One Grid, Twenty-Five Data Points, Zero Football: A Misrouted Record Inside the Sports Data Pipeline

On Monday morning a record landed on my desk with a label clipped to it: Domain: Football. My habit is to look for the formation first, then PPDA, then the xG differential across the last six matches. All three columns were empty. No team, no coach, no match. What I had was a public-safety alert from Mexico City — a warning about the alleged use of animals in esoteric rituals around Día de Muertos on November 1 and 2, 2026. The list included black and white dogs, cats, goats, roosters and ravens. Twenty-five information points, not one of which answered a football question, and yet there it sat inside our pipeline. That is the interesting part of the story.

Context: how the label got attached

The source is the Consejo Ciudadano para la Seguridad y Justicia de la Ciudad de México — a citizen council, not a prosecuting authority. CompAnimal programme data is layered on top. The claim is 203 complaints linked to esoteric practices since 2026, against 59,588 total complaints since 2026. January to June 2026 produced 6,616 complaints, 28 per cent higher than the same window in 2026. Among the recommendations, one stands out: temporarily suspend adoptions of black-coated dogs and cats.

Now the mechanics. Sports data desks no longer sort records by hand. An entity-extraction model pulls names, places and keywords from the text, and domain assignment follows keyword weighting. Here, "Consejo Ciudadano" was probably read as a club or supporter body, "roosters" and "ravens" as team nicknames, and "crisis" as a run of poor form. That is inference, but it comes with a clean test: if other non-football records in the same batch carry the same label, the fault sits in the classifier, not in one file.

I still keep a grid from 2026, built live from the gantries at Loftus Road and Selhurst Park — six vertical channels, three horizontal bands, 127 matches in a season. It taught me something I use weekly. When data arrives through the wrong door, the danger is not that it is useless. The danger is that it looks usable.

Core analysis: numbers that cut against their own headline

Start with the scale check the source never performs. 203 ritual-linked complaints since 2026 against 59,588 total since 2026 is roughly 0.3 per cent. The base periods differ, so that is an order-of-magnitude estimate, not a ratio. But if the headline sells one threat and the caseload describes another, the gap is the story.

The second data point speaks more plainly: 86 per cent dogs, 9 per cent cats. That distribution is not a ritual story. It is a story of domestic neglect, physical aggression and animals kept tied up — chronic welfare failure. The document itself concedes that abuse extends beyond the October and November celebrations. The source's own dataset half-refutes its own framing.

The third and largest failure is metric discipline. Complaint volume is not abuse incidence. Reporting-channel friction, media attention and a concurrent awareness campaign all push the count upward. The 28 per cent rise for January to June 2026 is real, but it is a compound of incidence and reporting propensity, and the split is nowhere recorded. Football makes the same mistake every week — touches in the box read as dominance, tackle counts read as defensive quality. Both are activity proxies, not outcomes.

Fourth, the source chain. Nine of twenty-five information points carry no source. Both quantitative datasets trace to a single institution, so neither independently corroborates the other. There is no named spokesperson, no byline, no publication date, only an iStock credit. Most seriously, the authority that should receive a complaint is never named. The public is told to report, without being told under which offence or to whom.

Fifth, the species sit on different legal tracks. Ravens are wild birds under wildlife rules; dogs and cats under companion-animal welfare; goats and roosters straddle livestock and wildlife provisions. A single undifferentiated warning collapses three regulatory regimes into one.

Contrarian: the most benign-looking recommendation carries the most risk

Suspending adoptions of dark-coated animals sounds logical — take the target off the shelf for a few days. But shelter data has documented black dog and black cat syndrome for years: demand for dark-coated animals already runs below average. Removing them entirely for a 48-hour window may entrench the very prejudice the alert exists to fight. No impact assessment accompanies the advice, and neither does any estimate of the backfire.

There is a second thread here that runs straight back into football analytics. The same conflation happens on our desks constantly — an injury note becoming a form slump, three events becoming a "crisis", a transfer rumour forwarded as a medical report. The pipeline that turned an animal-welfare alert into football content can turn a rumour into a scouting report tomorrow. Deleting the record and moving on is the easiest response and the wrong one. This record is a canary, not rubbish.

What to watch next

Complaint volume for November 3 to 10, 2026 will show whether the alert changed behaviour or only reporting. Watch whether "suspend" softens into "supervise" — that language shift will signal whether the stigma-backfire critique landed. And watch for the moment a prosecutor's office or a specific statute is named. On that day the alert stops being advice and becomes enforceable guidance. Until then the question is not how the record entered the pipeline. It is why nobody noticed it for so long.