The Behavioural Layer Above the Combine

A modern scouting stack is deep on the body: combine numbers, force plates, GPS load, tracking data, shot charts, advanced box scores. That layer is mature, and front offices are right to lean on it.

What the stack does not measure is temperament under pressure. How a prospect carries themselves in the final four minutes of a one-possession game. Whether composure recovers fast after a turnover or leaks across the next three plays. Whether the confidence in the pre-draft interview is congruent or performed. Those are the questions that separate two players with identical analytics profiles, and the behavioural layer adds the dimension the body metrics cannot see.

There is no new capture session and no controlled lab. The input is the footage a front office already has: broadcast and all-22 game film, pre-draft and combine interview clips, workout video. We read facial action units from 468 MediaPipe FaceMesh landmarks, head pose, and where the data supports it, body posture and voice. The signal was always in that film. It was never scored.

The Facial Action Coding System (Ekman and Friesen, 1978) has spent four decades establishing that specific muscle movements map to specific behavioural states, and that the coding is reproducible across observers. What changed is speed: a trained coder needs roughly a hundred minutes per minute of video, and the same pass now runs automatically against a prospect's full clip library.

Because it is the existing film, you can run it on the whole board, including prospects you will never get a private workout with.

The Composure Index measures decision readiness under pressure. It tracks how often stress-indicator action units fire across rolling windows and, more tellingly, how fast the face returns to baseline afterward. A prospect who shows a brow furrow for three frames after a bad possession and resets within half a second is a different competitor than one who carries that signature for two seconds while the next play develops. Recovery speed is the read, not the absence of stress.

The Authenticity Signal works the interview tape. It measures emotional congruence: whether the affect on the face matches the moment, or whether you are watching suppression and a practiced answer. It maps where the congruence holds and where it thins, so your interviewers know which threads to pull in the next conversation.

The Momentum Signature reads the clutch profile across a session: whether focus is reached, held, and self-reinforcing through a run, or whether it fragments when the game tightens. Together the three turn 'great motor, questions about the moment' into something you can point at on a timeline.

The quiet advantage is consistency. Human evaluation drifts: the prospect scouted on a Tuesday after a long flight gets a harsher read than the one watched fresh on Saturday. The engine scores the late-round flyer on the same scale as the projected lottery pick, with the same definitions, each clip.

That consistency is auditable. When two prospects land in the same composure band, the room looks at one timeline and the same action-unit evidence rather than two scouts' competing memories. The argument moves from 'I just trust this kid' to 'here is where his recovery speed degraded against ranked opponents.'

Each read carries a confidence level, and the engine abstains when the clip cannot support a score. Thin footage, a face turned from camera, heavy motion blur, a forty-second interview cut: the system returns 'not enough signal' rather than a number built on nothing. A read you can trust is one that tells you when it has nothing to say.

Use it the way you would a strong scout's report: as one input into the room, weighed against medicals, analytics, character interviews, and the eyes of people who have watched the player live. What it adds is the dimension the body-metric stack could never measure, scored on one scale for every name on the board.