Wrong Tag, Wrong Table: Anatomy of a Misclassification Inside the Sports Desk

5 p.m. I climbed into a rickshaw outside the training ground gate. In my bag: the notebook—two on-record lines from the coach, three lines of my own observation about a young defender's knee, and one question I will not write because it was off the record. Phone in hand, I opened the desk's feed folder. Seven items had landed. One was flagged red, because in the system's language it was 'football'.

I opened it. Inside: a court, a provincial government, three names, and an extension of a detention—thirty days. I read the seven information points twice. The first time, I thought I had opened the wrong file. No club named. No competition. No match date. No scoreline. No player, no coach, no contract, no league table.

The locker room tells you before the scoreboard does. This story had no locker room and no scoreboard. It had only the tag.

Let me be clear about one thing first, because many readers have never seen the inside of a desk. A domain label is not a measure of a story's importance. It is an address. It is the mark that decides which door the item walks through. An item arrives; the system reads the words inside and decides—cricket, football, something else. Then the label routes it to a desk, a page, an archive, and these days, a dataset.

This is not new to our trade, only the scale has changed. Sports desks once took news by telephone, paper came off a teleprinter, and a sub-editor sorted it by hand. Now feeds arrive, get scraped, get tagged, get sorted. If I file a Bangladesh Premier League match report from Sylhet, that report travels to eight or ten places. If a wrong label gets attached somewhere, it stays there, because nobody re-reads an archive.

Four or five feeds a week, two or three hundred items a day, one or two people—in that reality the label is the desk's biggest crutch. When the label is wrong, nobody notices.

Look at this record and what strikes you is the arithmetic of absence. Seven information points, one journalistic source, zero clubs, zero matches, zero players. Yet the label is entirely confident. In my working method there is a 'minimum viable record'—a football item must carry at least four things. A competition. At least two teams. A date. And an event—a result, an injury, a substitution, a contract, a transfer, a sanction. If one of those four is missing, it is not football to me, however boldly the word sits on the label above.

I wrote that rule on the first page of my notebook, because the temptation arrives daily. When an item comes in, you have to look at what it is actually about. In my experience the danger comes from somewhere else. The story is accurate within its own house. The facts are not wrong. The wrong thing is the route—the address. That is why the error survives: people go to verify facts, nobody goes to verify addresses.

A wrong fact is correctable, because somebody shouts about it. A wrong label usually is not, because nobody shouts about a label.

How the error happens, I have a hypothesis but not proof—and the distinction matters. The likely cause is a keyword collision. A three-letter abbreviation that simultaneously denotes a political party and, in some feeds, a marker for a football-related index. What the system literally saw was that token. It has no capacity to read context. People had that capacity, but people leaned on the label to save time.

The price of that leaning shows up downstream. A sports desk's archive is not only for readers. Match data, player profiles, transfer trackers, broadcast graphics—all of it today is built from records stored in one place. Discarding a bad record costs nothing. Make a bad record part of a trend and the damage cannot be undone.

Consider this: I track one team's PPDA across three matches. It falls—four, five, six—and the graph points down. Someone pulls a fourth reading from a different team's dataset and drops it in. The graph still points down. It just lies. The most dangerous thing in football analysis is not a wrong number; it is a wrong number with confidence. A label does exactly that—it gives confidence to a number even when it is wrong.

This is where single-source reporting becomes most dangerous. The evidential base of this record is one report. An unlabelled single-source item makes people check. An item wearing a clear label makes nobody ask—because the question is not about the label, it is about what is inside the label. And nobody reads what is inside.

In 2026 in Saransk that lesson settled into my body. Two hundred journalists in the media tribune, eight of them women. A European colleague said women do not understand tactics. I did not answer. I tracked Japan's 4-2-3-1, wrote about how Yuya Osako's hold-up play served the whole team, and filed 1,800 words at two in the morning.

In that piece I had to place every tag myself. Which touch was a hold-up, which duel was won, which pass was progressive. One wrong tag and the story of the match would have been wrong—Osako would have become a different footballer, one he is not.

After filing, I spoke to two fans in Sylhet who had been awake since one in the morning. They watched that match for football alone. They do not read my labels; they read what is written beneath them. At 1 a.m., the shape of Japan looked like a promise—and when wrong information slips inside that promise, it is worse than damage to them, because they have already decided to believe.

Now the outside reading. When this kind of thing happens, everyone points in one direction—the algorithm. The machine erred, artificial intelligence failed, the system is guilty. I locate the fault elsewhere. The machine did exactly what it was told. It was told to look at tokens. It looked at tokens.

The charge is editorial, not technological. A desk that does not ask 'what is this story actually about' before publishing is not filing any story correctly—it is merely moving labels from one place to another. Investment in sports desks has fallen, sub-editors have gone, and the verification step was dropped in some overnight reorganisation. Nobody wants to pay to restore the step that was dropped, because the error is invisible. Its invisibility is what keeps it cheap.

And something nobody wants to say. Records like this are often not discarded but kept—because they draw readers. Something unexpected landing on a sports page lifts clicks. When clicks rise, nobody asks whether it has anything to do with sport. Here is my second objection: if sports journalism borrows another room's news to inflate its own page, it stops knowing its own room's boundaries.

I am in favour of timestamped interim notes. Because a beat is not a final word; a beat is continuity. What I write in today's notebook may need correction in three weeks. Keeping the correction path open is how label errors get caught too. Every entry in my notebook carries a time beside it, because without time nobody owns a fact.

The same rule should apply to labels. Who placed it, when, on what basis—without answers to those three questions, a label is an opinion, not a fact. Put an opinion where a fact should sit and a dataset rots silently, without a sound.

The beat keeps time even when the story breaks. My job is to remember that, every day, at five in the afternoon, when the notebook opens.

I count the small details because they are the ones that stay. Seven information points, one source, zero clubs—those zeros are the real story. Why a football label sits where there is no ball is not a small question. Because if this error is not corrected once, it will happen again tomorrow, at another desk, on another feed, in another collision of words.

Looking ahead, what I see is the audit trail. An apology is something to accept, but an apology has no data audit. My question is therefore bigger than the answer: a desk that cannot say what its own story is about—what exactly is it filing, and who is reading that file forty years from now?