The Empty Cell: When Football Analysis Walks the Line Between Truth and Fabrication

On the night of July 6, 2026, in Nizhny Novgorod, I sat in front of the screen with a notebook, logging every pass. France beat Uruguay two-nil. But what held me was not the scoreline; it was a paradox on the statistics sheet. Didier Deschamps' side controlled just 39 percent of possession, yet generated 2.1 expected goals, while Uruguay, with 61 percent of the ball, generated only 0.4. I underlined that number and circled it three times. For the next three weeks, I rewatched every knockout match and built my own xG table for each team, the exact opposite of the emotional commentary on television.

But there is a detail I rarely tell. My spreadsheet had empty cells. Not because I was lazy. Because the data did not exist. Some matches gave me no touches-in-the-final-third figures, some players gave me no verified minutes, and some passes had two sources reporting two different numbers. It was from those empty cells that I learned the hardest lesson of this craft: the hardest part of analysis is not reading the number, but knowing when there is no number worth reading.

When there is no data, the only honest answer is "insufficient information to assess" — and that is the answer the sports analysis industry rarely dares to give.

Before 2026, I watched football. After 2026, I read it. But reading football does not mean turning every gap into a conclusion. That is the boundary I want to talk about here.

Context: an industry allergic to silence

When I began writing data analysis for the sports market, I thought my job was to collect as many numbers as possible. I downloaded data from FBref, from Understat, from StatsBomb, cross-checked every metric, built spreadsheets thousands of rows long. I thought a good analysis was one with many numbers. I was wrong in one very specific way: a good analysis is not one with many numbers, but one that knows which numbers to trust and which not to.

The modern sports analysis industry operates under a strange pressure. Readers follow every match, every signing, every press conference. They want answers immediately. Media platforms want content published within hours of the final whistle. And within that short window, the data has not yet been verified. That is when a temptation appears: filling the empty cell with speculation.

I have seen this in many forms. An article claims a striker is "in form" based on three goals in four matches, without checking whether those goals were tap-ins from a metre out. Another calls a defence "the worst in the league" based on goals conceded, without separating the goalkeeper factor, the fixture factor, and plain luck. These pieces are not exactly lying. They are simply filling a gap with something that sounds plausible.

The problem is this: a gap in the data is not a gap in the story. The human brain hates emptiness. When we see a blank cell in a table, we tend to fill it with a hypothesis. In football, that hypothesis is usually wrapped in certain language: "clearly", "certainly", "everyone can see". That is the first sign of a conclusion lacking foundation.

Data does not make revolutions. It only strips the paint off legends. But when there is no data, we should not paint a new legend over the gap either.

An empty cell is not a zero

The most common mistake I encounter at work is confusing "no data" with "data equal to zero". These are entirely different things, but in practice they are often merged into one.

Imagine a stats table with a cell for a midfielder's chances created. If that cell is blank, there are three possibilities. First, the player created no chances, so the true value is zero. Second, the data is missing due to a collection error, so the true value is some unknown positive number. Third, that metric is not defined for this position within that particular data system. Three possibilities lead to three different conclusions, yet looking at the table, we see only one blank cell.

In my daily work, I am forced to classify these three cases clearly. With the second case, I must not fill in a zero, because doing so is fabrication. With the third case, I must not compare this player with another, because the metrics are incompatible. And with the first case, I may only draw a conclusion once I have verified the data source is complete.

This may sound extreme. But it is precisely this extremism that protects me from larger errors. A wrong conclusion built on complete data can still be corrected, because there is a traceable basis. A wrong conclusion built on fabricated data cannot be corrected, because it has no root.

I remember once analysing a match in the national championship. I wanted to assess a team's pressing ability. The metric I needed was PPDA, the number of passes a team allows the opponent to make before committing a defensive action. But my data source did not have this metric for that match. I had two options: skip it, or estimate from raw figures. I chose the second, but stated clearly in the piece that this was an estimate with a margin of error. That is the difference between a transparent estimate and a fabricated number wearing the mask of precision.

Every number tells a story. The story is not in the number. And when there is no number, we need to be honest that we are missing a chapter, rather than writing a fake chapter ourselves.

The Chiesa lesson: when a small sample wears a star's mask

In 2026, at 21, I followed the Euros and was captivated by Federico Chiesa. The Italy winger made the crowd rise every time he had the ball. The media called him the tournament's "breakout star". The basis for that claim was simple: two goals and one assist.

I decided to dig deeper. I pulled data from three different sources — FBref, Understat and StatsBomb — and compared them. The result stopped me. Chiesa's xG across five matches was only 1.8, yet he scored two goals. His shot-on-target rate sat at 41 percent, below the average of the top European wingers of that period. In other words, Chiesa scored more than the quality of his chances should have allowed.

That is the signal of an over-performing run relative to its base. In data analysis, when a player scores far above his xG in a small sample, there are two explanations. One is that the player has exceptional finishing skill and will sustain it. The other is that the player is benefiting from luck, and the number will regress to the mean. With a five-match sample, the second explanation is far more likely, because the sample is too small to separate skill from randomness.

I wrote a 2,000-word analysis for my personal blog, arguing that Chiesa's performance was unsustainable and likely to decline as the sample widened. I did not say Chiesa was a bad player. I said the data was not yet sufficient to call him a star at that level. The following season, Chiesa suffered a ligament injury and his form fell. My caution was confirmed, but I did not celebrate. Because I know I could have been wrong, and in that case I would have to answer for a conclusion drawn from a small sample.

Chiesa did not break the data. He broke the way we read the data. The lesson here is not whether Chiesa is good or bad. The lesson is how a small sample can be read as a large trend. Two goals in five matches does not make a star. It makes a question without an answer.

And that is exactly where the analysis industry tends to fail. When we have a small sample, the honest answer is "not enough data". But that answer does not sell. It does not make a headline. It does not make controversy. So it gets replaced by a conclusion more certain than the data allows.

Anfield empty, and noise as a variable

In 2026, when the pandemic left stadiums empty, I was 20 and writing my bachelor's thesis in sociology. Liverpool endured a run of five straight home defeats at Anfield, an unprecedented phenomenon under Jurgen Klopp. The media called it a crisis. I wanted to know where the crisis lay.

I collected data on Liverpool's PPDA. The previous season, the figure stood at 8.2. During the no-crowd period, it rose to 12.5. A lower PPDA means a team presses more aggressively. When PPDA rises from 8.2 to 12.5, it means Liverpool pressed less, giving opponents more time on the ball. Their high defensive line became more fragile once the front line no longer applied effective pressure.

The next question was why PPDA rose. There were several hypotheses: a congested schedule, injuries, individual decline, and the absence of a crowd. I separated each factor. I compared home and away, compared rest periods between matches, compared with the crowd-present phase of the previous season. The results showed a clear correlation with the home factor and the crowd factor, but not enough to conclude causation.

That is the point I want to stress. When 53,000 fans fall silent, the numbers begin to speak. But what the numbers say is one thing, and why they say it is another. The absence of a crowd may reduce the mental pressure on opponents, making them more confident on the ball. It may also reduce the motivation of the home players themselves. These two mechanisms are different but produce the same metric. With observational data, we cannot decisively separate them.

I wrote about this crisis calmly, separating the random factor from the structural one. I did not say Klopp was finished. I said the data showed a change in how the team operated, and that change could have multiple simultaneous causes. Some colleagues said I lacked emotion. I took that as a compliment.

The empty stadium taught me that noise is data. But it also taught me that a variable being measurable does not mean we understand it. The gap between "measurable" and "understood" is where analysis turns into speculation if we are not careful.

The transfer market: where impatience is priced

Nowhere is fabrication in analysis as dangerous as in the transfer market. Here, a wrong number does not just ruin an article. It can skew an investment decision.

I have followed this market for years. What I realised is that transfer value is not priced by a player's true ability, but by the buyer club's degree of impatience. A club in crisis will pay more for the same player than a stable club. That difference is not in the data about the player. It is in the data about the buyer.

The transfer market is where impatience is priced. And when impatience is priced, data analysis is easily bent to serve a decision already made. A club that has decided to sign a player will find the metrics that support that decision. That is confirmation bias, and it appears everywhere in the industry.

There is one type of deal I watch especially closely: free transfers with high signing fees. In this form, the club pays no transfer fee, but pays a large sum to the agent and the player to sign the contract. On paper, it is a free deal. In reality, it is a large outlay recorded under a different heading. This makes it hard to monitor under financial fair play rules, which focus on transfer fees and wage bills.

I am not saying this deal type is illegal. I am saying it creates a grey zone, and the grey zone is where data is easily misread. When a club announces it has signed a player "for free", that is a technical truth but not the whole financial truth. If an analyst reads only the headline, they will conclude the club is saving money. If they read the balance sheet, they will see another story.

This is why I always cross-check at least two data sources before writing about a deal. Not because I distrust everyone. But because I know a single number can be technically correct yet wrong in meaning. And in the transfer market, meaning is what has value.

The trap: correlation is not causation

Here, I want to address the mistake I consider most serious in sports analysis: turning correlation into causation.

When two variables move together, we tend to conclude that one causes the other. But in football, most variables are related indirectly. A team that presses well often scores many goals. But that does not mean pressing causes goals in every case. Both may be the result of a third factor, such as squad quality.

The problem is that the causal story is more attractive than the correlational one. "Pressing creates goals" is a good story. "Pressing and goals appear together in teams with good squads" is a boring story. But the boring story is usually truer.

In my analysis of Liverpool, I was careful not to say the crowd caused the decline. I only said these two phenomena appeared together. This difference in language sounds small, but it decides between analysis and propaganda.

This is also why I am sceptical of every praise that sanctifies. When a player is called "unstoppable", I want to see how many matches he scored in, against which opponents, with what chance quality. When a coach is called a "tactical genius", I want to see whether his results exceed what his squad quality allows. These questions do not diminish a player's or coach's value. They simply place that value in its correct position.

One thing I always remind myself: when I cannot reproduce a conclusion with independent data, that conclusion is not yet ready for the piece. Not because it is wrong. But because I am not yet qualified to assert it is right.

What comes next

There is a paradox in the craft of sports data analysis: the more you understand data, the more you realise you know less. Every empty cell in a table is not a failure. It is a reminder that the truth of a match is always larger than what we can measure.

I do not want this piece to end with a summary. I want it to end with a question I am asking myself. As the sports industry accumulates ever more data, will it learn to say "I don't know" more often, or will it find more sophisticated ways to pretend it knows everything?

Data does not erase emotion. It explains why emotion exists. And in an industry that runs on emotion, the honest analyst is not the one who gives the fastest answer. The honest analyst is the one who knows how to stay silent when the data has not yet permitted a word.

I will keep watching the empty cells in my spreadsheets. Because it is in those empty cells that the boundary between truth and fabrication is drawn.