The mislabelled football tag and the crack in the sports data pipeline

On 15 September, a text file landed in a sports content analysis system in Shanghai. The label on the file read: football. The content inside was a message marking the UN International Day of Democracy from Pakistan's President Asif Ali Zardari and Prime Minister Shehbaz Sharif, issued by the President Secretariat and the Prime Minister's Office. No club. No player. No scoreline. No transfer fee. I read that file three times in a single morning. First to find the error. Second to make sure I had not misread. Third because I understood the problem was not the file. A few years ago, systems like this were still called support tools. Now they decide a good deal: which piece leads the front page, which item is pushed to whom, and most importantly, which number is treated as true. A mid-sized sports newsroom in Asia now runs thousands of documents a day, and most of them are never read end to end by a human editor. Tagging is usually automated, driven by keywords. And keywords are a very easy instrument to fool. Four words repeat throughout that file: governance, rule of law, constitution, accountability. To a classifier trained on sports corpora, those four sit extremely close to the vocabulary of competition rules and financial fair play. Governance in football means UEFA, means FFP, means disciplinary committees. Accountability means sanctions, points deductions, transfer bans. The system cannot tell a national constitution from a federation statute. The result is a political document pushed into a football analysis template. And the real problem sits in the next step: once it is inside the template, the system does not stop. It fills the blanks. I have sat long enough in analysis rooms to know what follows. The framework has seven sections, and every section demands data. With no real data there are only three choices: leave it empty, stop, or manufacture something that looks like data. The third option always wins, because it is the only one that produces a product to sell. That is when the metrics sprout. PPDA. xG. Wage structure. Debt-to-revenue ratio. Those numbers look professional, look credible, and have nothing to do with a message about democracy. Based on my experience tracking matches and deals, a metric only carries value when people know the circumstances in which it was measured. xG does not explain why a striker misses in the 89th minute, and it does not explain why a referee blows the whistle. Overused, it becomes decoration for conclusions that were written in advance. Concrete numbers make it clearer. Across the 2026-2026 season I entered all 128 deals in China League One and cross-checked them against 47 deals in the Chinese Super League. Second-tier clubs paid 22 percent more for under-23 forwards than top-tier clubs did. That figure appears in no prospectus. It surfaced only when I compared each contract by hand. In the second tier, the cleanest data is usually the data that has been polished the hardest. In 2026, in the corridor of the Luzhniki stadium, I heard a Portuguese agent describe the purchase of Nigerian striker Emeka Nwankwo, aged 19, seven goals in nine U20 matches, for 3.2 million euros, with an unusual clause: his former club retained 40 percent of any future transfer fee. That contract was never published in full in any outlet. But it existed, and it shaped this player's price for years afterwards. There is a deeper layer few people notice. Most sports content today is produced on a line: collect, tag, analyse, publish. Each stage has its own owner, and nobody sees the full route of a single fact. If the label is wrong at stage two, everything downstream is wrong too, but wrong in a tidy way, with structure, with a headline, with tables. That kind of error is more dangerous than an error that shows itself. An error that shows itself gets fixed. An error presented beautifully gets shared. Many people worry that machines will write badly. I think that worry aims at the wrong target. The machine does not write badly; it writes exactly to the label it is handed. The failure sits in the human check, and that check is thinning out. I learned more in the Luzhniki corridor than in the press conference room. The biggest lesson was that information does not travel from person to person on its own. Somebody has to carry it across. In an automated line, nobody carries anything. In Riyadh, I once chased a deal that an English paper called fabricated. I spent two weeks, met three officials and one bank, and found an 18-million-euro debt from an older deal that had broken the debt-to-revenue ratio. Riyadh taught me a lesson: money cannot buy FFP, it can only buy time. Sports reporting works the same way. A wrong label does not disappear on its own. It simply waits long enough for someone to forget it was ever wrong. The consequences do not stop at one faulty article. Scouting departments, investment funds and data companies all buy content from the same pipelines. A mislabelled report can pass through three or four layers before anyone traces it back to the root. Every number on a screen is a story that was never told outside the corridor. Erase the corridor and keep only the screen, and information does not become cleaner. It only becomes harder to verify. People argue about whether machines can write sports news well enough. The real test lies elsewhere: if a document about the International Day of Democracy can still be labelled football and travel the entire pipeline without anyone stopping it, how many numbers now circulating in the transfer market were born in exactly the same way? The transfer market does not run on money. It runs on promises that never make it into the contract. And a promise that was never written down is a promise nobody can check.