When the Source Field Said 'None': A Non-Football Story at a Football Desk

At 2:40 in the afternoon, at my desk in Ahmedabad, I opened the file that had arrived labelled 'football'. There were twenty-five information points. I did what I have done for thirty years — read the source column first, the story second. Thirteen of the points carried the same line: Source — none. No club, no player, no scoreline, no gate figure. I still have the notebook from that night in Salt Lake — 13 July 2026, East Bengal 4–1 Mohun Bagan, official attendance 131,781 — but there was nothing in that file I could set beside the page. There was an American actress's death, a recording of an emergency call, a coroner's date — and a label stuck on top: football. The team sheet first, the match report second. That is my only opening habit, and it is not confined to football. Every claim is made in a room — someone said something, someone wrote it down, someone put a date on it. In my notebook every figure carries the name of who gave it and when. 131,781 is not a magic number on its own; it came off a ticket counter on a particular afternoon, and if someone challenges it tomorrow I can produce the receipt. That is the difference between a press release and a filing. A release says 'record crowd'; a filing says how many, when, and who counted. Between July 2026 and March 2026, when I worked in the AIFF's media department, that difference became the day job. Documents arrived with both a date and a signature on them, and then went out as press notes that had lost the date and the signature. In football we level exactly this charge at institutions — the I-League launched in 2026, the ISL in 2026, the NFL in 2026, and the three launch dates are still not gathered in one place. The August 2026 suspension, or the club licensing paperwork: the record exists somewhere, just not in a public register. Then the same thing happened that we accuse others of. The file in front of me looked clean on the surface — dates, institutional names, a coroner's office, two national news channels. But the moment I read the source column line by line, the frame began to open. Thirteen of twenty-five points had no source recorded. In football terms: it is a match report with no stadium, no date and no referee — only a score. You cannot file that, however correct the score. Attribution density — the share of information points that carry a named source — is the most important number in the whole document, and it was printed nowhere. It comes out at roughly 52 per cent. In my experience, below 60 per cent the confidence scores turn into noise. That should be the first filter on any desk, ahead of the headline. One thing in the file was sound, and it was procedure, not content. ABC News obtained the audio from Greenville County emergency medical services through an access-to-information request; this was the citizen-to-operator channel, not internal police-paramedic traffic. Separately, forensic authorities classified the death as accidental on 22 September 2026, cited toxicology and found no traumatic injury. The chain is legible: coroner, classification, public-records release, newsroom publication. A football desk recognises it — club, league filing, public disclosure. Every record has a room it was made in; I went to find the room. The classification date was a Tuesday — 22 September 2026 — a day when most people had work in the morning. Records fall on Tuesdays, not at press conferences. The same applies to football: transfer dates get logged on Mondays and Tuesdays, and later become 'deadline day'. And a warning for myself. Do not use 131,781 as the yardstick for every number. That was a Federation Cup semi-final in 2026, in a football city; today's ISL playoff happens in a franchise market. If a comparison is being made, say which one. The same discipline applies to sourcing — the 2026 page is not evidence for a claim made today, it only tells you who was counting then. Now the part outsiders will misread. The easy explanation is that this is a technical bug — the classifier stuck on the wrong label, fix it, move on. I do not think the classifier is at fault. It slipped because the extraction layer stripped the sources. The primary documents were present in the raw reporting — the coroner's classification, the access-to-information audio, two newsrooms competing — and all of it was removed. What survived was a claim without a source, and a claim without a source will sit under any label you choose. That is the same disease I watched for three years at the AIFF: document present, signature present, nothing in the register saying who said what and when. The content value of this item is low; the process value is high. The access-to-information chain is a positive template — it shows how a claim should be traced. And what is most striking is that two national outlets chased the same record at once. That is the pressure of a transfer story: the second outlet confirms the first, and nobody confirms the player. The rule here is plain: publish the document, attribute on record wherever possible, and log the denial with its date beside the claim. In this file, the denial was the absence of a source. The question is no longer what was in that file. It is who keeps the record now. The suggestion is simple: sample the Stage-1 output at random, and if attribution density falls below 60 per cent, the confidence scores belong in the bin. And when the next non-football story arrives wearing a football label, will anyone read the source column before the headline? You learn more from an empty stand than a full one — the crowd makes noise, and the empty seat tells you who did not come. In this file the crowd's place was taken by an empty source field. The stage is new; the nerves are the same ones from 2026.