How we measure your swing
No black box. This page explains exactly what the engine measures, where the target numbers come from, how a red flag gets decided — and what a single phone camera honestly can't see.
What we measure
We track 33 points on your body — shoulders, elbows, wrists, hips, knees, ankles, head — in every frame of your video, using Google's BlazePose model. From those points the engine finds the nine standard swing positions teaching pros use, P1 (address) through P9 (finish), and at each position it measures five checkpoints:
Spine tilt. Your forward bend, measured from vertical. The iron target at address is about 35°.
Hip turn. Signed degrees relative to YOUR OWN address line — 0° is square, negative is turned away from the target, positive is open toward it.
Shoulder turn. Same convention as hip turn. Measuring against your own address line means the number never depends on how the camera was aimed.
Knee flex. The average bend of both knees. Around 25° at address for an iron.
Head drift. How far your head has moved off your hip centerline, as a percentage of your shoulder width — a body-relative unit, so it works at any distance from the camera.
On top of the nine checkpoints, the engine reads continuous curves — shoulder and hip rotation across the whole swing, weight shift, tempo — so a fault between positions doesn't slip through.
Where the target bands come from
Every “target band” in your report is a tour-average number, not our opinion. The reference set is built on tour-average motion-capture biomechanics — the SwingTRU Motion Study reference set — combined with a tour-average rotation chart for iron swings filmed face-on, checked position by position by our founder.
A few of the anchors, so you can see the grain of it: at the top of the backswing (P4) the tour average is about 89° of shoulder turn and 48° of hip turn away from the target. At impact (P7) the hips lead — roughly 42° open — while the shoulders trail at about 30° open. Shoulders MORE open than hips at impact is an over-the-top signature, not tour impact geometry, and the engine treats it that way.
Targets are club-aware: a driver swing is judged against a bigger pivot (about 12% more shoulder and hip turn through the backswing) and is allowed a slight drift off the ball that would be flagged as sway with an iron. And because rotation is measured against your own address line, the targets never punish your natural setup.
How a red flag gets decided
For each checkpoint we take the distance between your measurement and the tour-average target. Each target carries two tolerances — a yellow band and a wider red band, scaled to how much tour players naturally vary on that move:
The measurement is inside the yellow tolerance for that position. Example: shoulder turn at the top (P4) has a target of −89° with a ±12° yellow band, so anything from −77° to −101° is green.
Past the yellow tolerance but inside the red one. For P4 shoulder turn that red band is ±25°.
Past the red tolerance. This is what the doctrine sorts first — the earliest red in the P1→P9 chain becomes your one thing.
Each position's score is the average of its five checkpoints (green counts 100, yellow 60, red 20). One honesty rule matters more than any formula: a fault is only flagged when the checkpoint actually measures that fault. We removed several plausible-sounding fault detectors (casting, scooping, deceleration) because the 2D proxy metrics behind them couldn't truly see them — a confidently wrong diagnosis is worse than silence.
What one camera can — and can't — see
Kinetic works from a single phone camera, so every angle is estimated from a 2D image — a consistent yardstick for YOUR progress against the same references every week, not lab-grade 3D motion capture. We'd rather tell you that here than let a decimal point imply precision we don't have.
Per-metric confidence flags. Every number is checked against the visibility of the exact body points used to compute it. If a hip is occluded at the top or your ankles are out of frame, that metric — and only that metric — is tagged “low camera confidence” in your report, meaning treat it as an estimate, not a measurement.
Rotation is a 2D projection. True 3D turn can differ from what a single camera sees, which is why we compare you against references measured the same way (face-on iron data) instead of against 3D lab numbers.
No club face, no ball data from video. A camera on your body can't measure face angle or spin. If you have a launch monitor (Garmin R10, TrackMan, Golfzon and more), Kinetic imports those sessions and uses the real numbers instead of guessing.
The order we fix things in
Measurement is only half the method. The other half is the teaching doctrine our founder built across roughly 8,000 real lessons, and it decides what your report leads with:
- 01
Weight shift first. If the low point and weight shift are wrong, nothing else matters yet — a sway outranks a louder-looking fault every time.
- 02
The chain, P1 → P9. The swing is nine linked positions; we always fix the earliest broken link first, because a break at P2 can be the real cause of what you see at P6.
- 03
Two thoughts max. You leave every report with one priority and one feel — the way a real lesson ends.
See it on your own swing
Three free analyses every month, no card required. Every number in your report is measured exactly the way this page describes.