When a Football File Contains Not a Single Player: Notes on a Misclassification Case

When a Football File Contains Not a Single Player

04:12 in Munich. A record lands in my archive, labelled clearly: football. Sixteen information points. I read all sixteen, following a reflex that has hardened into discipline — read the whole thing before believing any part of it.

Not one team. Not one player. No coach, no contract, no academy, no league table, no governing body. Instead: a comedian, a fictional television character, an actress, a film premiere, and a question about a corporate merger thrown out on a red carpet.

The label says football. The contents do not contain a single gram of football.

I am writing this case up not because it is dramatic. I am writing it because it is the kind of error I have made myself, with one difference: my error was never in the label. It was in failing to force a file to prove its own identity before letting it pass.

A label is not evidence

In my trade everything begins with a label: U17 Bundesliga, Regionalliga, national youth league, European cup. Labels file documents into the right drawer. But a label has never been evidence of content. A forty-page dossier can open with "central midfielder, sixteen years old, 78% passing accuracy" and close with a player who has never once played that position across a full season.

Football's data pipelines now move far faster than people do. A record is tagged at collection, pushed to extraction, pushed to synthesis, and every stage trusts the label of the stage before it. Nobody re-asks the simplest question available: does this record contain any entity that actually belongs to the field it claims?

When that question goes unasked, the system does not collapse in front of you. It quietly produces something worse: copy that reads plausibly, uses the correct vocabulary, cites in the correct format, and is entirely invented.

Entity audit: sixteen points, not one name

I ran the check I run on every record before it enters the archive: list every named entity, then test each one against the field the record claims for itself.

The result: one comedian, one fictional television character, one actress, two media conglomerates, one magazine, one film. Not a single entity belonging to football. A match rate of zero across the whole sample, not part of it.

With a scouting file I always accept a margin of error. With a file whose match rate is zero, there is nothing to analyse. Every tactical conclusion, every financial judgement, every risk assessment in this case would be a product of imagination rather than of the record.

And here is where I have to be blunt with myself: the trade trained me into a filling-in machine. Hand me a template with nine boxes and I will tend to fill all nine, because an empty report looks more like a professional failure than a conclusion.

My career does not collapse because of an empty report. It collapses because of an invented one.

The trap of filling the gaps

I ran a small experiment, and I would recommend it to anyone who synthesises data for a living. Take those same sixteen information points, keep the football label, and see what an automated writing system produces.

A typical output: a media merger becomes broadcast-rights analysis; broadcast rights become club revenue; club revenue becomes transfer capacity; and finally a club that does not exist receives a transfer plan that does not exist. Each step is reasonable. The chain is baseless.

That is the mechanism of the error, and it requires no malice. It requires only a ready template and a source too thin to fill it.

I have seen exactly this mechanism in youth scouting, with real people. A fifteen-year-old scores four goals in a friendly, and within forty-eight hours a three-page dossier describes him as a complete number ten. Nobody checks who the opponent was. Nobody checks how many minutes he has actually played in that role across a season.

Every sediment layer tells a story, if we are willing to dig. In this case the first layer already said it plainly: there is no football here. The only remaining task is to listen.

The cost of a record sent down the wrong road

If this record travels onward into an automated summary product, the result lands at one of two poles: empty, or fabricated. Both do damage, but different damage. An empty output costs the reader's trust in the system. A fabricated output persuades the reader of something that does not exist — and a wrong belief is far harder to repair than a missing one.

After every large data audit I still do something younger colleagues find odd: I count the debris. After the storm, I count the forty-seven fragments again. They reassemble into a picture nobody wants to see. Counting is not self-punishment; it is how I know exactly what was lost and at which stage.

The heat cycle of a story, not of a season

One part of my method transfers to this case without bending the data: tracking the life cycle of a media narrative.

A television character debuts in 2026. A film arrives in 2026. Then a long silence. Then a return announcement, an essay written by the character himself, a premiere with a provocative remark, and a release date: 23 October 2026.

Mapped onto the four familiar phases — emergence, acceleration, peak, backlash — this record sits in acceleration. The peak most likely lands in release week, when named critics publish rather than anonymous commenters.

And here I apply the scouting lesson directly: compare sample size with the size of the claim. One premiere. One remark. Against twenty-eight years of the character's history. The controversy sample is small, self-selected, and unattributed. Not enough to call a backlash. Enough only to call a noise spike.

In football I call this a goal from a single phase of play: beautiful, real, and silent about the season.

Source tiering: who speaks, and why

I still grade sources in three tiers, and the grading works for transfer news as well as cultural news.

Tier one is a primary source with an identity: an essay written by the artist himself, published in a major magazine. Clear content, coherent argument. But it is advocacy, not neutral reporting. The author has a direct interest in readers believing his thesis.

Tier two is general-press aggregation: release date, biography, launch context. Reliable on facts, thin on analysis.

Tier three is phrasing like "some viewers feel" and "others say". No names, no outlets, no quotes. This is the weakest link, and it is also the only link representing the critical side.

Does that structure look familiar? It should. Transfer reporting runs on it weekly: on one side a club or an agent speaking on the record, on the other "reportedly" and "according to some sources". The named side always beats the unnamed side, not because it is more accurate but because it is easier to quote.

My own experience tracking many transfer windows offers a crude observation: most "some sources" claims I have logged were never confirmed, and were never retracted either. They simply vanished, leaving a thin, unverifiable sediment layer sitting permanently in the archive.

An unfalsifiable argument

One detail in this case deserves to be kept as a professional lesson.

The central claim of the essay is that being out of date is the point. Criticise it as dated and that proves it still bites. Criticise it as tasteless and that proves it still lands somewhere.

On the surface this is a strong argument. Look closer and it is a closed one: any counter-evidence can be reinterpreted as confirmation. No data can make it false.

In scouting I meet the same structure in the sentence "this boy has not broken through because he has not been placed correctly". It sounds reasonable. It cannot be wrong, and because it cannot be wrong it has no predictive value. A good scouting judgement must state the conditions under which it would be judged wrong.

If I say a player will not exceed 30 km/h maximum speed over the next two seasons, that is a judgement. If I say he has not reached that speed because he has not met the right opponent, that is a story.

The counter-intuitive angle: the fault is not the label

The first instinct on finding a mislabelled record is to fix the label. Fix it and the problem disappears.

I do not think so. The label is a symptom. The disease sits in a pipeline that cannot refuse.

A system that only classifies and never refuses will always find a drawer for whatever falls into it. And every time it finds a drawer, it produces a record that looks legitimate. Fixing one bad label cures one case. Adding one gate — requiring at least one in-domain entity before a record is accepted — cures a whole class of error.

There is a deeper layer, and it is the part I find most uncomfortable when auditing myself.

A report with forty lines of "insufficient information to conclude" looks rigorous. It proves the writer is honest. But it can also become a shelter: a writer can hide ignorance behind a wall of blank lines, and nobody can verify whether the information was genuinely absent or simply never read.

I have fallen into both sides of that trap. Side one is inventing to fill the template. Side two is staying blank to stay safe. Both are avoidance; only the form differs.

There is a parallel I cannot unsee. The academies of the biggest clubs are talent warehouses by design: each intake gathers dozens of teenagers, and fewer than 10% genuinely have a path to the first team. The rest are not rejected — they are retained, tagged "prospect", and gradually disappear from every list. A system that cannot refuse accumulates volume rather than quality. Data pipelines run on exactly that logic.

The Werner lesson: which number did I trust wrongly

In 2026 I assessed a sixteen-year-old midfielder named Lukas Werner at the Bayern Munich academy. The file recorded 78% passing accuracy in the U17 Bundesliga. GPS data recorded a maximum speed of 28 km/h, below the squad average. I held my position and declined to recommend promotion to the U19 side. The boy moved to the RB Leipzig academy that same summer.

I was wrong, and I have accepted it. But how I was wrong is the part worth telling: I trusted the 78% because it carried the right label — right league, right age group, right position — and I doubted the 28 km/h because it came from a new system that had not been validated. I let the label set the confidence level, exactly as the pipeline is doing with today's record.

After the 2026 World Cup I compared the 19 players under twenty who started knockout matches and found that 14 had been rejected by German academies on physical grounds. I wrote a forty-page internal memo admitting the limits of the old evaluation method, and since then every report of mine must carry at least three independently verifiable indicators.

Today's record is the second examination of the same lesson, at a different layer. I do not trust my eyes; I trust what the file leaves behind. But a file can also lie about itself. And when a file lies about itself, the person reading the file is the last line of responsibility.

What I want to know before next season

I will keep this record in the archive, but I will change the label and record the reason. I will not delete it, because a logged error is worth more than a cleaned-up one.

Old files never die; they wait for someone patient enough to read them again. I still believe that. But I have to add a clause: the reader must be patient enough to recognise that sometimes both the file and the person who wrote it are deceiving themselves.

And I wonder how many dossiers in my own archive carry a football label without ever having been forced to prove that there is football inside.