You come up the escalator at the MRT exit a little after ten, phone in hand, the walk home familiar enough to do with your eyes closed. Fifteen metres away, someone else is doing exactly what you're doing: watching how you move. Not staring. Reading the way you carry yourself down the length of one block, the way you'd glance at a car in a car park to check if a door had been left open.
Eleven years with the Singapore Police Force, then two decades reading exactly this kind of moment as a regional security manager across Asia-Pacific, is where what follows comes from. I've stood on both sides of that fifteen-metre gap enough times to know it isn't instinct or bad luck deciding what happens next. It's a read, done in seconds, against a short list of signals sent whether you know it or not. That background is on the trainers page in full. What matters here is simpler: the read can be changed. Alter the signal, and more often than not, you simply stop being a candidate.
Predators don't choose targets at random, and they don't choose them because of who those people are. The choice happens in seconds, running off a handful of visible, temporary signals in how you move through a space. None of those signals are fixed. All of them can change before your next block. This is what a selector actually reads, and what happens when you read it back.
How do attackers choose their victims?
Selection is the first of four stages in what Jeffrey Fleury calls the Anatomy of a Predatory Encounter: Selection, Testing, Positioning, and Attack or Withdrawal. Everything in this article lives inside Stage 1, because it's the only stage you can end by yourself, before the other person has done anything at all. At Selection, an attacker isn't choosing a person so much as scanning a stream of people and running a rough cost-benefit calculation on each one, weighing resistance against risk against how much time he'd have before anyone noticed. He isn't reading who you are. He's reading what your body is doing in this specific minute, and that read is fast.
How fast isn't folklore. A well-known 1981 study in the Journal of Communication had convicted violent offenders rate strangers as easy or difficult targets using nothing but silent video of the way they walked, with no names and no backstory attached. The offenders still sorted them consistently, and the split traced back to gait, the rhythm and balance of an ordinary walk. Take away everything except how a person moves down a street, and the read still holds. That's the premise this whole article rests on: the read is real, and it runs on behavior, not identity. Behavior is something you get to set.
None of this is an argument that Singapore is dangerous. It's a city where most walks home end exactly as uneventfully as they're supposed to, and that fact doesn't cancel out Stage 1. It just means the calculation runs less often here than it does elsewhere. Rare is not the same as never, and the handful of times in a life this actually matters are exactly the times worth being ready for.
What a selector is actually reading
Gait is one input, not the whole test. On streets built around MRT platforms and void decks, the signals worth knowing run wider than a single lab study from 1981 could capture. Five come up again and again: how you move and at what pace, where your head and attention sit, what your hands are doing with a phone, the choices you make about space and distance, and how you respond in the first two seconds if a stranger tests you with a question or a sudden bump into your space.
Pace is the easiest signal to read from across a car park, and one of the hardest to fake under pressure. Someone walking with a destination already decided moves differently to someone drifting, hesitating at junctions because the bearings are gone. Attention runs on the same logic. A head that stays down and a thumb that keeps moving across a screen tell a watcher that the next stretch of pavement comes with a few free seconds attached, before anyone would notice anything at all.
Space and response work together. Walking close to the kerb or against a wall, past a lit shopfront or through a gap between two vans, tells a watcher how much thought is already going into the next block. And if that watcher tests the read with something small, a question for directions, a stumble into your path, the first two seconds of the reply either confirm the read or break it.
The table below sets out what each of those five signals communicates, and the specific, ordinary change that flips it.
| Signal | What it communicates | The change that flips it |
|---|---|---|
| Gait and pace | Uncertainty about direction, or attention elsewhere entirely | Walk like the destination is already decided |
| Head position and attention | A few free seconds before anyone would notice anything | Head up, eyes moving across the street now and then |
| Phone use | Hands and attention both occupied at once | Call or text before you start walking, or after you've arrived |
| Spatial choices | Predictability: the same gap or shortcut, every time | Choose the lit, open route over the short one, and vary it |
| Response to a test approach | Whether hesitation reads as an opening | A flat, unhurried answer and a step that keeps distance open |
Why passing Selection matters more than anything that comes after
Selection isn't the end of the four stages. It's the gate to the other three. Testing is a small interaction, a question or a stumble, built purely to see how you respond. Positioning follows only if that response reads as compliant: the maneuvering that closes distance or steers things toward somewhere quieter, an empty stairwell instead of a busy platform. Attack or Withdrawal is the stage most people picture when they think about self-defense, and it's also the stage that stops mattering the moment Selection doesn't go a certain way.
That's worth sitting with. None of the training that matters at Stage 4 gets used if Stage 1 ends things first. A step back. A flat answer given without hesitation. None of it is dramatic. It just makes you a worse bet than the next fifty people on the same street, and for someone running a cost-benefit calculation in real time, a worse bet is enough.
Difficult, not dangerous
A selector isn't measuring whether you could win a confrontation. He's measuring whether you'd notice, and whether the next block hands him a few unwatched seconds. Take those seconds away and the calculation changes on its own, well before anything approaches a fight. None of this is about who you are. It's about what your body is doing for the length of one block, and that part is yours to set. That's the whole reversal this article is named for: the same criteria a selector uses to build a shortlist is the checklist you use to take yourself off it.
Common questions
Does this mean someone who's been targeted before did something wrong?
No. A selector's read runs on a few seconds of visible behavior, not on character or worth. Everyone sends these signals sometimes, tired at the end of a long day or distracted by a call. Being read a certain way in one moment says nothing about you as a person. It says something about that moment, and the moment is the part that's changeable.
Can changing how I walk actually make a difference?
It changes what a selector has to work with. That's the whole claim, nothing about guarantees. Fewer easy signals to read is a real shift in the calculation, regardless of how any single encounter would otherwise have gone. The goal was never to make you unreadable. It's to make you a slower read than the person thirty seconds behind you, and that comparison is one you can actually influence.
Isn't this just another way of saying stay on edge all the time?
No, and the difference matters enough that it gets its own answer in the three-minute awareness habit we teach first. The signals in this article live as ordinary habits: walking like the destination is decided, head up rather than fixed on a screen. Neither requires scanning every stranger as a threat. It requires being present in your own body for the length of a walk, which is a smaller ask than it sounds.
What if I already do one of these things, like using my phone on the walk home?
Then that's the one to change first. Not everything at once, just the highest-traffic habit, calling or texting before you start walking or after you've arrived, instead of mid-block. A specific change beats a general resolution to be more careful, because a specific change is the kind you actually keep.
Does answering a stranger's question always mean I'm being tested?
No, and treating every stranger as a test would be exhausting, and wrong far more often than it's right. Most questions on the street are exactly what they look like. What's worth carrying is the response, not the suspicion: a flat, unhurried answer and a step that keeps distance open costs nothing if the question was genuine, and it's the same response that changes the read if it wasn't.
Do these signals matter the same way at night as during the day?
They matter more at night and in quiet stretches, because Selection depends on privacy as much as it depends on you. The same walk through a crowded market at midday and an empty car park at midnight isn't the same test, even if you're sending identical signals in both. Match the effort to the setting: the quieter and more isolated the stretch, the more those signals become the only information anyone has to go on.
None of this replaces training for the moment avoidance genuinely fails, and it was never meant to. What it should do is take the mystery out of how that moment gets decided long before it arrives, so the choice you're making on an ordinary walk home stops being invisible to you. Reading the street back is a skill like any other. It's one part of what I build with students, from the ground up, in the room at JF's programmes.