Wrong Label, Wrong Conclusion: When Football Trusts Unverified Data

On the night of 30 June 2026, I sat in a small apartment in Marseille, a dim desk lamp, a pen and a squared notebook. France beat Argentina 4-3 in the World Cup round of sixteen, and I recorded the number that kept me awake: France had 38 percent of the ball but took 14 shots, two more than Argentina. Mbappé made six acceleration bursts on counterattacks, covering 312 metres in total. I wrote 4,000 words in one sitting about Didier Deschamps' 4-1-4-1 block, the way he baited the press and then exploded down the flanks. The post went up on my personal blog and hit 12,000 reads in 48 hours, twenty times my average.

Back then I believed numbers did not lie. Seven years later, after reading through thousands of data tables, I understood something else. A number is only honest when the label attached to it is honest, and most of the mistakes in this trade happen because we name the thing wrongly from the start, not because we calculate badly.

In modern football, the analysis department of a mid-table European club can receive tens of thousands of data points per match. Providers such as Opta, StatsBomb or SkillCorner resell event data and tracking data to clubs, to journalists, to bloggers like me. Brentford and Brighton emerged as two cases of using data to buy cheap players and sell them dear. The story is beautiful, and because it is beautiful people forget something banal: data is only as good as the labelling stage. Football is a game of chess with pawns that can run, but before analysing the moves you have to know which piece is which.

I started noticing this through a small detail. A centre-back I was tracking in the French league played a few matches as a defensive midfielder because of injuries. At once, in the scouting database, his position label flipped to "defensive midfielder". From then on, all his metrics were read through the wrong lens: people praised his tackling, criticised his turning speed, and forgot he was a centre-back with a habit of man-marking. One team sheet, and a whole player profile veers off course. That is not the player's fault. It is the label's fault.

Public platforms such as FBref or Transfermarkt admit that their position labels are coarse. A player can only be tagged "forward", "midfielder" or "defender", when in reality he plays left of a back three, drops deep out of possession and tucks inside in possession. I make a habit of watching a player for at least four matches before writing, and I keep a separate note for "real position" apart from "listed position". The gap between those two notes is where real football lives.

The problem does not stop at positions. It spreads to the definitions of the metrics everyone assumes are objective. Take "sprint". Data providers define a sprint with different speed thresholds, usually around 25 km/h, but each sets a different number and a different duration. A player who tops the sprint chart at one provider can fall mid-table at another, even though both watched the same match. PPDA, the passes allowed per defensive action, depends on where one counts a defensive action from. Tracking data does not say who is right; it says who showed up at the right moment, under the definition someone chose.

The same issue surfaces with expected goals, which analysts call xG. xG measures chance quality from shot location and situation, but it does not know who is standing in front of the shooter. A striker who scores many tap-ins will have a lower xG than reality if the model cannot account for a pass that tore the defence apart. Another striker who shoots from distance constantly will have low xG and be called a "poor finisher", when his role is to stretch the defensive block. I once wrote wrongly about such a player and needed two seasons to undo the "inefficient" label I had stuck on him.

Then there are sources. This is where football and every branch of journalism are frighteningly alike. A transfer rumour spreads on the strength of one post with no source. No club name, no agent, no timestamp. Source: unspecified. Yet after a few shares it becomes a "leak", then "almost certain", then a headline. I set myself a rule: never write a transfer story unless I can trace at least two independent sources, or one source with a long enough record of accuracy to verify. A "reliable source" label stuck on the wrong thing is enough to ruin a whole transfer window in the reader's eyes.

Injury data is worse. The line "the player will return in three weeks" that clubs publish is almost always a guess presented as fact. Some players are listed with a "minor knock" and miss two months; others are listed "long-term" and start within ten days. Why? Because diagnosis and announcement are two different jobs, and the label is written to serve the media, not medicine. When analysing a team's fitness I learned to ignore the label and count the actual minutes on the pitch.

There is one area where I am especially sensitive, and I state my view through the examples I choose. Young players are often pushed into adult rhythm before their bodies have matured. The physical data of a seventeen-year-old in youth football cannot transfer directly to senior football, because the opponents differ, the space differs, and so does recovery rhythm. In the Asian leagues I follow fairly closely, more than a few young midfielders are given exploding minutes after a few good games, then break down at the most important moment of their careers. In South America, Brazilian clubs sell eighteen-year-olds to Europe labelled "ready to start", when he has only played ninety minutes a handful of times. The "ready talent" label is applied too early, and that label is worth a whole contract.

I once learned to mislabel in the other direction, and it was a lesson I will not forget. In 2026, before the Euro final, I spent a whole week analysing Roberto Mancini's Italy. I counted 612 passes in their semi-final against Spain, 23 of them line-breaking into the final third. At first I called them a "possession team". But on review I realised the label was wrong. Italy did not try to keep the ball for long; they kept it at the right moment. The way they tucked a full-back inside to build a 3-2-4-1 in possession, then collapsed into 4-1-4-1 out of it, showed a team playing by situation rather than by player position. I had to rewrite the headline. Mancini's Italy did not own the ball; they owned the moment.

That lesson took me to what I believe is the root of everything. The error is not dirty data. Dirty data can be fixed. The error is building a beautiful analytical frame and then forcing it onto a subject it has nothing to do with. Every analyst has fallen into this trap, myself included. We love our frame so much that we see it everywhere, and when the data does not fit, we bend the data instead of admitting the frame is in the wrong place.

I will tell a true story, because it concerns my trade directly. Once I received a data file labelled "football". I opened it and read it carefully. Inside there was no team, no player, no coach, no match. It was a report about a political press conference, with figures who have nothing to do with a ball. The label said one thing; the content said another. Had I been sloppy, I could have "analysed" a few very impressive tactical conclusions out of nothing. But doing so would be fabrication. And fabrication, in this trade, is the heaviest sin.

So I chose another path: I marked the whole thing "insufficient information to assess", and recorded why. It sounds banal, but this is the hardest skill of an analyst. Everyone wants a conclusion. Everyone wants a firm closing line. Nobody wants to publish a gap, because it looks like failure. But the ability to say "I do not have enough data to conclude" is exactly the boundary between analysis and arbitrary interpretation.

I set myself a scale, and I advise young writers to do the same. A metric counts as evidence only when I can answer three questions: by what definition was it measured, who measured it, and across how many matches. A trend counts as a conclusion only when the sample is long enough and there is at least one counter-example I can explain. Otherwise, I call it a hypothesis, and I write the word "hypothesis" plainly in the piece. Readers have the right to know where I am certain and where I am guessing.

There is another temptation, subtler, that I once fell into: the temptation to be contrarian for its own sake. When you read a lot of data and feel you understand the system better than others, you easily want to take the opposite view just to prove you are different. I wrote a few pieces like that, and looking back, they were contrarian because I wanted to be contrarian, not because the data forced me to be. Since then, before writing a contrarian piece, I ask myself: if this view aligned with the crowd, would I still want to write it? If the answer is no, that is not analysis. That is ego.

There is also a temptation about origin. I live and work in France, so every tactical comparison I make is very likely to take French football as the standard. That is a hidden label, and it is dangerous because it is invisible. I force myself to include in every piece at least one data point from a league outside Europe. A J.League match, a Brasileirão round, an Asian Cup tournament. Not to fill space, but to remind myself that football has many standards, and the one I am used to is only one of them.

On the night France beat Argentina 4-3, I wrote a line I still believe: France 4-3 Argentina was the day organised chaos beat talented disorganisation. Looking back, the line is right tactically, but it is also a label I stuck on a complex match. Argentina were not naive in their disorganisation; they lost structure in specific moments, and France punished exactly those moments. If I only kept the "disorganised" label, I would never see the most interesting thing about that match.

So what is to be done? I do not believe in a perfect path. No data is perfectly clean, and no analytical frame is right for every subject. What I believe in is a process: verify before concluding, separate definition from number, state source and date, distinguish hypothesis from conclusion, and have the courage to publish a gap when the gap is the truth. It sounds slow. But football, by its nature, always rewards patience and always punishes labels stuck on in a hurry.

Tracking data does not say who is right; it says who showed up at the right moment. I keep that line as a reminder. Because, in the end, my job is not to decide who is right, but to reconstruct honestly what happened on the pitch, even when what happened is not enough for me to tell a tidy story.

Tonight I am again sitting with a fresh data table, for a match in a league I have never watched closely. Perhaps this match will teach me a new label to peel off. But before I write a single line, I will ask the question I ask every time: does the label I am about to attach actually describe it, or does it only describe what I want to see?