Mislabeled Data and the Silent Flaw in Youth Football Scouting

Autumn 2026, at AS Roma's Trigoria training centre, I held a twenty-two-page scouting report in my hands. The sixteen-year-old in the file was rated B-plus potential, ran the thirty metres in 4.02 seconds, and ranked in the top fifteen per cent of his age group for line-breaking passes. Six years later, he plays in Italy's fourth tier. The reason was not in his feet. It was in a single mislabeled data field: his primary position was recorded as central midfielder, when thirty-eight of his forty-one matches were actually played on the right wing. The entire forecasting model was built on a skewed field from the very start. Nobody in the meeting room checked it, because the number looked reasonable, and reasonableness is the most dangerous accomplice in this trade.

Professional football today runs on data as much as on the eye. A single Serie A match generates hundreds of thousands of raw data points: each player's coordinates every tenth of a second, touches, distance covered, heart rate through GPS vests, and hundreds of events logged ball by ball. Major providers such as StatsBomb, Opta and Wyscout resell that data to clubs, to betting companies, and to youth academies that want to believe they are judging objectively.

But data does not generate itself. It travels a long chain: cameras, motion-recognition algorithms, then human hands labelling every pass, every position, every type of event. The error is not at the head of the chain, where everything is clean and quality-controlled. The error sits in the final mesh, where a tired analyst, under time pressure, assigns a label to a passage of play he barely glimpsed. A single wrong label is nothing. A wrong label repeated every week across two seasons becomes a false fact, confirmed by the system itself.

For senior players, that error is washed away by volume. Three hundred matches, ten thousand actions, and the algorithm corrects itself through the crowd. For a fifteen- or sixteen-year-old, the data is thin as paper: sometimes just four matches on a shaky recording, passed from one assistant to another, with nobody asking where it came from.

I once received a file in which the same player had three different birth dates on three pages. That boy, if you are not patient enough to cross-check, was two entirely different people. In analysis circles we call it a domain-mismatch error: a document framed into the wrong place, then treated as if it always belonged there. When an entire academy leans on files passed around like merchandise, the mistake is no longer an exception. It becomes a habit.

To see where the error lives, split a scouting report into layers stacked on one another, like sediment in an excavation pit.

The top layer is the position label. It sounds simple, but position in youth football is not fixed. A left winger can play as a central midfielder in a back three, then become a full-back in a back four. When an analyst labels him a central midfielder, the entire physical programme behind him changes: short-sprint volume rises, explosive work falls, and his defensive range is pushed into midfield. A wrong label does not merely describe a player incorrectly. It creates a new player, and that new player is sometimes far worse than the real boy on the pitch.

The middle layer is the event label, where every pass, shot and duel is categorised. This is the zone where the smallest errors do the greatest damage, because every forecasting model learns from it. I once spent an entire session reviewing footage of a Coppa Italia match against Virtus Entella and found that Nicolò Zaniolo landed on his left foot at a sixty-two per cent asymmetry, loading his right knee unevenly. It was a technical sentence, but I filed it under medical notes. The report lay still in a drawer. Fourteen months later, he tore his anterior cruciate ligament in a turning movement. The data was there. What was wrong was the label I put on it.

The bottom layer, and the most neglected, is the context label. A player scores seven goals in youth football, but those seven goals come against three weak sides, behind defences that have never faced pressure. If the system records the number seven without recording who the opponent was, the reader sees a sharp striker. If the system records the context, the reader sees a question mark. A wrong label, repeated long enough, becomes a truth nobody dares to challenge.

The frightening part is that all three layers are usually produced by the same person, on the same evening, under the same pressure to file before dawn. When an academy leans on that chain, errors compound exponentially. One mislabeled report is an accident. A thousand identical mislabels, repeated across many academies, become a scouting system that teaches an entire generation of young players that they should play where they do not fit.

Meanwhile, the signals that truly predict the future sit outside every database. In 2026, when stadiums closed, I built a small programme for fifteen Roma academy players: fifteen minutes of skipping rope a day, resistance bands in the living room, a messaging group to keep the rhythm. An eighteen-year-old midfielder named Edoardo Bove raised his maximal endurance by twelve per cent and was promoted to the first team for the Europa League match against Young Boys. No data provider recorded it, because a living room is not a pitch, and a skipping rope is not an event. Yet that was the answer to why an undervalued player rose.

That same year, at the European Under-21 Championship, Italy lost to Portugal on penalties. I did not spend two hours logging the scoreline. I spent two hours logging eighteen line-breaking passes by Sandro Tonali, who kept dropping deep to pull opposing centre-backs out of position, opening space for team-mates to run into. The result was a forty-page report on the space between the lines. No data provider marked that space. Because space, in the end, has no owner. To measure it, you have to sit down, rewind and rewind again, and place the label by hand, something no algorithm can do for you.

The paradox is this: clubs are buying more data than ever while checking it less than ever. The market pays for volume, not for cleanliness. A data package of five hundred metrics sounds more attractive than a process in which someone sits and cross-checks every label. Nobody signs a sponsorship deal to boast that they re-verified the birth date of a fifteen-year-old boy.

The result is a reverse spiral: the more data there is, the more confident clubs become, and misplaced confidence is worse than doubt. The old scout's instinct was to sit back and ask where this number came from. That instinct is being replaced by a single click on a metrics leaderboard.

There is a deeper layer few mention. Mislabeling does not only produce wrong decisions; it produces institutionalised bias. When an academy assesses a whole age group with the same model, a skewed field teaches that entire group to play to one template. And the next generation will believe that template is the nature of football, rather than a mistake repeated long enough to look like truth.

Sediment does not lie; it only deceives those not patient enough to dig. The next race among academies is not who collects the most data, but who dares return to check the labels nobody has bothered to look at for years. A living room becomes a gym, because talent waits for no one to make its bed, and it also waits for no one to pin the right label on it.

If tomorrow you had to choose between a model of ten thousand metrics and an evening spent rewinding footage, which would you choose. The next player deserves to know your answer.