an escape route. "As they retreat, they wear themselves out, and are ultimately more demoralized by the retreat than by any defeat he might inflict on the battlefield."2
That is a genuine tactical claim and it is not a softening of the law. It is a claim about where destruction is cheapest to produce.
Sit with the second half of that sentence, because it is the part that carries the technique.
A defeat is an event. It happens, it ends, and the survivors have a story about it — we were beaten, here, on this day, by that force. A story is something an army can hold on to.
A retreat is a process with no event in it. Nobody was beaten at any particular moment. The force simply dissolves over weeks of marching, and there is no engagement anywhere in it to organise around afterwards.
Greene's claim is that the second is more destructive to morale than the first, and it costs the pursuer nothing. You are not spending soldiers to produce it. You are spending your enemy's own distance.
Then the qualification, and it is not decorative: "When you have someone on the ropes, then—but only when you are sure they have no chance of recovery—you might let them hang themselves."3
Only when you are sure they have no chance of recovery.
That is the entire law of Law 15 reinstated as a precondition for its own Reversal. The Reversal does not permit stopping short. It permits changing the instrument once the outcome is already determined — you may substitute their exhaustion for your effort, provided the result is not in doubt.
Which is why this is not really a reversal. It is a cost optimisation inside the law, and the chapter's own framing gives it away.
The second half of the Reversal is where the chapter meets its strongest objection, and it is worth quoting the whole movement because the shape of it is the finding.
"Finally, sometimes by crushing an enemy, you embitter them so much that they spend years and years plotting revenge. The Treaty of Versailles had such an effect on the Germans. Some would argue that in the long run it would be better to show some leniency."4
That is the objection, stated fairly and in Greene's own voice, and it is the most serious one available against Law 15.
The answer: "The problem is, your leniency involves another risk—it may embolden the enemy, which still harbors a grudge, but now has some room to operate. It is almost always wiser to crush your enemy. If they plot revenge years later, do not let your guard down, but simply crush them again."5
Take the exchange apart, because it is the clearest instance in this build of the corpus's argumentative structure.
The objection is empirical. It cites a specific historical outcome: total crushing produced decades of accumulating revenge, at a scale that dwarfed the original conflict.
The answer is not. It offers no counter-evidence, no case where leniency produced the predicted emboldening, and no threshold distinguishing the two. It restates the law's premise — half-measures embolden — and then adds simply crush them again.
That last clause is where it collapses. Versailles was the crushing. The objection is that crushing produced the revenge. Answering it with "then crush them again" applies the treatment to the condition the treatment caused, and the reply's grammar — simply — presents the repetition as obvious rather than as the thing in question.
And there is a structural feature worth naming: the recommendation is unfalsifiable in this form. If crushing works, the law is confirmed. If crushing produces revenge, the remedy is more crushing. No observation can count against it, which is a property of the argument rather than of the world.
Note also which position the chapter closes on. The qualification (banishment, not murder) came twenty lines earlier and softened; the Reversal re-hardens and gets the last word.
One clause in the Reversal is doing psychological work rather than strategic work, and it deserves separating: "Let them be the agents of their own destruction. The result will be the same, and you won't feel half as bad."6
Read that against the chapter it sits in. Four pages have argued that pity is fatal and that you must not take an enemy's hatred personally — that the operator's feelings are an obstacle to be removed.
Here the operator's feelings reappear as a benefit to be optimised. Same result, less discomfort.
Which means the chapter has two incompatible positions on the operator's interior life: it is an impediment when it argues for restraint, and a legitimate consideration when it can be satisfied at no strategic cost. Feeling is a bug in one paragraph and a feature four pages later, and nothing in the text marks the switch.
Someone is doing badly and you could finish it.
Not dramatically — you could make one call, forward one thread, say one accurate sentence in a meeting, and the thing that is already happening to them would be complete.
Greene's Reversal says do not, and its reason is not kindness. Work through it.
If you act, the event has an author, and the author is you. Everything that follows attaches to your name: their version of the story, the sympathy that accrues to people things are done to, and the specific fact that everybody watching now knows you finish people. If you do nothing, the same outcome arrives with no author at all, and there is no story for anyone to hold.
Then apply the condition, because it is strict and it is where this goes wrong. Only when you are sure they have no chance of recovery. If the process is not already terminal, waiting is not patience — it is leaving a live opponent alone while they get their footing, which is the failure the rest of the law is about. You have to actually judge whether the thing is finished, and the honest answer is often that you do not know.
And notice the clause about feeling better. It is true and it is worth distrusting. The relief of not having done it yourself is real, it is available whether or not the strategic reasoning holds, and it will happily supply you with a strategic reason for inaction that you would not otherwise have found.
The strongest content here is the retreat-versus-defeat distinction, which is a genuine mechanism and is independently plausible: a process with no event in it gives the losing side nothing to organise around.
Tension: the Reversal does not reverse the law. Its stated precondition — certainty of no recovery — reinstates total victory as the requirement, and only substitutes the instrument. Third instance in this build of a Reversal that completes or qualifies its law rather than inverting it, after Law 7 and Law 14.
Tension: the Versailles exchange is argument-shaped without being an argument. An empirical objection is answered with a restated premise and an instruction to repeat the treatment. Documented, not resolved.
Tension: the operator's feelings switch status mid-chapter — an obstacle when they counsel restraint, a benefit when they cost nothing.
🚩 SINGLE SOURCE · 🚩 CONSENSUS MISREPRESENTATION — "a good general knows" and "some would argue" are both consensus claims with no source, and the Versailles reading is a large historiographical position compressed into one sentence and used as a foil. [POPULAR SOURCE]
Open questions. Does the retreat-demoralisation claim have evidence behind it, or is it received military wisdom repeated? And the one the Versailles exchange raises and cannot answer: if both crushing and leniency are said to produce revenge, what observation would the chapter accept as evidence against its own recommendation?
The Reversal is where Law 15 comes closest to the vault's Siu material and refuses the meeting. Siu's Law of Reversal — "stretch a bow to the very limit, and you'd wish you had stopped in time" — is precisely the Versailles objection generalised, with a mechanism attached (Galbraith's countervailing power) that Greene's version lacks. Greene raises the objection, answers it with repetition, and closes. Siu raises the same objection and derives a limit from it.
The difference is that Siu's position can be wrong and Greene's cannot. Siu predicts something specific: opposition self-organises as dominance approaches completeness. That is checkable. Crush them again predicts nothing.
Within the chapter, the Reversal also sits oddly against Sinigaglia, which is the law's best-executed case and does the exact opposite of what the Reversal recommends: Cesare's first move was taking the bridge, closing the escape route the Reversal says to leave open. Both are presented approvingly. The variable that sorts them — whether you want the enemy exhausted or the enemy finished tonight — is available in the two cases and named in neither.
Psychology — The Amygdala-Aggression Link
That page's finding reorganises how the cornered army reads. Fear and aggression are not opposites; they live in the same structure. The amygdala specialises in danger detection, and danger resolves downstream into either fleeing or fighting. An animal can freeze or lunge at an identical threat. "We mistake an aggressive person for someone unafraid, when the truth is often the reverse: they are terrified, and aggression is their panic response."
That is the mechanism under Greene's tactical claim, and it supplies what he asserts.
The cornered army does not fight harder because it has become braver. It fights harder because removing the escape route removes one of the two available resolutions of the same activation, and the energy goes into the remaining one. The soldiers are not less frightened than the ones retreating. They are identically frightened with one fewer option.
The insight neither generates alone: leaving an escape route is not mercy and not even really strategy — it is preserving the enemy's ability to convert fear into flight rather than into force. Which reframes the whole manoeuvre. You are not being generous with an exit; you are managing where a fixed quantity of arousal goes. And it predicts something Greene does not say and that generalises well past armies: any pressure applied to a person with no available exit converts into aggression by default, regardless of their temperament, because the route it would otherwise have taken is closed. The de-escalation trade knows this; the strategy manual arrives at the same place and calls it demoralisation.
Behavioral mechanics — The De-escalation Formula — Re-orient, Validate, Observe, Action
A woman is screaming at a gate agent, and the opening move is not I understand you're frustrated — it is a strange joke about trains, specific enough that she cannot place you as an employee or a lecture. Re-orient first, then validate, then observe, then act.
Set that beside Greene's escape route and the two are the same operation aimed at opposite ends.
Both begin by declining to meet force with force and instead altering the situation so the other party's arousal has somewhere non-destructive to go. The de-escalation formula opens a cognitive exit — an unexpected frame she can step into. Greene opens a physical one — a road out of the valley.
What the pair produces: the technique is identical and the aim inverts, which means the method carries no ethics of its own. The de-escalator opens the exit so the person recovers. Greene opens it so the retreat does the destruction his own soldiers would otherwise have to. Same first move, same underlying model of arousal, and the entire moral difference lives in whether the exit leads anywhere. That is a genuinely uncomfortable result for anyone who has learned de-escalation as a humane practice — the humaneness is not in the manoeuvre, it is in what is waiting on the other side of the door, and the person walking through it cannot see that from inside.
Sharpest implication. The Reversal's real content is not restraint but authorship. The same outcome arrives either way; what changes is whether it has your name on it, and whether the losing side is left with an event they can organise around. That is available without any of the chapter's ruthlessness, and it is the part worth keeping.
Generative questions.