When the Machine Misnames: A Football-Tagged News Item and the Lesson for Sports Data

One October morning, I sat before a screen in my small apartment in Incheon, opening the aggregated sports-news dataset the engineering team had sent overnight. Among hundreds of rows tagged football, one made me stop for a long time. It mentioned a large square in Mexico City, metal fences erected around a government residence, a committee named after people who died in 2026, and a female president. No club. No player. No scoreline. Not a single minute of stoppage time. Yet the label at the head of the row said, clearly: football.

I sat still. Thirty-four years in this profession have taught me that most errors in our trade do not come from malice, but from the laziness of a system designed to run faster than the humans inside it. A wrong label makes no one bleed. But it is a symptom of a larger disease: we are building machines that classify sports content without anyone checking whether they truly understand the sport they serve.

This is not a story about Mexico. It is a story about how the sports industry, and especially women's football, is being misnamed at industrial scale.

To understand how a political news item can fall into the football basket, you have to understand how sports content operates today. Most news you read does not travel straight from reporter to reader. It passes through a pipeline: collection, parsing, tagging, scoring, distribution. At each stage, an algorithm decides where the content belongs. And at the tagging stage, the algorithm does not read the way a person does; it counts words.

That is where shared vocabulary becomes a trap. The fences in that story were police barriers, but in the football dictionary, a fence evokes a defensive line. A sealed residence evokes a low block. A blocked area evokes the penalty box. A few keywords landing in the right slots, and the machine nods and tags.

I have seen the same thing in my own field. Years ago, an article about labour law in the entertainment industry was tagged as sports simply because it contained the word league. A press release about a car brand landed in the motorsport section simply because it contained the word track. Such errors are harmless until they flow into products people trust: automated standings, prediction models, aggregated feeds sent to partner newsrooms.

Women's football sits in the most dangerous position in that pipeline. Why? Because data on women's football is already thin. When sources are scarce, the system is forced to loosen its criteria to fill quota. When criteria loosen, errors rise. When errors rise, trust in women's football falls. That spiral feeds itself.

I remember the summer of 2026. I was invited to commentate at a U-20 women's tournament in France, thanks to a series of data-driven analyses. In one group-stage match, I mispronounced a striker's name three times in the first half. Social media exploded. I was so ashamed that I spent the following two weeks reviewing every piece of footage, learning the pronunciation of more than three hundred players at the tournament.

From that stumble, I built a note system: pronunciation, shirt numbers, biographies, hometowns, and the private sorrows of each young woman. Not because I wanted to become a dictionary. Because I understood that calling a person by her correct name is the minimum act of respect. And if I do not forgive myself for misnaming a player, I cannot forgive a machine for mislabelling an entire sport.

Let us talk about the mechanics of error, because error does not arise by itself. It is the product of three forces at once.

The first force is the economics of volume. In the content industry, volume is king. A publishing system is judged by articles per day, impressions, sessions. No one pays for a machine that publishes slowly but accurately. So whenever it must choose between verification and speed, the system chooses speed. A political news item tagged as football reduces no one's KPI; it is one row among thousands.

The second force is shared vocabulary. Sport and politics share a frightening lexicon: strategy, confrontation, defence, attack, fence, campaign, extra time. Alone, each word is harmless. Together, they are enough for an algorithm to believe it is reading a sports story.

The third force is the absence of expert human review. As newsrooms cut staff, they replace editors with filters. But a filter does not know that women's football has a different passing rhythm, that a female midfielder can play as a right-leaning number eight and still be the centre of the play. A filter only counts keywords.

These three forces explain why misclassification is not rare but systemic. And they explain why women's football is a permanent victim: few writers, few checkers, few fixers.

Surveys of sports media over more than three decades, from the United States to Europe, consistently show that women's sport receives only around four to ten percent of total sports news coverage. I cite that figure not to complain but to point out that when you have only one-tenth of the space, you cannot afford another one percent of error. Every mislabel on women's football costs ten times more than on men's football.

I have spent years proving that the fragility of women's football data is not due to a lack of talent, but a lack of attention. In 2026, when I began a WK League analysis programme on a digital platform, I tracked the data of Incheon Hyundai Steel Red Angels for six months. I found a substitute striker named Choi Yu-ri. She played only 214 minutes all season, but her expected goals figure was 3.2, the highest in the squad.

That number appeared in no news report, because no one had the patience to count the minutes of a player on the bench.

I wrote a piece on her potential. A month later, she was given a starting spot in the final match, scored twice, and won the title. I do not tell this story to boast. I tell it to prove one thing: when we take the trouble to parse the right data, we find a person. When we mislabel, we erase a person.

In the summer of 2026, in Tokyo, the South Korean women's national team under coach Colin Bell took only one point from three matches. Looking at the results, everyone concluded: total failure. But I dug into the data and found a telling number: the team's PPDA was 8.6, the lowest in the tournament. That means they pressed harder than any other team, despite far weaker fitness and resources than the European sides.

PPDA, put simply, is the number of passes an opponent is allowed before your team makes a defensive action. The lower the number, the higher the pressure. A team judged physically weak that records a PPDA of 8.6 has chosen to die standing rather than lie down. And no league table records that choice.

I wrote a piece arguing that this failure had meaning. It reached two hundred and fifty thousand reads, the highest of my career. Since then, I have used concepts like pressing and PPDA as metaphors for structural injustice: a weaker team can still play more bravely, and that deserves to be named correctly.

Now put the two stories together. A machine misnames an entire sport. A person correctly names a substitute striker and changes her season. That contrast is the whole problem.

There is a field I always see running parallel to this story: esports. I have written that professionalisation is turning players into assembly-line products, that individual play is being sanded smooth in digital training. The same is happening to sports journalism. When every writer is pushed into the same content line, personality is sanded smooth, and error becomes the standard.

I also think about VAR. I have often said that excessively long VAR reviews are shredding the rhythm of the game, that two minutes of waiting is enough to cool a goal. But there is an inverse lesson here: bad VAR is slow and opaque, while good VAR corrects a clear error. The sports data industry needs something like VAR: a review mechanism that may be slow, but that can fix an error before it hardens into a conclusion.

Imagine what happens when a wrong label travels one step further. A political item tagged football flows into a prediction model. The model learns that words like fence, lockdown, and committee are signals of football. Then one day it reads a genuine piece about a derby with police barriers around the stadium, and it can no longer distinguish football from politics. That is the moment data stops being a mirror of truth and becomes a mirror of its own error.

I once sat in a meeting where people argued about whether to keep a manual review step. One said: manual review slows the system by three percent. Another said: that three percent is the price of truth. The debate ended without a winner. But I have never forgotten the second person's line, because it mirrors exactly what I learned after my 2026 stumble: the price of naming correctly is always cheaper than the price of naming wrongly.

Someone will say: a wrong label, what is the big deal? The content still reaches the reader, it just sits in the wrong drawer. Why exaggerate?

I understand that logic. But it overlooks three things.

First, in the age of machine learning, a label is not decoration; a label is training data. Every time a political item is tagged football, the machine remembers this as an example of football. Repeated a few thousand times, it learns wrongly. And once it learns wrongly, it starts misnaming things that matter more: it confuses a female player with a male one, a friendly with a final, a fake transfer rumour with a real one. A small error at the input becomes a bias at the output.

Second, a wrong label destroys discoverability. Women's football fans live by search. They do not have hundreds of TV channels covering their team every week. They have to hunt. If a correct article about women's football is buried under a pile of wrong labels, the people who need it most will never see it. Invisibility is not only a lack of news; it is news buried in the wrong place.

Third, and perhaps most importantly, how we name a thing is how we value it. When the press calls a female player number seven instead of her name, that is not harmless laziness; it is a statement that her identity matters less than the number on her shirt. When a system tags a political item as football, it states that accuracy matters less than abundance.

Misname once, remember a lifetime; correct it, and you value it more.

I have watched several generations of female players being misnamed. I have also lost my temper with careless colleagues. But I have learned that anger fixes nothing. What fixes things is turning correction into an act of respect: I too have erred, everyone errs; what matters is that she is named correctly from now on.

There was a night during the pandemic, when every league was suspended and I lay still in this apartment for three weeks without writing a line, that I called Ji So-yun, the legendary midfielder who once played for Chelsea. She told me that at fourteen she had to leave her family to train alone, and my tears spilled over. A phone call in the middle of a pandemic taught me that sport heals without an audience. From that night, I wrote differently: starting with a person, then arriving at a number.

And I want to say something hard for my own industry to hear. We cannot demand seriousness for women's football while accepting careless data pipelines. If we want women's football treated fairly, we must hold it to a standard as strict as men's football, no more and no less. The greatest respect is not leniency; the greatest respect is fair criticism, and correct naming.

I am not an engineer. I cannot rewrite anyone's algorithm. But I can tell the people who build the systems one thing that thirty-four years in this trade has taught me: install in your pipeline a small door where a human being can stop and ask whether this row of data truly belongs to football. That door will make the system a little slower. But it will save the system from its own cleverness.

Back to that October morning. I could have deleted the wrong row and moved on. But I chose to record it in a separate file, named mistakes worth remembering. Because I believe a mature sport is measured by how it treats its smallest errors, not by its largest victories.

Every contract is a life relocated. So is every label: it relocates a person within the world of attention.

I do not know whether that machine will ever correct itself. I only know that at fifty, I have learned something I wish I had learned sooner: the pitch has no borders, but the heart has coordinates. And my coordinates, for thirty-four years, have always been wherever women who play football are named correctly.

Tomorrow, perhaps, another wrong row will appear in my inbox. I will stop again. I will ask again: which life is this number hiding? And I will go and find that person before I write. Not because I am a perfectionist. But because I believe sport, at its deepest layer, is not about winning and losing. It is about people waiting to be named.

From the numbers, I see a person waiting to be named. And when a machine misnames someone, my job is not to delete the wrong row. My job is to remember that behind the wrong row lies a real world, with real people, waiting for someone patient enough to call them by their true names.