A Football Label on the Wrong File: How the San Judas Tadeo Notice Exposed a Data Pipeline Fault

On the morning of 27 October 2026 the file on my desk was labelled Football. Inside, a single page carried a straight count: 350 police officers, 32 vehicles, 8 motorcycles, 5 tow trucks and one helicopter. My first assumption was that a club had sent its match-day operations brief for a derby. The next line placed the scene at colonia Centro in Mexico City, alcaldía Cuauhtémoc, for the annual San Judas Tadeo religious celebration. The notice came from the Secretariat for Citizen Security (SSC) and set out road closures, diversions and safe routes for pilgrims across 27 and 28 October. There was no football anywhere in it. Not a formation, not a transfer, not a single recorded minute. I had counted 131,781 seats before I trusted a single attendance figure. Here there were no seats to count. That was the moment the whole analytical frame tilted under its own weight. This is a story about a system, not a story about a news item. A large share of football coverage now runs on automated entity tagging. A feed is scraped for keywords — celebration, date, crowd, operation, deployment — and a classifier assigns a domain. If words like club, match or stadium appear anywhere nearby, the label lands on Football. That is what happened here. The San Judas Tadeo observance returns every 28 October; it is cyclical, and cyclical events carry a keyword fingerprint that overlaps heavily with the sports calendar. The root of the error sits in process, not in vocabulary. In 2026 I coded ten ISL matches for an event-tagging exercise, logging every sprint, every recovery interval, every line drop. That work taught me one durable rule: a tag and a fact are not the same object. A tag is a claim. A fact is a verification. When a pipeline treats an input tag as a verified fact, the error propagates downward without friction. Run the standard nine-dimension audit on this file and the result is absolute. Tactical analysis: not applicable. Club finance: not applicable. Transfer market: not applicable. League landscape: not applicable. Governance compliance: not applicable. Dressing-room dynamics: not applicable. Risk profile: not applicable. Media narrative: not applicable. Industry transmission chain: not applicable. Nine out of nine. This was not a partial failure. It was a complete domain mismatch. The real danger lives exactly here. Faced with such a file, an analyst has two roads. The first is to state that the material does not exist and return a null. The second is to force-fill the template, because the template is sitting there and an empty return looks like a bad day's work. The second road is cheap and the most expensive. One could write that the 350 officers represent a pressing structure, that tow trucks are defensive cover, that the helicopter is aerial scouting. All of it is writable. All of it would be false. Football does produce genuine deployment documents, and they have a settled format: steward counts, sector-by-sector gate allocation, medical team positions, fire marshals, turnstile throughput. When I was reconstructing the 2026 Federation Cup semi-final at Salt Lake from a single wide-angle tape, the stadium management's sheet measured crowd density per square metre while my own sheet logged East Bengal's seven completed switches in the first 25 minutes and Bhaichung Bhutia's three right-half-space entries. Two different languages, both counting. The San Judas Tadeo notice also counts, but its numbers do not speak the language of the game. A counter-intuitive conclusion follows, and it is the most useful thing in this whole episode. The SSC notice is a clean document. It names the responsible agency, gives dates, states concrete resource figures, and then openly flags what it has not supplied — which streets, closed at what hour — directing the public to official updates. The agency declared its own limits. Compare that with how football communicates. On injury, the standard output is muscle injury, out for a few weeks. Which muscle, what grade, which rehab phase, what return-to-play criterion: silence. Clubs stay vague until share price or playoff arithmetic forces their hand. The Mexico City traffic police disclosed more about their operation than most football clubs disclose about a hamstring. The comparison is uncomfortable and it holds. The contrarian point is that the mislabel is not the primary failure. A confidence gate that does not exist is the primary failure. Every classifier will err; that is expected. An error becomes dangerous when no checkpoint in the pipeline takes the first hundred words, holds them against the label and asks: is there a team, a venue, a fixture, a recorded minute in here? If the answer is no, the file leaves the system. The deeper sickness is incentive design. Pipelines measured on daily output volume treat a null as a defect, so analysts learn to force-fill. Minutes are the ledger; marquee names are only the cover page — and here the marquee name was Football itself, sitting on the cover and hiding the empty pages beneath. A third, more restrained observation is needed so that I do not fall into my own trap. Central Mexico City is shared urban infrastructure. The streets that carry a pilgrimage also, on other days, carry crowds to a stadium. A link can be theorised — urban logistics, match-day traffic, stadium access. But this notice names no fixture, no club and no stadium. The link is speculation, and my job is to label speculation as speculation. The spreadsheet said one thing, the touchline said another, and the touchline won. Here the touchline was empty. The correct classification for this file is civic public safety or local news, not sport. It should be removed from the football analysis pipeline immediately, because a single mislabelled file contaminates every ledger beneath it. Had this item reached a content engine, the output would have been a headline about a football security operation in Mexico City — entirely invented, and entirely publishable. Before the hype curve there is a measurement nobody bothers to take. Here it was simply this: read the first paragraph, not the label. When the next batch is ingested, watch two numbers — how many files returned a null, and whether that null was suppressed as a failure. A pipeline that forbids the word null will fill every blank page, and the reader will never learn that he is being told tow trucks are defensive cover. The question is not why the classifier erred. The question is whether a human was positioned to catch it, and if so, why that human did not have the nerve to write null.