The little vest under the training shirt with a pod between the shoulder blades is now standard equipment from the Premier League down to reasonably serious semi-professional football. It records position, speed, acceleration, and with the integrated accelerometer, a whole family of derived metrics.
The technology is mature and reliable. The interpretation is a mess, and I don't think that's widely understood outside of sports science departments.
What's actually being measured
Two separate sensor systems, usually.
A GNSS receiver — GPS plus other satellite constellations — giving position at typically 10 or 18 Hz. From position you derive distance, speed, and acceleration. Position accuracy outdoors under good conditions is on the order of a metre or so, which is fine for total distance and less fine for anything involving short movements.
And an inertial measurement unit — accelerometer, gyroscope, magnetometer — sampling much faster, often 100 Hz or more. This picks up impacts, changes of direction, jumps, and the general mechanical loading that satellites can't see.
Everything on the dashboard is derived from those two streams.
The metrics and their problems
Total distance. The most reliable number and the least informative. A midfielder covering 11.5 kilometres tells you almost nothing about whether he had a good match. Distance correlates with position, tactical role, and match state far more than with performance.
High-speed running distance. More useful, but definitionally unstable. What counts as high speed? Different providers use different thresholds — 19.8 km/h, 21 km/h, 25 km/h for "sprint" — and some use individualised thresholds based on the player's own maximum. A player's high-speed distance can change by 30% depending on which convention you use, with no change in what he actually did.
Accelerations and decelerations. Counting the number of efforts above a threshold. Very sensitive to the smoothing applied to the position data, and satellite-derived acceleration is noisy at the timescales that matter. Two systems on the same player in the same session will disagree, sometimes substantially.
Player load or similar composite scores. Proprietary formulas combining accelerometer output into a single number. Genuinely useful for tracking a single player over time on a single system. Completely meaningless for comparing across players or across providers, because the formula is different and often undisclosed.
The comparability problem
This is the thing that most undermines the field. There is no standardisation.
Different hardware, different sampling rates, different filtering algorithms, different thresholds, different derived metrics with the same names. A club that changes supplier loses the ability to compare against its own historical data in any rigorous way.
Published research inherits the problem. A study finding a relationship between high-speed running and injury risk used one system with one threshold. Whether it applies to your setup is genuinely unclear, and this caveat rarely survives into the practitioner conversation.
There have been efforts at standardisation, and some progress on validation protocols. But the commercial incentives point the other way — a proprietary composite metric is a lock-in mechanism.
What it's genuinely good for
Having complained at length, the technology is valuable when used for the right questions.
Within-player trend monitoring. Same player, same system, over weeks. Is his output declining? Has his acceleration profile changed? This is where the data is most trustworthy, because all the systematic errors are constant and cancel out.
Session design. Knowing that a particular drill produces a certain load lets a coach construct a week with a deliberate distribution rather than a guess. That's a real improvement over the previous method, which was largely intuition.
Return to play. Progressing a player back from injury by matching training loads to a target profile derived from his own pre-injury data. Probably the single best-evidenced application.
Catching outliers. A session that produced twice the expected load, or a player whose numbers are wildly out of line with the group. The value is in prompting a question, not answering one.
The framing error
The mistake I see repeatedly is treating these numbers as performance measures rather than exposure measures.
The data tells you what a body did. It does not tell you whether what the body did was any good. A winger who covered less high-speed distance than usual might have had a poor match, or might have been positionally excellent and therefore not had to sprint to recover.
Conflating the two produces some genuinely bad decisions — players being criticised for low distance numbers in matches where they were tactically flawless, and praised for high numbers earned by being out of position.
The sports science staff generally understand this perfectly well. The problem arises when the numbers leak into the coaching conversation without the context, which happens constantly.
What I'd want next
Honestly, standardisation more than any new capability. An agreed set of definitions and validation protocols would make every existing dataset more useful overnight, at no hardware cost.
Beyond that, integration with tracking data so that physical output can be interpreted positionally — not just how far he ran, but where and in relation to what. That's technically feasible now and it's where the interesting questions are. The current situation, where we can measure a footballer's body with great precision and understand almost nothing about the football, feels like a solvable problem that nobody's prioritised.