A 'Football' Label on a Human Tragedy: Data Misclassification and the Unexplained Death of Joselyn Sandoval Calderón

There is a file in my tracking system, labelled "football". Inside it there is not a single player. Not a single match. Not a single club, coach, contract, or league. There is only a name: Joselyn Sandoval Calderón, aged 21.

That was the moment I realized that the system I had trusted for years — the system that classifies news, tags topics, and pushes data into the right shelf — had made a mistake I could not overlook. Not a purely technical mistake. An operational ethics mistake: a human story placed in exactly the place it did not belong.

I am not writing this piece to analyse football. I am writing it because in my profession, the line between the right topic and the wrong topic is not an administrative detail. It is the foundation. And when the foundation collapses, everything built on it must be questioned again.

Context: a story that does not belong on the pitch

What we know, as of this moment, comes from a preliminary news report. Joselyn Sandoval Calderón, 21, was a student at the Centro Universitario UAEMéx Valle de Teotihuacán. She was reported missing. The university community circulated a public search appeal. After two days of searching, her body was found.

The location reported is the Otumba area of the State of Mexico, near the highway linking Mexico City with Tulancingo. The response unit credited with the search and recovery is named Protección Civil y Bomberos de Otumba. The family performed the formal identification. The State of Mexico prosecutor's office, Fiscalía General de Justicia del Estado de México, abbreviated FGJEM, opened an investigation to establish cause of death.

That is the entirety of what can be stated with certainty.

No cause of death has been published. No suspect has been detained. No official hypothesis has been made public. The report stops exactly at the threshold where a decent report should stop: it records the event, names the parties, and does not speculate.

In the world I work in, that is a rare kind of patience. The sports data industry lives on speed. A goal goes in, and ten minutes later there are three data tables, five threads of argument, two trending topics. We are paid to run ahead of events, to pre-empt conclusions, to turn a moment into a model. But here — where the subject is a person who has died, and the central question is what happened to her — every professional instinct I have has to go quiet.

That is why I decided to write about the very process that put this story into my system. Because that is where the error lives. And that error, in the end, is the same kind of error the sports data industry commits daily, differing only in severity.

The mechanics of a misclassification

Let me explain how a report about the death of a university student can carry the label "football".

News aggregation systems in general, and sports monitoring pipelines in particular, operate on a set of signals. There are semantic signals: keywords in headlines and body text. There are entity signals: recognized people, organizations, and places. There are source signals: the section of the outlet the item came from. And there are propagation signals: who shared it, and in which groups it appeared.

When all four signals point the same way, the label is assigned and almost nobody checks it again. The problem arises when one signal is noisy, or when a weak link is read by the system as a strong one.

In this case, I can imagine at least three paths to the error.

The first is an institutional link. Centro Universitario UAEMéx Valle de Teotihuacán is an educational institution. Universities in Mexico, as elsewhere, field student sports teams, inter-university competitions, and recreational football activities. A system harvesting data on university football might have scanned this institution's name and filed it under sports. When that name later appeared in a different report — this time, about a student's death — the system dragged the old label along.

The second is a geographic link. Otumba and the Teotihuacán area are toponyms that can appear in local sports coverage, amateur matches, and community football leagues. A regional filter may have pulled this report in by accident.

The third is a propagation link. If the story about Joselyn was shared in community groups that carry a sports element, or if the first sharer was an account that normally posts about sports, the propagation algorithm may have learned the wrong lesson.

All three paths lead to the same result: a human story turned into a sports data entry.

This is not a theoretical hypothesis. I have watched the same thing happen for years while working with football data. A report about a player's injury can be misread as a transfer report. A piece about a coach's private life can be pulled into a tactical category. An announcement about a youth team's fixture list can be promoted into a first-team story. These errors are usually harmless. They distort a few indices, skew a few models, blur a few judgements.

But here, when the subject of the error is a person who has died, the line between data noise and disrespect disappears.

What the system cannot read

Every number tells a story. The story is not inside the number.

I learned that line after years of analysing football metrics. A player has 1.8 xG across five matches and scores two goals. The number says he is outperforming expectation. It does not say he is running into good spaces, or missing clear chances, or playing in a side that does not create enough, or peaking in a window where fitness and psychology allow him to shine. The story is not inside the number.

The same holds for news classification systems. A "football" label says the system recognized a set of signals. It does not say the signals are correct. It does not say that underneath the label is a human being. It does not say that the name in the file belongs to a 21-year-old student with a family, classmates, and a future.

There is a structural gap between data and truth. Professional data practitioners must know how to stand inside that gap and acknowledge it, rather than filling it with guesswork.

In the case of Joselyn Sandoval Calderón, what sits in that gap is far too large to ignore. Because the unresolved question is not a professional one. It is a basic one: what happened to a young person.

And one more detail must be said clearly. The threshold at which the report stops — no cause, no suspect, no hypothesis — is not superficiality. It is compliance. In an open investigation, withholding conclusions is responsible conduct. It prevents two things: unsupported speculation, and further harm to the living.

I have written many times that data does not make revolutions. It only strips the paint off myth. But that line has a limit, and the limit is now in front of me.

Data cannot always strip the paint off myth. Sometimes it strips only itself, and leaves a naked question without an answer.

Contrarian angle: the wrong balance

Here, the counter-intuitive element is not in the story. It is in how the sports news industry operates around the story.

When I watch a report like this circulate, I often ask myself one question: which two kinds of error is our system balancing between?

The first kind is omission. A report is not labelled properly, not delivered to the right readers, not placed in the right analytical stream. In sports, omission means losing a signal, losing a market movement, losing a piece of a larger picture.

The second kind is mislabelling. A report is placed in a topic that does not belong to it. In sports, mislabelling means your model is polluted, your metrics are skewed, and your readers are misled.

Modern news systems, driven by speed and by the demand for broad coverage, typically optimize to avoid the first error and accept the second as a cost of doing business. The operating principle is: better to label broadly than to miss. If an item has a 20 per cent chance of being sports-related, take it. If it is not, it will drift away.

But that logic only works when the things under the label are events that can drift away. A match result can drift. A transfer rumour can drift. A squad list can drift. A death cannot drift. It was not designed to drift.

This is the point my industry tends to avoid: we build systems on assumptions about the subject matter, and those assumptions are not always correct. We assume everything passing through the pipeline is sports data, that it can be processed, dissected, converted into takes. But there are types of content where processing them is itself the error.

The second counter-intuitive point concerns public expectation. When this story appeared, the natural reaction of most people was to seek more information. Has anyone been arrested. What was the cause. Was this a gender-based attack. In the broader context of the State of Mexico and of Mexico as a whole, where violence against women and girls is a seriously discussed social problem, that expectation only grows.

But great expectation does not create truth. It only creates pressure. And pressure, uncontrolled, pushes every party — investigators, media, public — toward worse choices.

The FGJEM prosecutor's office must work at the pace of forensics, not the pace of media. The media must choose between satisfying the demand for information and respecting the limits of an open investigation. The family must endure loss and attention at the same time.

These three parties do not share the same interests. And within that structure, a report labelled "football" drifting into an entirely different stream of interest is a miniature picture of a larger problem: we process human stories with tools designed for events.

Why this matters to the sports data industry

Some will argue this is just a small error in a small system, not worth discussing. I disagree.

The first reason is data propagation. In modern data architecture, a wrong label does not die in place. It is copied. It is stored. It is fed into training models. It is referenced by other systems. One error at source can multiply into thousands downstream. If an automated pipeline learns that "UAEMéx Valle de Teotihuacán" is a sports entity, then for months afterward, every report containing that entity may be pulled in wrongly.

The second reason is professional ethics. The sports data industry is built on an implicit belief: that data is neutral, that numbers are objective, that models do not judge. That belief has value, but it can also become an alibi. When a system processes a death as it processes a match result, the alibi "we only do data" no longer holds.

The third, and perhaps most important reason, is about the limits of professionalization. We have built a sports data analytics industry with standards, processes, and quality measures. But we have not built strong enough rules for identifying content outside the scope. A good analyst knows when a sample is too small to conclude from. A good analyst should also know when content is too sensitive to process.

Data does not erase emotion. It explains why emotion exists. In this case, emotion appears because there is a person. And explaining it, right now, simply means acknowledging that the system was wrong, and that the error must be recorded transparently.

What cannot yet be said

I want to use this section to be explicit about what this article does not do.

This article does not offer a hypothesis about the cause of Joselyn Sandoval Calderón's death. There is no basis for one. The source report held the line, and I hold it too.

This article does not speculate about suspects. No suspect is named in any source reviewed.

This article does not assert a link to gender-based violence. It is a possibility that the broader context raises, but it has not been confirmed by any information in this case. It matters to be clear: context is not evidence.

This article also does not turn the story into material for tactical analysis, predictions, or any other content within my professional field. The temptation is large. When you work with football data daily, everything can become a variable. But a death is not a variable.

What this article does is record an event, record the state of the investigation as of this moment, and raise a professional problem that the event exposes.

And one more thing must be said clearly. One of the principles I follow when writing about sport is not to sanctify anyone, and not to turn anyone into a number. But that same principle requires me to remember that every number begins with a person. Joselyn Sandoval Calderón was a 21-year-old student. She had a name, a school, and a family who had to perform a formal identification. Those are facts. And they deserve to be recorded as facts, not as administrative data points.

A forward-looking thought

There is one question I will carry with me in the coming weeks, when I return to analysing matches and numbers.

It is this: if our industry can build systems sophisticated enough to measure a player's movement across one square metre and half a second, why have we not built systems decent enough to recognize when a piece of content does not belong where it is?

The distance between those two things is what the "football" label in my file is pointing at. We are better than ever at reading the world, and still clumsy at deciding when to stop reading.

When 53,000 spectators fall silent, the numbers begin to speak. But there is another kind of silence, not in any stadium, but inside our systems. It is the silence when an algorithm decides that a death is a data entry. That silence generates no profit, no views, no argument. It generates only a question.

And that question, in the end, is not only for sports data analysts. It is for every industry using data to retell the world. What happens to the content we cannot, should not, and must not process? The answer, if one exists, is probably: leave it alone, call it by its right name, and hand it over to the log of a human investigation, where the FGJEM prosecutor's office, the family, and the university community must work together.

The rest, for now, does not belong to me. And I will not pretend that it does.