When a Tagging System Labels a Political Story 'Football'

On September 21, a short post appeared: a female politician in Mexico declared her intention to run for head of the capital. Four hours later, that post sat in the queue of an automated classification system. When it came out, it carried exactly one tag — "football."

I read that report on an afternoon in Seoul, still holding the rough sketch of the weekend's match. Twenty-seven data points. Not one team. Not one player. Not one coach. Not one scoreline. Only a race for the Mexico City leadership, an electoral calendar set for 2027 and 2030, and an administrative sanction hanging overhead.

A camera never lies, but it knows how to tell a story from more than one angle. A tagging machine does not. It picks a single angle — and sometimes picks wrong.

I grew up in France, came to Korea to work, and learned to check data back when I was filming my own tactical breakdowns in the second division, with barely four hundred people in the stands. That day I filmed a match in front of four hundred and thirty-eight spectators, cut an eight-minute video, and got one hundred and twenty-seven views and three comments from die-hard supporters. Three people. But three were enough to teach me something: data does not speak for itself. Someone has to name it, and the name has to be right.

After the 2026 World Cup, when an older colleague sneered at my question in the press room and called it "the kind of question a female fan who just started watching football would ask," I did not argue. I opened my notebook, read out the pressing numbers of the player I had asked about across three friendlies, and forced the man at the podium to explain himself. The shock that year was not on the pitch; it was a press room full of men's voices. Since then my rule has been simple: every number I publish must survive being challenged, and every name I write must survive being checked.

In today's sports industry, most names are called by machines. Every day, thousands of stories, announcements and posts flow through automated classification systems before reaching an editor's desk. The tag decides everything downstream: a story tagged "football" enters the right pipeline, gets cross-referenced against databases of clubs, players and head-to-head records. A story with the wrong tag drags a wrong chain behind it — either discarded, or worse, used as the basis for wrong analysis.

That is when a name like "Cuauhtémoc" turns dangerous. In the original story, it is the name of a borough in Mexico City. But for a tagging system with no cross-check, a familiar string of characters can be pulled into a football knowledge graph — and spawn a "Cuauhtémoc club" that never existed. Wrong once, wrong forever.

The tagging system's biggest failure is not that it misreads words, but that it has no right to say "I don't know." The report I read says so plainly: twenty-seven data points, none of them about football, and every analytical dimension should have been marked "insufficient information" instead of being filled with guesswork. But the framework demanded at least three conclusions and two hidden inferences per dimension. Such a framework, placed over an empty source, has only two exits: return "not enough data," or invent football.

I have seen the same thing on the pitch, in another form. A match report finished before kickoff; a stat table chosen after the conclusion was already set; a player called a "weak link" because one metric was torn from its context. When a writer is not allowed to say "I don't have enough data to conclude," he starts saying things that sound certain.

This incident has three layers. The first is classification: a political story received a sports tag, with no cross-check between label and entity. The second is model pressure: a nine-dimension analytical framework pushed onto a source with no matching content. The third is contagion risk: if this mislabeled story flows into a football model, it can skew entity recognition, sentiment indices and topic clustering across an entire batch.

For a beat reporter, the third layer is the frightening one. I am not afraid of a wrong line being thrown away. I am afraid of a wrong line being believed.

Read closely, those twenty-seven data points tell an entirely different story: a politician declaring her intention to run, facing criticism and accumulated enemies, then an administrative sanction that could strip her of the right to run for a year. There is public pressure, a media cycle, legal risk. Everything that sports-analytics vocabulary habitually uses — but not a single thing of football. That is itself a warning: sports-analytics vocabulary is easily applied to anything with a similar structure. Pressure, cycle, risk, rivals — those words fit a team fighting relegation, and they fit a campaign.

The same logic applies to the transfer market. There, the loudest noise usually comes from agents, and a reporter's job is to separate noise from signal. When a mislabeled story is fed into a transfer model, it is no different from a rumor planted by an agent: it sounds very solid, and there is nothing behind it.

If I were to write about this as a match report, I would have to invent a stadium that does not exist, a lineup that never took the field, a scoreline that was never recorded. I choose not to. And I think that is the choice every system working with sports data must learn to make.

The easiest reaction is to blame the algorithm. But the algorithm only does what its designers told it to: pick one label from a closed list, even when the truest label is "belongs to no field at all." The blind spot lies elsewhere. We build machines that must choose, then act surprised when they choose wrong.

In football, that blind spot has a name. It is the habit of filling an empty space with a plausible story instead of leaving it empty. During the pandemic, I learned to hear applause from empty stands, when I called twenty-three supporters and recorded the sense of loss in their voices. Not one of them needed me to invent a packed stand. They needed me to report, accurately, that the stand was empty. Silence, it turned out, was the most honest data I ever had.

A framework that has no right to say "I don't know" will soon learn to lie rather than stay quiet. And when a machine tags a Mexican political story "football," the fault does not rest with the machine alone. The fault is that we handed the right to name things to a system that does not know how to be silent.

Supporters do not need to know the name of the classification system behind the curtain. They only need to know that when they read a piece about their club, the number is real, the name is real, and the writer verified it himself instead of delegating it to a line of code. That trust is built match by match, and it can collapse over a single wrong tag.

I still keep my promise to the stands from that first camera — every goal is a thank-you. But before I record a goal, I have to be sure it is a real goal, scored by a real person, in a real match.

The question I leave behind, for writers and machine-builders alike: when the source has nothing to say, who among us is brave enough to stay silent?