The File Labelled 'Football': An Autopsy of a Misclassification

Last night I opened a spreadsheet expecting confirmation and found a confession. Fourteen rows, each marked "Source: None" in the right-hand column. At the very top sat a single-word label — Football. I searched for a club. Nothing. A coach. Nothing. Formation, pressing trigger, set-piece geometry, xG, PPDA, transfer fee, amortisation, league table — eighteen years into this craft I do not write a paragraph without them, and not one of the fourteen information points carried even a shadow of them. What I found instead was an account of the third elimination gala of a television competition. One line stopped my hand: "the third farmer to leave." The word farmer has nothing to do with soil here. In the internal vocabulary of La Granja VIP 2, the contestants themselves are called granjeros. To a machine, though, the word was a green light. Elimination, season, competition, nomination, support — put those five words in one paragraph and an automated classifier nods and writes: Football. What the record actually contains is this: TV Azteca's La Granja VIP 2, the third elimination gala of the season, and the exit of Carlos Trejo, the contestant who drew the lowest support under a positive-voting mechanic. The other names are no more football than he is — Manelyk González, Daniela Alexis "La Bebeshita", Kenny Avilés, Julio Camejo, Abigail Pak "La Coreañera", María Karunna. The format's only rules are La Salvación and La Traición. No club. No player. No league. No governing body. No transfer, no contract, no match, no goal, no point. I opened the spreadsheet expecting confirmation and found evidence of a pipeline failure. Anyone who works on a football analytics pipeline knows what a label really is. When an article enters the system, the first job is to identify club, player, competition. The label decides which questions get asked next. A football label triggers nine dimensions — tactics and technique, club finance and the transfer market, results and the public-opinion cycle, league landscape and team positioning, rules and governance, management and dressing room, risk profile, media narrative, industry transmission. A wrong label means wrong questions, and wrong questions mean every answer is fabricated. I know this because I once started a match breakdown from the wrong end myself. At the 2026 Abahani Limited Dhaka 2-1 win, I did not trust my new expected-goals model; I charted fourteen pressing sequences and twenty-three line-breaking passes by hand and waited ten matches before I said the model's name out loud. Without that patience at the level of the label, analysis is only arranged guesswork. So this time I will conduct the autopsy — but on the body as it actually lies. Dimension one, tactics and technique. There is no subject. No system, no formation, no in-game change, no pressing scheme, no set-piece design. The only competitive mechanism in the record is a broadcast format rule: viewers vote to retain a contestant, and the lowest support triggers elimination. That design and match tactics are not the same species of thing. Dimension two, club finance and transfers. Zero. No fee, no wage, no contract length, no sell-on, no amortisation. The real hazard hides here. When a data system proceeds with mandatory fields empty, an empty field can drift over time into being read as "verified zero" rather than "not applicable." A player's transfer value is not zero when the player does not exist. Dimension three, results and public opinion. This is the one dimension with a loose analogue in the record — an elimination outcome and a public vote. As football judgment, it is still void: no table, no points per game, no fixture difficulty, no relegation or title stake. Three eliminations cannot establish a trend, and even by entertainment standards that is a small sample. Dimension four, league landscape. Imaginary. The only landscape present is the Mexican television celebrity ecosystem — a broadcaster, its programme, its contestants. That cannot be passed off as team-tier positioning in a football pyramid. Dimension five, rules and governance. The source does contain rules: La Salvación, La Traición, positive voting. These are production regulations, private governance written and enforced by the broadcaster itself. They bear no relation to FIFA, UEFA or any league rulebook. This is where the language trap cuts deepest. The words rules, sanction and elimination belong to sport, but here they are studio procedure. Dimension six, management and dressing room. There is no dressing room. No ownership, no sporting director, no coaching staff. The only authority embedded in the record is the production team. Carlos Trejo's exit is a voting outcome, not a contract event, so contract-year effects, renewal brinkmanship and agent manoeuvring are simply not calculable. Dimension seven, risk profile. This is the only dimension with a genuine, measurable risk — and it is not a sporting risk but an information risk. The largest is that a football-labelled record containing not one football sentence is entering an analytics pipeline. Alongside it sits an equally serious sourcing risk: thirteen of fourteen information points are unattributed. The date is given only as "Sunday, September 27" with no year, so the event cannot be placed in any timeliness window. And the final paragraph is truncated, its list of remaining contestants missing. None of these four defects is a football defect. Every one is an editorial one. Dimension eight, media narrative. Here the source is legitimately analysable — not as football, but as a media artefact. And what it shows is entirely familiar. Carlos Trejo is known in Mexican press culture as a "ghost hunter," host of the Cañitas strand. His exit was reported by the same broadcaster that produces and airs the show. Producer, broadcaster and publisher are one entity. In that closed loop, the narrative will never say its own programme's weak point aloud. I have seen this pattern in football many times. When a club's own channel explains the club's tactical victory, the analytical question shifts — from "is this true" to "whose interest does this framing serve." La Granja is the same structure, with a studio in place of a pitch. Dimension nine, industry transmission. Zero. No youth pathway, no agent, no rights holder, no licensing, no betting-market signal. Nine dimensions written out, nine of them null. Here lies the most important lesson: structural completeness is not substantive validity. A report can tick every box and still be false. In sports journalism this error usually runs the other way. There, a story is built from statistics whose conclusion was written in advance. What happened here is a deeper event: the machine misread an entire domain, because entertainment formats borrow sporting vocabulary to describe their own competitions. Selection, elimination, season, support, nomination. Each word is innocent alone. Together they mislead. That lexical collision is the likely root of the mislabel. The cost? If this record enters training data, a model will learn that the word elimination signals football context. Once learned, it will stop distinguishing relegation coverage from a reality-show eviction night. A bad label that gets into a pipeline goes on to produce its own offspring. I will not construct the tactical bridges the record invites. Positive voting is not a low block. Lowest support is not a relegation battle. La Traición is not a dressing-room betrayal. Each bridge looks professional. Each is fabricated. What stopped me was not a rule about publishing but a rule about pre-registration: before reaching a conclusion, write down what evidence would make you discard it. Here the discard condition was simple — at least one football entity, a club, a player, a competition or a governing body. I found none, so the conclusion was refused before it was written. The second, counter-intuitive claim deserves equal space: the most valuable output of a football pipeline is often a refusal, not an insight. Commercial models reward recall — catch more results, look better. In football, ten false positives do less damage than one missed signal, because false positives spread silently. Correct refusal is the only defence. The one genuinely profitable thing in this episode is a negative control. This record is a near-perfect false-positive sample. Show it to a classifier and ask whether it is football; if the machine says yes because it found vote, support and elimination, you have measured the edge of its judgment. Looking forward, three tasks. First, install a gate at every intake: no football record without at least one football entity. Second, write "not applicable" explicitly in every empty field rather than leaving it blank, or zero and absence will blur together. Third, never store an undated record, because a Sunday without a year fits no timeline. The largest task remains old inventory. On my own spreadsheet, how many files still sit labelled Football with no football inside them? That number is unknown to me today, and the unknown number is the only honest conclusion this article can offer. When the next batch arrives, I will not look at the label first. I will look for a club's name.