Mislabeled Feeds in Sports Media: The Lesson of a Story With No Football in It

2:47 in the morning in Hamburg. The screen in my study was still on, and a wire line came through from San Luis Potosí, in central Mexico. I read it slowly. Forty-four years of reading night wires have taught me that any word you skip is a word you may later pay for. The line contained "first-year student", "Faculty of Engineering", "Zona Universitaria Poniente campus", near building T, and a short sentence about the university suspending academic activities. Above the line, in the field reserved for the section label, sat one tag: football. I sat still for a long while. Not because the tag made me curious. I sat still because I knew that somewhere in the state of San Luis Potosí a family had just taken a phone call nobody wants to take. And in some newsroom, a labelling engine had filed that phone call under sport. Both things happened on the same night, and only one of them is real. Football never lies; only the viewer lies to himself. This time the one lying was an editorial workflow. I called the night editor. She was brief: the wire came from an aggregator, the system tags automatically by geographic entity, and someone confirmed the tag with one click, or confirmed nothing at all. Either path ends the same way: a story about a student's death placed beside transfer news, scorelines and betting odds. I am not writing about that death as a sporting event. I am writing about the tag. The incident, as far as I have it, took place at the Universidad Autónoma de San Luis Potosí, UASLP. The victim was a first-year engineering student, male, roughly eighteen or nineteen. The report does not name him. Rector Alejandro Zermeño Guerra has spoken publicly about the case. The only authority empowered to determine the cause of death officially is the state prosecutor's office, the Fiscalía General del Estado, FGE. As of my reading, no cause had been published, and the rector declined to say whether the death was accidental or a suicide. The student's father was present. A group of UASLP students publicly demanded stronger mental health provision from the university. That is everything I know. I add nothing, I infer nothing, and I certainly do not reconstruct a moment at which I was not present. So why does a story like this carry a football tag? The answer lies in how news systems have been built over the last fifteen years. Most modern newsrooms no longer let humans assign a section to every wire item. They run an automatic labelling pipeline: raw copy from agencies passes through an entity extractor that identifies people, organisations and place names, then checks them against a label dictionary. That dictionary is usually built around whatever monetises best: football, transfers, scores, betting. And this is where the string "San Luis Potosí" detonates. That string is the name of a state, the name of a city, and the name of a professional football club in Mexico's top division. A tagger based on string matching cannot tell those three entities apart. It sees "San Luis Potosí", it sees a university with teams in the university league system, it sees the word "Engineering", and it applies the tag. This is the most basic failure in natural language processing: conflating entities that share a name. The industry calls it the entity resolution problem. Newsrooms call it "the system doing its thing". In practice, people call it what happens every Friday. I want to describe two occasions when I watched a correct tag being earned, so the distance becomes visible. On 20 May 2026, the final matchday of the 2026/17 Bundesliga season, I sat in the commentary box at the Volksparkstadion in front of 57,000 people. Hamburg SV hosted VfL Wolfsburg. Hamburg won 2-1, the winner came in the 88th minute, and the club finished the season on 38 points, exactly two above Wolfsburg, who were relegated. The coverage of that match deserved its tags: football, Bundesliga, survival. No machine assigned that tag. Ninety minutes assigned it, plus one shot in the 88th minute, plus twenty-two years of my sitting long enough in this city to understand what people cry about. On 2 July 2026, at Rostov Arena, I watched Japan lead Belgium 2-0 through Genki Haraguchi on 48 minutes and Takashi Inui on 52. Jan Vertonghen pulled one back with an overhead kick on 69. Marouane Fellaini headed the equaliser on 74. In the 90th minute plus four, from a corner in Japan's half, Belgium broke and Nacer Chadli scored to make it 3-2 after exactly fourteen seconds. Fourteen seconds can open a life, or bury a legend. That tag was also written by a human, after I stood motionless for two minutes with nothing to say. Now compare the football tag applied to the wire from San Luis Potosí. It took less than a millisecond, required nobody to stand still, required nobody to cry. And it is still in the database, possibly forever. My experience of watching matches teaches one simple thing: a tag is a promise. When I wrote "survival" on Hamburg against Wolfsburg, I promised the reader they would read about the life and death of a club inside ninety minutes. When the machine wrote "football" on the UASLP wire, it promised the reader football. That promise breaks by the third line. And the consequences do not stop at a broken promise. Football pages run on betting money, sponsor money and bookmaker money. When a story about the death of an eighteen-year-old sits inside a football feed, the advertising system serves it like any other football page. A father searching for news about his son may see a betting ad in the right-hand column. Nobody intended that. The system simply did what it was programmed to do. That story will also remain in the football index. Three years later, an editor searching football data for San Luis Potosí will meet it. Ten years later, a larger retrieval system will treat it as training material. A death becomes noise data in a football model. It sounds dry, but I regard it as the most serious form of memory violation this trade produces in silence, because nobody is accountable for a caption that has vanished from the page. In many digital newsrooms, nobody reads it back. People trust the tag. A correct tag means correct trust. A wrong tag spreads structurally: the section page pushes it into the morning digest, the digest pushes it into an email, the email pushes it into the weekly roundup. Nobody in that chain reads the whole line, and nobody has to, because the tag has already answered. I once sat in a German newsroom when a system like that went live. They told me it would save forty percent of an editor's time. That was true. What they did not say was that the forty percent was taken from precisely the work that used to keep section labels honest: reading, and judging. At the sports desks I have worked with, both at home and in Europe, the checking editor's role thinned out very quickly over about a decade. People called it streamlining. I call it transferring risk to the reader. A stadium with no singing is where I hear the truth most clearly. In May 2026, after sixty-three days of Bundesliga suspension, I was assigned Hamburg SV against Holstein Kiel in front of an empty Volksparkstadion. Hamburg lost 1-2. I sat alone in the box, hearing studs on grass and players breathing. I left at 11 p.m., went home, and did not sleep for three days. I tell that story for one specific reason. In an empty stadium, no singing can cover a misplaced pass. Every error stands bare. Our labelling pipeline works the other way: the more noise, the more wires, the more tags, the easier a wrong tag slips through. It slips through amid transfer noise, among thousands of daily items, under speed pressure nobody dares to slow down. There is a technical detail worth naming, because it explains why this class of error does not disappear on its own. Tagging systems are judged on two opposing measures: coverage and precision. To avoid missing football, you lower the matching threshold, accepting more non-football items. To avoid admitting foreign items, you raise it, accepting that you will miss some football. In a section that earns money from traffic, nobody chooses to miss. The balance tips toward coverage, and the price of that tilt is always paid by the category of story nobody wants mislabelled: deaths, violence, accidents. There is another issue, less discussed: most content management systems still run on a single primary label. One story, one section. Reality does not work that way. A single incident can belong simultaneously to education, public health and law. A single-label model forces the system to choose, and it chooses by its own criteria. When there is one box to fill, the engine fills the box with the highest weight, not the box that is most true. Name collisions are also far more common than people assume. Across Spain and Latin America, countless universities, hospitals and public bodies share names with local sports clubs. A naive label dictionary will keep tagging sports onto stories about city councils, general hospitals and secondary schools. Those errors cause no harm, so nobody fixes them. They sit quietly in the system until a story arrives whose consequences are unacceptable. That is why I do not accept "the system did it" as an explanation. A system that errs thousands of times while nobody repairs it is not a system. It is a decision that was made and that nobody owns. The blind spot here is not that a machine mislabelled something. Every tagging system mislabels, including the best ones, and anyone who has worked the trade long enough knows it. The blind spot is the assumption that a clean taxonomy protects us. It does not. The harm happens at the moment two things are placed side by side, not at the moment the tag is applied. A death placed beside a betting price causes harm whether the tag was applied by a person or a machine. So a project to improve dictionary precision, however worthwhile, never reaches the root. The root is that we turned editorial judgment into a measurable metric. Once judgment becomes a metric, it gets optimised. Once optimised, it chases volume. Football becomes the magnet: it pulls stories in, it pulls pageviews, it pulls advertising, and it pulls in things that do not belong to it. The football tag on the UASLP wire is the inevitable by-product of a pipeline designed never to miss football. There is a deeper layer still. Our collective memory of events like this is shaped by the frame journalism places on them. When a death is archived inside a football frame, that frame becomes part of the incident's identity. It does not remove the family's grief. It blurs the question the UASLP student community is raising: what was missing in a university's mental health provision. That question belongs to education and public health. It is buried under a tag nobody will consult when researching student mental health. Singing does not make victory, but it makes memory. Here the reverse happened: the noise of the section made a false memory. And a false memory outlasts a false report, because a false report can be corrected, while a false memory sits in the index. I propose one simple rule for any sports newsroom: no content about the death of a human being may run inside sports advertising inventory, whatever tag it carries. That rule needs no algorithm. It needs one accountable person to press a button. I propose one more thing, about how we read. When you see a story about a death sitting on a football page, I hope the reader remembers that no cause has been concluded, that the state prosecutor's office remains the only authority empowered to publish one, and that behind the wire there is a father and a student community demanding something very specific about mental health. None of that fits inside any taxonomy, including a perfect one. The people of Hamburg did not die on the night of 20 May, and I wept on air because of it. Tonight, an eighteen-year-old student in San Luis Potosí has died, and none of us in this trade is permitted to turn that death into a tag.