What actually counts as a close pass?

Every rider has a story about the one that made them wobble. Almost none of them start with a number. They start with a noise, a shadow at the shoulder, and the specific full-body jolt of realising how little room there was. That gap between how a close pass feels and how it has to be described is the whole difficulty.

The legal definition, where one exists

In much of the United States, the answer is a fixed lateral distance written into state code: most commonly three feet, with several states requiring four, and a handful requiring more at higher speeds. In the UK, the Highway Code was revised in January 2022 to ask drivers for at least 1.5 metres when overtaking a cyclist at speeds up to 30 mph, and more room above that. Australia and much of the EU use a metre-and-a-half-ish figure with similar speed conditions.

These numbers are not really claims about physics. They are administrative thresholds: a line that is simple enough to teach, testify to, and enforce. Which is a genuine virtue. It also means the threshold is the floor of acceptable behaviour, not a description of what is safe.

Why distance alone undersells the danger

Two passes at an identical 90 cm are not equally dangerous. The variables that actually decide how bad a pass is:

• Speed differential.Being passed at 15 mph over your speed is a different event from being passed at 45 mph over, at the same clearance. Energy scales with the square of velocity, and so does the time you have to react.
• Vehicle size and shape. A box van produces a pressure wave and a wake that a hatchback does not. Trailers and towed loads track wider than the vehicle in front of them, which is why a pass that clears the cab can still catch the rider.
• What the road gave you. Ninety centimetres with a smooth, wide shoulder is survivable. The same gap next to a drain grate, gravel, or a kerb removes the escape route that made it survivable.
• Duration. A long vehicle at close range holds you in the danger zone for several seconds rather than a fraction of one.

A rider integrates all of this instantly and involuntarily. A statute integrates none of it.

The measurement problem

Say you want to report a pass. You now need a number, and the honest position is that eyewitness distance estimates, your own included, are not reliable. Under stress, from a moving bike, with the vehicle in peripheral vision for under a second, people routinely misjudge lateral distance by a wide margin, in both directions. Riders under-report as often as they over-report.

This is the gap a camera closes, but only partly. Raw footage shows that something happened; it does not carry a distance. Getting from pixels to centimetres needs the camera's field of view, its mounting geometry, and a reference of known size in frame. That is the work our pipeline does, and it is why we report a measured clearance with an error band rather than a single confident number. An honest 0.8 m ± 0.15 is far more useful in a conversation with an officer than a precise-sounding figure nobody can defend.

The working definition we use

For flagging purposes we treat a pass as a close pass when the measured lateral clearance falls below the statutory minimum for the rider's jurisdiction, and we escalate it when the speed differential is high, the vehicle is large, or the road offered no escape. The threshold triggers the flag. The context decides whether it is worth a rider's time to escalate.

A pass that clears the legal line by ten centimetres and still frightened you is not a false alarm and it is not you being dramatic. It is the threshold doing what thresholds do. Log it anyway. The pattern across a season of riding is the part that eventually persuades somebody.