Law 17 spends four pages arguing that predictability is the condition of animals — the thing that makes creatures huntable, the comfort most people never escape.
Then the Reversal:
"Sometimes predictability can work in your favor: By creating a pattern for people to be familiar and comfortable with, you can lull them to sleep. They have prepared everything according to their preconceived notions about you."1
And it names two distinct uses, which are worth separating because they have different costs.
"First, it sets up a smoke screen, a comfortable front behind which you can carry on deceptive actions. Second, it allows you on rare occasions to do something completely against the pattern, unsettling your opponent so deeply he will fall to the ground without being pushed."2
The smoke screen is continuous. You behave consistently, people stop watching, and the space behind the pattern is where you operate. This is a cover — it is renewable, it costs only the discipline of remaining consistent, and it can be used indefinitely because nothing is spent.
The single break is a one-shot expenditure. You behave consistently for years, and then, once, you do the opposite. This is a deposit and a withdrawal, and when you break the pattern you have spent it. The next time, there is no pattern to break.
Greene lists them in one sentence as though they were the same benefit. They are opposite in the most important respect: the first is an income, the second is a capital asset with one liquidation event.
And you cannot fully have both. Running the smoke screen means the pattern must hold — which is exactly what the single break destroys.
Here is what the Reversal establishes and the chapter never reconciles.
The body text: predictability is animality, a failure of the distinctively human capacity to improvise, something most people accept out of a mistaken belief they will be left alone.3
The Reversal: predictability is a resource that accumulates while you behave consistently and can be spent for a large effect.
Those are not compatible instructions. Under the body text you scramble continuously. Under the Reversal, continuous scrambling means you never accumulate anything, so the powerful move is permanently unavailable to you.
Ali settles it, and he appears immediately after this paragraph. Ten years of one style, spent in a single night. On the body text's account, those ten years were ten years of huntability. On the Reversal's, they were the deposit that made the knockout free.
The chapter argues both and marks neither, which means a reader cannot tell what Law 17 recommends.
The mechanism deserves stating precisely, because the phrase lull them to sleep undersells it.
An observer watching an unpredictable person is attending — every action gets processed, because none of it can be predicted. That attention is expensive for them and it is a cost you are imposing, which is Law 17's whole argument.
But attention cuts both ways. A person watching you closely also sees what you do.
An observer watching a predictable person eventually stops processing individual actions and switches to running the model. They are not watching you; they are watching for deviations from the model, which is a much cheaper operation and a much coarser one. Anything that fits the pattern passes through unexamined.
The smoke screen does not blind the observer — it changes what they are looking at. They are checking for anomalies rather than reading events, and anything conducted inside the shape of your normal behaviour is not an anomaly.
That is why the two halves of Law 17 are genuinely in conflict rather than merely differing in emphasis. Unpredictability buys you an exhausted observer who is watching you very closely. Predictability buys you an unobservant one. You cannot have a target who is simultaneously worn out from trying to read you and no longer bothering to look.
Put that way, the two strategies serve different objectives and the chapter has the material to say so.
Unpredictability is for when you want the other party occupied and unable to act independently. Visconti's court: nobody can plan, nobody challenges, nothing happens. The output is paralysis in the target.
Predictability is for when you want to act unobserved. The output is freedom of movement for you.
Those are different goals. Visconti did not want freedom of movement; he wanted an unchallenged court. Ali did not want Foreman paralysed for ten years; he wanted one night in which Foreman had prepared for the wrong fight.
The deciding question is whether your objective requires the target to be stuck or requires them to be looking elsewhere — and Law 17 has both cases, recommends both techniques, and never poses the question.
You want room to do something without it being examined.
The instinct from the rest of this chapter is to become harder to read. That is the wrong instrument for this job, and understanding why is the whole content here.
Being unreadable makes people watch you more. They cannot model you, so they process every action individually, which means the thing you wanted unexamined gets examined along with everything else. You have bought their exhaustion at the price of their attention.
What you want instead is to be boring in a specific, legible way. Same rhythms, same channels, same kinds of decision, so that people around you switch from reading your actions to checking for deviations. Then conduct the thing you need conducted inside the ordinary shape — same meeting cadence, same reporting line, nothing that registers as a departure.
The discipline this requires is unglamorous and it is the actual cost: you have to be consistent for long enough that people have stopped looking, and there is no way to shortcut the interval. A pattern established last week is not a pattern; it is a recent development, and recent developments are exactly what an anomaly-checker is tuned to catch.
And decide, in advance, which of the two things you are buying. If you spend the pattern on one dramatic break, the smoke screen is gone and everyone starts reading your actions individually again. You can have the cover or the coup. Choosing by accident means you usually get neither.
The strongest thing here is that the Reversal is specific — it names two distinct uses rather than offering a general hedge, which is unusual for this book and lets the two be separated and tested against the chapter's own cases.
Tension: the Reversal contradicts the body text on what predictability is, and the chapter never chooses. Animality versus stored asset; continuous scrambling versus accumulate-and-spend. A reader cannot determine Law 17's recommendation.
Tension: the two named uses are mutually limiting and are presented as complementary. The smoke screen requires the pattern to hold; the dramatic break destroys it.
Tension: the chapter's own two flagship cases sit on opposite sides. Visconti scrambles continuously and buys paralysis. Ali accumulates for a decade and buys one night. Greene presents both approvingly under one law.
🚩 SINGLE SOURCE — the Reversal is Greene's own argument rather than a sourced claim, offered with no case of its own; it borrows Ali, who appears immediately afterwards. [POPULAR SOURCE]
Open questions. Is there a partial version — can a pattern be broken slightly, spending some of the deposit and retaining the cover, or is it a binary asset? And: the smoke screen requires people to stop watching, which takes time. Is there any way to know from the inside when the interval has elapsed, or does the operator find out only by being caught?
This Reversal is the fifth in the build that completes or qualifies its law rather than inverting it, after Laws 7, 14, 15 and 16. The pattern is now firm: some of Greene's Reversals supply the criterion the chapter withheld. This one goes further than most — it does not qualify Law 17, it identifies the resource Law 17's main argument tells you to destroy.
Against Law 16's Reversal, the structural similarity is exact and worth naming. Both say the same thing about a different quantity: the powerful move is a withdrawal from an accumulated stock, and the accumulation phase is the part the body text disparages. Law 16 accumulates presence and spends it as absence. Law 17 accumulates predictability and spends it as a break. In both cases the chapter's headline instruction, taken literally, prevents the accumulation.
Two laws apart, same architecture, and Greene connects neither — which suggests the corpus has a systematic blind spot about phases. Instructions are stated as permanent postures when the cases consistently show them as the second half of a two-stage sequence.
The convergence with Law 3's material on concealment is closer than the chapter admits: a smoke screen is a cover, and covers belong to the concealment family rather than to the unpredictability family. Greene has filed a Law 3 technique inside Law 17's Reversal, where its incompatibility with the surrounding chapter is easy to miss.
Business — Consistent, Persistent, Predictable Actions — The Trust Builder
That page's argument is that trust is constructed from three behaviours: consistent (same stance each time), persistent (over time), predictable (they can forecast your next move). Predictability sounds negative and is a positive, because the other party stops spending energy guessing.
Set it against the smoke screen and the two describe the same behaviour producing the same effect in the observer, valued in opposite directions.
Both agree: consistency causes the other party to stop actively monitoring and start assuming. The sales page calls the result trust. Greene calls it lulled to sleep.
And the mechanism has no way to distinguish them, which is the finding. Whether reduced monitoring is trust or vulnerability depends entirely on what the consistent party does with the space, and that is invisible to the person who has stopped watching — because having stopped watching is precisely the state in question.
That gives a rather bleak reading of trust as an operational matter. Trust is not a judgement that someone is reliable; it is the cessation of the verification that would tell you. The sales page is right that it is necessary — no relationship can afford continuous verification — and Greene is right that it is exploitable, and both are describing one thing. Which means the defence cannot be watch more carefully, since that is the state trust replaces. It has to be periodic re-derivation on a schedule, the same countermeasure this build arrived at from Law 12's prior-installation mechanism.
Psychology — Error Management Theory
That page's logic is that biases are tuned by the asymmetry of error costs — where one kind of mistake is much more expensive than the other, the system is set to make the cheaper error more often, and the resulting bias is adaptive rather than defective.
Apply it to the observer and it explains why the smoke screen works at all.
Watching a person continuously is expensive. Missing something important is expensive. The observer cannot have both, so they settle on a policy — model-and-check-for-deviations — which is cheap and catches the errors that matter given the model.
The insight the pair produces: the smoke screen does not exploit laziness or credulity, it exploits a correctly calibrated policy. The observer is not failing. They have optimised for a threat distribution in which unusual behaviour signals danger and usual behaviour does not — and that is true of nearly everyone nearly always, which is why the policy is right.
Which locates the vulnerability precisely, and it is not fixable by effort. The policy's failure mode is by construction the case where a threat is deliberately shaped to fit the model — and an adversary is the one source of threats that can do that. Error management explains both why the observer's policy is correct and why it is defeated only by opponents, and Greene's smoke screen is simply the name for occupying that gap.
Sharpest implication. Unpredictability buys an exhausted observer who is watching you closely; predictability buys an unobservant one. They are not two intensities of one technique — they are opposite trades, and Law 17 recommends both without ever asking whether your objective requires the target stuck or the target looking elsewhere.
Generative questions.