Where there was no cricket, a cricket label: nine frames from Bogura and a quietly spreading error

Nine frames. Only nine. In the last one the afternoon has already tilted and the fish are laid out on sacking by the roadside, sorted by size — the big ones at the edge, the small ones in the middle. A hand is bent over them, turning one to face the light. No full face in any frame, no shop sign, no date. The numbering runs 1/9 to 9/9, which tells you this is not a clipping from a print page but a web-native photo gallery.

The file carried a label: cricket_asia.

That is the moment I want to hold. Since 2026 I have kept two versions of every match — the one that gets printed and the one that does not. The unseen page gets written first. And the unseen page here is not about fish. The error was never in the fish. It was in the stamp, and the stamp always falls where no camera is pointed.

Chander Bazar. Dekra village, Bogura sadar upazila, Bogura district, Rajshahi Division, northern Bangladesh. The fish come in from Nuruil beel, the shallow wetland that shrinks and swells with the monsoon and holds up the plate of a large part of northern Bengal. The report names six kinds: mola (Amblypharyngodon mola), puti (Puntius species), taki (Channa punctata), shol (Channa striata), koi (Anabas testudineus) and meni. The first five are identifiable. Meni does not map onto any single taxon from a transliterated local name and should be marked unconfirmed. The list is not accidental — this is a classic small indigenous species assemblage, the fish that food-security literature counts for micronutrient density rather than protein alone.

The trading window is afternoon to evening. Caught in the morning, on the roadside by midday, sold by dusk. Demand is explained in one word: freshness. Price is explained in one sentence: prices are slightly higher. No figure, no baseline, no comparison market.

Now the other column. There is not one cricket entity in the entire item — no team, no player, no board, no league, no match, no rule, no broadcast. Every information point carries the same source field: none. No byline, no dateline, no outlet.

And still it cleared the first stage and travelled onward wearing cricket_asia. Everyone inside sports data knows the sequence: the first stage watches volume, the second stage watches meaning. When the first stage is slack, the second does not get to improve things; it inherits the duty of inspection. This report is that inspection, and the inspection says there is no cricket here.

The item has two defects and they come out of the same illness. The label says cricket while the contents hold zero cricket entities. The item's single economic claim — prices are slightly higher — stands without a number attached to it. What both share is the absence of an anchor. No date, no reporter, no institution, no baseline. Content that cannot substantiate its claim cannot substantiate its subject either, and content that cannot substantiate its subject can be made to wear any label at all. That is the knot, and the knot is bigger than a fish market.

The damage runs both ways and both ways are silent. Downstream, a Bangladesh item has entered a cricket corpus with not a grain of cricket in it. Ten of those a month, and by the sixth month some analyst writes, with full confidence, that Bangladesh cricket engagement is rising — a sentence standing on the price of fish. Upstream, the loss is inverted: a researcher looking for beel supply or SIS nutrition will never find this, because it is filed under cricket. A wrong label blinds two communities at once.

Go back to the nine frames. Nine frames, one afternoon, one market. That cannot characterise Bogura's fish economy; beel supply swings with the monsoon, and a dry-season afternoon looks nothing like a monsoon one. 13 March 2026, Eden Gardens — India went from 98 for 1 to 120 for 8, seven wickets for 22, and the upper tier began to burn. I had to leave the scorecard that night. The crowd at Eden Gardens did not watch the story; it became the story. My rule since then is to describe the crowd before the scoreline. In data work the same rule reads: before you count a thing, record how it was collected.

2 April 2026, Wankhede. Dhoni put Kulasekara over long-on and a chase of 275 ended. The piece I filed at three that morning was not about the six. It was about the 1.8 seconds before contact — a stadium inhaling, a bowler pausing at the top of his mark. Data pipelines contain that same 1.8 seconds and no camera in it: the instant a label is applied to a file. Metadata never enters the frame, which is why the most expensive mistakes happen exactly where nobody is filming.

18 June 2026, Tunbridge Wells. No television carried that innings; the BBC was on strike. Kapil Dev made 175 not out off 138 balls, and I filed 400 words that my editor cut to 90. I kept the full copy in a shoebox for thirty years. It had a date in it, a ground, and my own eyes. The distance between an archive and a rumour is one line of attribution. The nine frames from Chander Bazar do not have that line, so they are not an archive.

Put the item in its correct drawer, though, and it is not bad work. The auction board taught me economics faster than any lecture hall, and the same logic applies here: a short chain from beel to roadside, few intermediaries, a modest premium earned on freshness. That is competition on time-to-sale rather than on price — a differentiation posture, not a commodity one. For a micro-economist it is a tidy vignette. The trouble is that it is a vignette of a single afternoon and it does not name the person who made it.

One number is enough: nine. Nine photographs from one afternoon now sit inside a cricket corpus, and nothing in the file records who took them or when. Which means they are as hard to remove as they were to find. Content you cannot pull back out is not data; it is silt.

The industry treats this class of error as housekeeping — metadata hygiene, a tagging fix, the cosmetic end of reporting. I read it the other way. A label is a claim, and a wrong claim about subject matter belongs in the same category as a wrong claim about a score. Data teams audit numbers, never subjects, which is how an item like this clears two stages.

The second habit is blaming the algorithm. The algorithm had nothing to grip — no byline, no dateline, no outlet, no institution. Automation does not manufacture opacity; it inherits opacity and then scales it. Journalism that has stopped keeping attribution cannot be rescued by a tag.

The third habit is cleverer. We assume an evergreen photo feature has no news value and can therefore rest in any drawer. In its own drawer its value is real: seasonal beel supply, the nutritional weight of small indigenous species, rural household cash. A wrong label destroys that value twice — once by hiding it from the researcher, once by feeding a false signal to the cricket analyst.

The way forward is dull and straight: before a cricket label goes on, demand the name of one cricket entity. One. A team, a player, a match, a board. No name, no label. That is an editorial rule, not a technical one, which is why it will hold.

The question left is the one no corpus records: how many items sit under the correct label and still carry that missing line of attribution? Nine frames are waiting quietly. The route to removing them was never written into the file.