A “Football” Label on a Traffic Incident Report: The Data Misclassification and Its Cost

On my desk sits a data file marked “Domain: football”. Inside, seventeen information points describe a roadside altercation on the Mexico–Puebla highway, at kilometre 26, near the Puente Blanco bridge in the State of Mexico. An app driver was struck with a blunt object during a traffic dispute. Family members, friends and fellow platform drivers set up a blockade on the road, demanding that the authorities locate those responsible and investigate the incident. The queue of vehicles behind them stretched more than three kilometres. Mexico's federal roads and bridges authority, CAPUFE, issued a notice on lane reductions and advised motorists to take precautions.

There is no club in that file. No player. No league, no match, no contract. I spent another twenty minutes simply confirming I had not opened the wrong document. The football label sat there, bare, and it became the most important fact of that working day. This is the kind of error I call a domain misclassification — attaching a specialist label to a text that belongs to an entirely different field.

The transfer window is when this error does the most damage. Volume spikes, newsrooms race against algorithms, and each record passing through the pipeline is checked by a single label field. When the label is wrong, everything downstream is wrong with it: recommendation models read it wrong, indices read it wrong, and the final report delivered to the reader is wrong too. A contaminated record does not sit still. It spreads.

The Stage-2 analytical framework runs across nine dimensions, from tactics and technique, club finance, results, league context, rules and compliance, through to coaching, risk profile, media narrative and industry transmission. All nine ran across that file and stopped at the same conclusion.

The tactical dimension asks for sophistication in build-up, quality of execution, personnel fit, and metrics such as expected goals or passes allowed per defensive action. The file offers not a single line. The finance dimension needs broadcasting revenue, commercial revenue, wage bill, net debt. Nothing. The results dimension needs a table, recent form, fixtures. Nothing. The rules dimension needs financial regulations, transfer registration, disciplinary sanctions. Nothing.

One detail is easy to misread. CAPUFE appears in the text as an authority issuing a public notice. It is Mexico's federal roads and bridges body, not a football governing body. Its lane-reduction notice is a transport advisory. If an automated pipeline scans for the phrase “regulatory authority” and files it under sports compliance, we have just created a false positive — and that false positive will sit in next season's training data.

The same thing happens with pressure vocabulary. The text describes a civic gathering: the victim's family, friends and fellow drivers demanding justice. In football terms, public pressure refers to heat on a manager, on a board, or on a few underperforming senior players. The crowd on the Mexico–Puebla highway is not a stand. Filing the two under one variable is a category error, and category errors cannot be fixed by adding more data.

The text also mentions a network of relatives, friends and platform colleagues. Some models will read that phrase as a dressing-room structure. But a dressing room is a professional space with hierarchy, contracts and generational turnover. A group gathered around a crime victim is a social support network. The two differ in kind, not merely in degree.

What matters here is not what that text is about. What matters is how a professional system must respond when it discovers it is holding a mislabelled record.

The principle I follow is transparent null handling: when a dimension has no underlying data, the correct answer is to state plainly “insufficient information — cannot assess”, not to fill the gap with speculation. It sounds simple. Production pressure pushes writers the other way. An empty analysis has nothing to publish. An analysis padded with inference gets page views.

I started from a battered spreadsheet, and it became the memory of an entire profession. In 2026, as a third-year undergraduate in Beijing, I tracked 240 matches of a domestic season and logged 127 penalty incidents. I spent three months cross-checking each incident against the laws before publishing a single line. That habit taught me something I still apply daily: time spent verifying a record is always cheaper than time spent repairing the consequences of a bad one.

A referee's mistake is never random — it is a blind spot you can chart. That blind spot has structure. Referees err in zones the human eye struggles to read, in phases with many overlapping layers, and at moments when physical capacity drops. Data pipelines behave the same way. They fail where the label field is the only thing checked, where one shared keyword is enough to join two different fields, and at the stage when output speed is pushed hardest.

If you think this is the private story of one stray file, look at the transfer reports running every day. A player is attached to club A because his agent had lunch with club B's sporting director. A deal is called “nearly done” on the strength of a deleted status update. Every one of those records shares something with the file on my desk: the label comes first, the evidence arrives later, and sometimes it never arrives.

Some information is not wrong. It simply arrives at the wrong time. And some information is on the right subject but filed in the wrong drawer.

The counterintuitive point sits here. The natural response to a mislabelled record is to rescue it — to find some football angle to write, to turn the classification failure into a bigger lesson, or to fold it into a wider feature. That response causes two losses at once. It legitimises dirty data inside the process itself, and it dilutes the real signal, making it harder for readers to tell evidence-based analysis from inference built on a misleading headline.

The correct handling runs against instinct: stop, flag the error, and produce no football content from that source. The greatest value of a mislabelled file lies in the hole it reveals in quality control, not in the article it might generate.

There is also a source-quality detail here that applies across sport. The original report rests partly on family testimony, in which the act is described as alleged, and partly on security footage reported to have captured the incident. That is a source type deserving its own marker: there is testimony, there is footage referred to, but verification is incomplete. In football, most transfer news has exactly this source structure, and it is usually presented as though confirmed.

Fans remember the incident; I remember the context. Context is always more reliable. A passage of play removed from its context becomes a short clip to argue over, not a fact to analyse. A news item removed from its domain becomes a line of noise, and the further that noise travels, the harder the damage is to trace back.

Looking ahead, I expect sports newsrooms to build a domain-validation gate at the entry point, before topic tagging. It need not be complex. It needs to answer one question: does this text contain at least one entity belonging to the domain it has been assigned — a club, a player, a league, a sports governing body? If the answer is no, the record must be halted and reclassified rather than flowing further down the analytical chain.

Alongside that sits a rule for marking unverified sources. Any claim resting on second-hand testimony or reported footage, rather than footage that has been checked, should carry its own tag. That tag does not diminish the report's value; it states the level of certainty, so readers know where they stand.

And perhaps the most important point for anyone in this trade: a good data system is not measured by the volume of records it holds, but by the number of records it dares to reject. The ability to say “this source is unusable” is a professional skill, not an evasion. In a transfer window where thousands of lines pass by every day, readers do not need another article. They need a filter they can trust.

That file will not become a tactical analysis. It will become an error ticket, and that ticket is worth far more than any speculation I could have written from it.