When the System Called El Chapo a Footballer: Wrong Labels and How We Misread a Match

On the hard drive in my Hamburg office sits a text file labelled "football" on its very first line. I opened it one winter morning, and inside there was not a single player.

No club. No stadium. No scoreline. What lay inside was a crime report: Joaquín "El Chapo" Guzmán, the Sinaloa cartel, a kidnapping in Puerto Vallarta on 15 August 2026, and the recent testimony of Renato Sales Heredia on a podcast. Sixteen information points, and not one of them touches a ball.

The file was not wrong. The label stuck to it was.

I have followed football for nearly four decades, writing analysis from the days of forums to the nights spent drawing diagrams with RB Leipzig's GPS data. My daily work is to dissect a match into blocks of space. That morning, the thing I had to dissect was not a match, but the way a system names something incorrectly.

And I realised this: it is the same work I do every weekend. People misname matches all the time. The difference is that when you misname a match, the cost is rarely as loud as when you misname a crime report.

The label arrives before the eye

There is an experiment anyone who has sat in an analysis room has run without meaning to. You switch on a match, the broadcast graphic flashes "4-4-2". For the first thirty seconds, your eye sees exactly four defenders, four midfielders, two forwards. You see the label, then you see the match agreeing with the label.

But switch the graphic off and watch only the runs, and the story usually changes. The left-back pushes higher than the right winger. The central midfielder drops level with the centre-backs when his team has the ball. The shape on the sheet says one thing; the real structure says another.

A label does not describe a match. It shapes what you allow yourself to see inside the match. This is the problem of every classification system, not just football. Once a file has been tagged "football", whoever reads it will go looking for players. When they find none, they assume the fault is theirs, and never suspect the label.

That is why I began to distrust the very system I use. I do not distrust the data. I distrust the name people give the data before anyone has read it.

Geometry is not on the blueprint

On the night of 21 June 2026, in the World Cup group stage in Russia, Croatia beat Argentina 3-0. I lost three nights of sleep re-watching that match through fourteen different camera angles. On the scoreboard, it was a win. On the tactical sheets, people called Croatia a 4-2-3-1 and Argentina a 4-3-3. But the sheet could not tell the story I found.

Luka Modric received the ball twenty-eight times in a patch of land that appears on no diagram: the zone between Argentina's two pressing lines. I call it the third space. Across ninety minutes, Croatia took seventy-four touches in that zone. Argentina took nine. Nine.

No one in the stands saw that number. No broadcast carried it. And if you only read the "4-3-3" label the television stuck on Argentina, you would go looking for three midfielders, find them standing in exactly three positions, and conclude that Argentina's midfield played badly. You would conclude about people, when the problem lay in the structure of space.

Geometry is not on the blueprint; it lies between the runs. I have written that line over and over for five years, and it holds for files that have been mislabelled too. The "4-3-3" tag and the "football" tag are the same kind of error: they describe surface form, then leave the reader to infer the substance.

When I wrote a 4,200-word piece on Croatia-Argentina, the desk asked me to cut it to 1,800 words. I refused and published it on my own blog. It was shared by a Liverpool scout with the comment: "This is what our coaching staff need to read." I tell that story not to boast, but to show that the "1,800 words" label the desk stuck on my article was also a misclassification: it assumed length decides value, when what decides value is the structure inside.

When an entire pipeline is contaminated

In the 2026-2026 season, I rebuilt a database from 1,240 Bundesliga matches. I wrote code to extract passing data, and found that teams who passed back to their centre-backs under high pressure saw their rate of fatal turnovers rise by 41%. No league table carries that number. It only appeared when I refused to trust the ready-made labels — labels like "possession side", "counter-attacking side".

At the same time, I noticed something that chilled me: those labels were not only wrong in other people's articles. They had contaminated my own pipeline. A defender tagged "midfielder" by the system drags every downstream metric out of true. A match tagged wrongly makes every conclusion drawn from it wrong too. The error does not stop at one line. It flows downward.

That is exactly what happened with the file labelled "football" that contained Sinaloa material. No one checked whether there was a player inside. The label came first, and every processing step afterwards trusted the label.

In football analytics we have very fine tools. Expected goals (xG) measures chance quality. Passes allowed per defensive action (PPDA) measures pressing intensity. Positional tracking draws every run of every player. But all of those tools rest on an assumption nobody checks: that the event recorded is the event that happened.

If a Modric touch is logged to a team-mate by mistake, both players' numbers drift. If a pressing action is tagged "duel" instead of "interception", the PPDA model misreads intensity. These errors are small, but they multiply over time. After 1,240 matches, they form a false picture.

I watched this from the stands

Based on my experience following matches, one moment repeats in almost every round. In the fifteenth minute, the commentator says: "The away side are sitting deep, defending in numbers." I look up, and I see the away side had six players in the opposition half in the previous phase. The "sitting deep" label was stuck on beforehand, and the phase that just happened was ignored because it did not match the label.

I once knew an RB Leipzig analyst. He showed me the team's GPS data. We found that Leipzig's pressing created triangles turning their backs to the opposition goal at an angle of 112 degrees. That figure of 112 degrees had never appeared in the German media. Not because it was secret, but because nobody thought to look for it. People knew the label "high press" and were satisfied with it.

My first piece on Julian Nagelsmann's "wide-attack geometry" ran only 800 words but came with fourteen animated diagrams. It drew 47,000 reads in three days. What I learned was not about writing short or long. What I learned was this: the value of an analysis lies not in its length, but in whether it dares to doubt the label everyone else has accepted.

The counter-intuitive angle: the fault is not in the machine

The first reaction when people see a mislabelled file is to blame the algorithm. Humans are subtle, machines are crude. That reasoning sounds reasonable, but it ignores one important detail: humans wrote the tagging rules. The algorithm did not invent the word "football". It only repeats what humans taught it.

What is notable is that humans make the same error, only nobody checks. In the stands, we label a striker "poor" for missing a chance, without measuring whether that chance was created by good or bad spatial structure. We label a coach "conservative" for a slow side, without counting how often they deliberately draw the opponent up before countering.

An individual error is often the product of bad spatial structure, yet we always hunt for the fault in the individual. That is the most convenient label football media sticks on every match. When a defender loses his man, the report talks about a lapse in concentration. It does not measure whether the defensive line was stretched so far that no covering distance remained.

At this point, every phase of play is a proposition; tactics are the logic of the body. A football situation, like a data file, can only be judged correctly if you accept that the surface label may deceive you. If you believe Argentina played 4-3-3, you will never count nine touches between the lines. If you believe a file labelled "football" is about football, you will go looking for players in a place where there are none.

This is also why I keep a scepticism toward the rumour rankings of the transfer window. People stick the "top target" label on a player from one tweet, and an entire chain of analysis is built on that label. What needs checking is not the rumour, but the structure of the release clause and the wage bill behind it.

What to verify next matchday

When a system calls El Chapo a footballer, the fault is not that people failed to find a ball. The fault is that nobody checked whether there was a ball to find. That is the trap anyone reading a match can fall into, every weekend, quietly.

Next matchday, I will ask myself one question before opening any graphic: when was this label stuck on, by whom, and on what evidence. Because if I do not ask, I will again see a match agreeing with its name — and miss the land lying between the runs that no one sees.