Mislabeled in Football's Data World: When the Machine Is More Confident Than the Editor

On an October afternoon in Marseille, in an editorial room still carrying the smell of printer's ink and cold coffee, a file slid into the processing queue under a familiar label: football. I opened it. Inside was a step-by-step guide on how to file an uncontested mutual-agreement divorce online in Mexico City — forms, electronic signatures, the virtual filing portal of the capital's judicial authority, confirmation timelines. Not a single player's name. Not a single score. Not a single stand. But the machine did not hesitate for a second. It stamped the file "football" and pushed it onward, like a linesman raising his flag for a goal on a ball that never touched the net. Thirty years in this trade have taught me that in football, mistakes rarely come from the weak. They come from the certain. A defender certain he tracked his man. A referee certain he saw it clearly. An algorithm certain it understood. That mislabeled file is a miniature portrait of a disease quietly spreading through sports media: we are handing over the act of understanding to machines that have never learned how to doubt. Football today runs as an information economy. Every match generates thousands of data points: pass counts, distances covered, top speed, expected-goal figures, pressing intensity per defensive action. Newsrooms no longer have enough people to read it all. They build automated pipelines: collect, label, distribute. The label becomes a silent gatekeeper. Once the label is wrong, everything downstream is wrong, and nobody sees it, because the error has been dressed in the clothing of correctness. Based on my own experience watching matches and working with data, I believe we underestimate how dangerous classification errors are. In football, a misplaced pass can be seen with the eye. A misplaced label cannot. It travels quietly through summary tables, automated bulletins, and answer engines, then returns to readers as a fact that looks thoroughly respectable. Picture a search engine's answer machine receiving that file. A user types "online divorce procedure Mexico City." The machine, trusting the football label, hesitates. Then it either discards the item or blends a legal answer with a paragraph of goal statistics. Both outcomes are bad. One buries useful information. The other poisons it. Either way, the reader loses, and trust in the system cracks one more line. The irony is that football prides itself on precision. We measure Kylian Mbappé's speed to the hundredth of a second. We quantify a shot to the last probability. Mbappé's speed exists not to run, but to cut a blade across time, and we record that blade in numbers only to prove it was real. Yet that very obsession with precision creates a blind spot: we inspect everything inside the box, and forget that the box may have been mislabeled from the start. I have seen this at a smaller scale. Years ago, the newsroom where I worked introduced a content recommendation system. An article about a youth team's tactics was tagged "transfers," simply because it mentioned a player who had just renewed his contract. For weeks, that article kept surfacing whenever a hot transfer story broke. Readers clicked, were disappointed, and left. None of us noticed until a loyal reader wrote in asking why the newsroom had deceived them. One wrong label had silently destroyed trust we had spent years building. At the scale of an entire industry, the consequences are far larger. The transfer market is a match with no referee, where every figure is a free kick. A false rumor labeled "confirmed" can send a player's value soaring within hours, force a small club to pay more for a signing, or rob a player's family of its peace. The machine does not know that. It only knows the label. There is a paradox worth pondering here. We build ever more sophisticated data systems to understand football more deeply, yet those very systems make us misunderstand faster. An old journalist like me spends three days verifying a single fact. An automated pipeline can spread a mistake to hundreds of thousands of people in three minutes. Speed does not make the truth faster; it only makes the error travel farther. Data does not score goals, but it knows where the ball is going. I still believe that. But I must add a clause: data only knows where the ball is going when it is read correctly. A number in the wrong place is more dangerous than a number that does not exist, because a missing number tells us to go looking, while a misplaced number convinces us we already have the answer. In football, faith in metrics has been abused to a dangerous degree. Expected goals was created to measure chance quality, yet people use it to judge a manager's decision, to rule on a player's form, to replace the eye. It cannot do those things. It does not see how many nights a player slept before the match. It does not see the fear in a young defender's eyes when he faces a striker faster than himself. And still people use it like a judge. The labeling machine works the same way. It does not see content; it sees patterns. It recognizes keywords, structures, frequencies. When that legal document contained words it had met before in a sports context — filing, submission, record, confirmation — it stitched them into a judgment. But a document is not the sum of its keywords. Neither is a match. A team is eleven people, and more than that, a system of equations that can run. A system of equations, if you read only the variables and ignore how they connect, will give you an answer that is formally correct and spiritually wrong. I remember the matches played without crowds during the pandemic. An empty stadium is a mirror: it does not reflect the spectators, it reflects the loneliness of the game. With no noise, I noticed players communicating more with their eyes, with nods, with pauses. Those signals appear in no statistical table. And I wondered: if a machine read that match through data, would it see the loneliness? It would see an ordinary match, with less noise. Our problem does not lie in the machine being weak. It lies in us granting it too much authority without demanding that it know how to doubt. A good system must be able to say: "I am not sure." So must a good journalist. Over thirty years, I have learned that the most important sentence in this trade is not "I know," but "let me check again." That humility is being traded away for speed. Return to that divorce file once more, but from another angle. It is a symptom, and a symptom always points to a larger disease. If one document is mislabeled, it is quite possible that other documents in the same batch are mislabeled too. Classification errors rarely travel alone; they travel in packs. In a pipeline where everything is connected, a mistake at the input stage can multiply into a mistake at the output stage, a hundred times larger. There is another story I want to tell. In 2026, at the World Cup in Russia, I was sent to Moscow. During France's 4-3 win over Argentina, I did not record the score. I recorded the moment a body changed direction. I described Mbappé's burst in the 64th minute as a knife slicing through the fog of old tactics. My editor complained that the piece had no numbers. But a young coach in Ligue 2 called to ask permission to use it as teaching material. What I learned from that was not that data is useless. What I learned is that data only means something when it serves understanding, rather than replacing it. Had I handed that piece to a machine, it would have written: Mbappé reached a top speed the eye could barely follow, creating a chance. Correct in terms of data. Wrong in terms of the human. That moment was not in the speed figure; it was in a nineteen-year-old player, amid tens of thousands of spectators, deciding that time could be bent. The machine does not know fear. It does not know surprise either. And something that does not know surprise cannot retell a moment. Back to our industry. I am not calling for data to be thrown away. I am calling for data to be put back in its proper place: a tool, not a god. An information pipeline needs human stopping points — checkpoints where an editor can say "hold on." Not to slow everything down, but to keep the truth from being swept away by speed. There is a lesson from football itself. A good defense runs slower than we think, but reads situations faster. It is a defense that knows when to drop, that covers for one another. Our information systems need such a defense too. They need people who know that a label is not a fact, that a keyword is not a meaning, that a number is not a story. I wonder what would happen if we applied football's principles to the way we report on football itself. In football, every goal can be reviewed. Every decision can be questioned. So why are labels, these automated judgments, exempt from that questioning? Why do we demand that referees explain themselves, but not that machines do? This asymmetry is the industry's biggest blind spot. We scrutinize every frame to find an offside, yet we trust a labeling system no one checks. We argue for hours over a penalty, yet accept that an algorithm decides which content readers get to see. In both cases, we are playing a match in which the rules apply to only half the pitch. I think about the young people in the newsroom. They grew up with data, they trust spreadsheets, they trust charts. They are not wrong to trust. But I hope they learn something my generation learned through wounds: every number is born from a human decision — a decision about what to measure, what to ignore, what to name. Data is not neutral. It carries the biases of its maker, even when that maker is a machine programmed by a human. In football, we have witnessed small teams achieve miracles. Saudi Arabia's 2-1 win over Argentina at the 2026 World Cup is one example. Back then, colleagues wrote about an offside mistake. I spent three days interviewing Saudi fans on the streets of Doha, gathering twenty-three stories. I described the high defensive line as a door left unlocked that no one dared walk through. It was called too poetic. But it was those stories that explained what the scoreline could not: why an underestimated team could stand firm against a bigger one. The same holds for how we handle information. A correct label may fail to tell the story. A wrong label is certain to tell the wrong story. What we do goes far beyond reporting; we are sowing fragments of memory. Sow one fragment wrong, and you ruin a whole season. I also think of football's great moments, where collective memory is woven from the smallest details. On the night of the 2026 Champions League final, Milan beat Barcelona 4-0. People remember the score. But what made that match was the rhythm, the way an underestimated team suffocated its opponent with composure. If a machine reads only the score, it will tell you a very different story. And if it reads a wrong label, it will tell you about a match that never existed. At this point, I want to return to the opening question: what happened to that divorce file? It caused no disaster. It was blocked, unlabeled, returned to its proper place. But it left a small crack in my faith in the system. Not because the system was wrong, but because the system did not know it was wrong. A system that does not know it is wrong is a system that will keep being wrong. I am not writing this to accuse anyone. I am writing it as a reminder. In football, we often speak of the moments that break a match's rhythm — a counterattack, a long shot, an unexpected red card. Our media industry is living through such a moment too. The data revolution has broken the old rhythm of the writing trade. What remains is a question: will we let it sweep us away, or will we learn to stand firm and read the match again? If there is one thing I want to leave behind, it is this: keep your doubt. Keep that small, uncomfortable, but essential habit of the journalist — the habit of asking whether this thing belongs where it is. In a newsroom flooded with spreadsheets, such a question sounds outdated. But I have seen it save more articles than any chart. If you are a reader, remember this: every time you read a number about football, ask yourself what question that number was born to answer. Sometimes the answer is a goal. Sometimes the answer is a mistake. And sometimes, the answer is a label applied by a machine that has never once watched a football match all the way through. I still believe in data. I believe it can be poetry, can be rhythm, can be light cast into the dark corners of a match. But I believe more in the human behind the data — the one who knows that behind every number is a story, behind every label a decision, and behind every decision a responsibility. The dream of football never lies in the result, but in the instant before the ball touches the ground. And perhaps the dream of writing about football is the same: it lies in the instant before the writer rushes to trust the first answer. When I left the newsroom that day, Marseille had gone dark. I thought about the divorce file, and I smiled. It reminded me that even in a world run by confident machines, there is always room for someone to sit back, open the file, and say: "Hold on. This belongs somewhere else."