When the Football News Machine Mistook a Mexico City Arrest for the Beautiful Game

Night in Acoxpa, in the Tlalpan borough of southern Mexico City. Two men — one aged 30, one 40 — were subdued after a scuffle with traffic police. A white vehicle, a request for documents that was refused, a few soft-tissue injuries, an ERUM medical team called to the scene. That report belonged on the public-safety page, among short briefs on public order. But inside the data system, it carried a single label, tidy and decisive: football. I sat for a long time with that label. In thirteen years on the job, I have grown used to football being misread — read through prejudice, through emotion, through cheap hot takes that I myself have sometimes authored. But this was the first time I saw football misread by a machine that has no idea what football is. The real story isn't in Mexico City. It's in Marseille, in Hanoi, wherever a newsroom runs on algorithms. Modern sports news is no longer written entirely by human hands. Millions of items every day flow through automated tagging systems: entity recognition, sentiment analysis, topic clustering. A machine only needs to catch a few keywords — "match", "team", "card" — to drop an item into the football bin. The Tlalpan incident has no football club. No player, no competition, no transfer, no tactics, not a minute of the ball rolling. It has only one word that can be misread: "card" — the card traffic police demanded, not the yellow card on a pitch. That's all. One word. One machine. And one report lost among millions of lines of data every day. What's more frightening is that this error is not isolated. When I went back through the entire analytical file, the systemic-risk item was rated High — not because of the arrest, but because of the process behind it. An item containing no football entity whatsoever — no club, no player, no competition — still slipped into the football dataset. That means the input filter is broken. And when the input filter is broken, everything downstream is contaminated. Picture what happens next. A sentiment model reads the Tlalpan report, sees the words "brawl", "subdued", "police", and records: football is showing a rising trend of violence. A trend-tracking model clusters topics and concludes: Mexican fans are furious. Those false signals flow into reports, into charts, into editorial decisions in a newsroom ten thousand kilometres from the scene. No one verifies, because no one imagines a traffic arrest could carry a football label. The deeper problem is in the sourcing. In the original report, almost every information point reads "source: none". Only ERUM — Mexico City's emergency rescue squad — is named, and only to confirm officers suffered soft-tissue injuries. That's the mark of a report copied from a police log, not an original piece of journalism. Weak sourcing plus automated tagging makes a perfect compound for data contamination. Football needs no enemy. It poisons itself through carelessness. Over thirteen years I have watched sports news move from being written by eye to being written by machine. Speed went up, accuracy went down. People no longer ask "is this story true" but "does this story have enough keywords". A machine cannot tell an opinion from a fact. It learned that "France won because of an easy draw" is a football sentence, and it spits that sentence out wherever the word "champion" appears. It does not understand that a title never comes from the fixture list, but people need an excuse to hate the strong. To the machine, it's all just word probability. And this is the part I don't want to write. Modern football didn't kill improvisation; it just locked improvisation inside a tactical cage. But the news machine locks football inside another cage — the cage of keywords. Inside that cage, an arrest in Tlalpan and a World Cup final carry the same weight. Both are data. Both are worth gathering into the same bin. Hold on. Blaming the machine is the easiest escape, and I don't want to take that road. The machine only does exactly what we taught it: prioritise volume, speed, and keywords enough to make a headline. If a report about two men in Tlalpan can slip into the football bin, the problem isn't that the machine misreads. The problem is that we stopped reading. I remember the night I sat writing about an empty stadium, when Ligue 1 returned after the pandemic. Applause in an empty stadium is the echo of fear, not of joy. We grew used to football without spectators, then grew used to football without real readers. Both are the same empty room: football in an empty room. I may be wrong. Maybe this is just a small technical error, one misapplied label, a few seconds in a vast pipeline, and no one cares. But I have learned one thing in thirteen years: small errors in large systems do not disappear on their own. They accumulate. They repeat. And then they become the standard. The night in Tlalpan will be erased from the dataset. The report will be untagged, the machine retrained, and a new filter will be born — one demanding at least a club, a player, a competition before any line passes through. But there will be another report, in another city, slipping through another gap. The final question isn't whether the machine knows what football is. The question is: does anyone among us still remember?