When the Spreadsheet Is Empty: The Art of Saying 'I Don't Know' in the Age of Football Data

There was a November evening when I sat in front of my screen in a small apartment in Liverpool, and on the screen was an analytical report with nine sections, each ending with the same line: "Insufficient information to assess." I laughed. Not at the wording, but at the strange coincidence with how I have worked for thirty-five years. Nine tables, nine empty cells, all at once, in a document longer than two thousand words. I read it three times, not to find answers, but to find the reason why a report could say so much while saying nothing.

For someone who works in the transfer market, an empty report is the strangest assignment possible. My first instinct is always to fill in the blank — a metric, a match, a name. But this time, the blank was the content. And I realised it might be time to write about something I have never written about: the silence of data.

Data whispers, and those who know how to listen will hear miracles. But what happens when data does not whisper at all?

In 2026, when I was a twenty-seven-year-old rookie joining the sports desk of Belgrade Television, I learned the first lesson of the trade: one only prints what can be verified. What cannot be verified does not go on air. I remember the old editor-in-chief saying something I have carried with me ever since: "If you don't know, write that you don't know. That is the only truth you have the right to print."

Thirty-five years later, that sentence returned to me in a completely different context. Football in 2026 has become an industry where every match generates millions of data points. Cameras track every step a player takes. Sensors inside the ball record the force of every shot. Algorithms calculate xG for each attempt, xA for each pass, PPDA for each pressing sequence. Nothing escapes the data net. And precisely because of this, we have forgotten something: most of what matters most in football cannot be measured by machines.

I remember a conversation with a Premier League analyst last season. He told me his club has an internal dataset tracking "ball circulation speed" — the average time for a pass to travel from one player's foot to another's. That metric appears on no public statistics site. And he told me: "Sometimes we discover something. But most of the time, we discover that we know nothing at all."

That is the kind of discovery no one wants to publish. Because in the modern football industry, "I don't know" is an expensive confession. A sporting director who says "I am not sure about this player" will lose his job faster than one who makes a bad signing. A manager who says "I don't know why we lost" will be torn apart by the press. An analyst like me who says "I do not have enough data to conclude" will be dismissed as useless.

But what is the truth? Every transfer window, hundreds of millions of euros are spent on the basis of data tables whose own authors are not certain of them. I know this because I have worked in those rooms. I have seen forty-page reports on a twenty-year-old player, filled with charts, heat maps, and development projection models. And I have seen those same reports thrown in the bin six months later when the player failed to prove anything.

Every number in the transfer ledger is a fate waiting to be written. But that fate is not to be found in the spreadsheet.

Let us begin with a concrete example I have spent years studying: Sam Allardyce at Bolton Wanderers from 2026 to 2026. In seven seasons, Allardyce took a small club into the upper half of the Premier League and reached a League Cup final. Bolton's basic metrics at the time were unremarkable: an average xG of 1.2 goals per match, possession below 45%, and pass completion below league average. And yet they won.

When I used my own xG model to dissect Bolton, I found something strange. Bolton's expected goals were exceeded by their actual goals in almost every season — meaning they scored more than the model predicted. In data, that is usually a sign of luck, of randomness that will soon vanish. But that luck lasted seven years. Seven years cannot be called luck.

It took me nearly a decade, and a re-education in certain concepts, to understand that my model was missing a variable: the ability to exploit set pieces. Allardyce's Bolton scored between 35 and 40 percent of their goals from dead-ball situations — corners, free kicks, long throws. That figure was nearly double the Premier League average at the time. But the earliest xG models did not fully account for this variable, because they were built on data from big clubs that played short, controlled football. Those models could not see Bolton, because Bolton did not fit the template they were taught.

This is the crucial point I want to stress: every data model carries the bias of the person who built it. When you build a model on data from Manchester City, Barcelona, and Bayern Munich, you are teaching a machine that beautiful football is short and controlled. Then you take that model to evaluate a long-ball team and conclude that the team "plays inefficiently." But the team is not inefficient. The model is.

In 2026, I wrote a long analytical piece on Juergen Klopp's concept of gegenpressing when Liverpool finished the season with 78 points. I calculated Liverpool's average PPDA at 8.2 — the lowest in the league — while Jose Mourinho's Manchester United sat at 15.7. I posted the piece on my personal blog and was heavily criticised for being "too mechanical." People said I reduced football to an Excel sheet. People said I understood nothing about the emotion of the game.

I remember sitting for a long time in front of those comments, wondering whether they were right. Then, on 19 January 2026, Liverpool beat Manchester City 4-3 in a match where their PPDA dropped below 7. I sat in my apartment in Liverpool, rewatched the match ten times, and realised that data and emotion are not opposites. My emotion watching that match — the breathlessness, the overwhelming intensity — was exactly what the PPDA metric was trying to measure in its own language.

I have stood before a table of numbers and felt as though I was witnessing a miracle at Anfield. But the miracle was not in the numbers. It was in the gap between the numbers and what I saw with my own eyes.

In 2026, I analysed all 64 matches of the World Cup in Russia using my own xG model and predicted France would win from the group stage, because their chance-creation metrics were the highest — an average of 2.4 xG per match. My article was mocked for claiming Croatia had "low xG but luck." When Croatia reached the final, I was emotionally exhausted. I hid in the Liverpool library for two weeks to review all the data. And in the end I discovered that my model ignored corners — a serious error. Croatia scored nearly a third of their goals from set pieces, and my model had not accounted for that.

That was the most painful lesson of my career as a data analyst. Not because I predicted wrong, but because I predicted wrong with confidence. That confidence is the enemy of accuracy.

In March 2026, the entire football world stopped for the COVID pandemic. Liverpool were 25 points clear of Manchester City, almost certain to win the Premier League, but the season was suspended. I lost my faith profoundly. If data could not predict a pandemic, what did data mean? I wrote three drafts and deleted them all. When football returned to empty stadiums in June, I realised that "no-crowd" data changed the metrics entirely: the home win rate fell from 46 percent to 39 percent. I had to sit alone for weeks to redefine my model.

An empty stadium does not distort the data, but it makes the truth feel empty. And I learned that every number needs a context in order to breathe.

There is one mistake I see repeated throughout modern football analytics: confusing correlation with causation. When we see a team with high xG win many games, we hastily conclude that high xG is the cause of winning. But what if both are the result of a third cause — such as squad quality?

This is the kind of question few analysts want to ask, because the answer could dismantle the entire football data industry as we know it. If the metrics merely reflect squad quality rather than predict outcomes, where does their value lie?

Look at the transfer market. Over the past decade, clubs have spent billions of euros on young players based on data analysis. But look more closely: how many of those expensive transfers actually delivered proportionate value? According to my research on publicly available data from Europe's five major leagues between 2026 and 2026, more than half of the players bought for over 50 million euros failed to deliver contribution matching their fee. More than half. That is the naked gamble that no data table dares to print.

xG is a revolution, but every revolution needs time before people accept it. And perhaps the football data revolution is now at the stage where we must learn to acknowledge what it cannot do.

I remember a conversation in July 2026, when I was writing a series on the Euros. I happened to connect with an Italian tactical analyst on social media. He shared internal training data from the Italy squad: they ran an average of 112 km per match — not the highest in the tournament — but their "ball circulation speed" metric was exceptional. Through that connection, I gained access to unpublished data and wrote a piece arguing that the Italians were not a defensive team — they were a movement machine. The article was shared more than 10,000 times.

But what I remember most from that conversation was not the data. It was what my Italian colleague said when I asked about his prediction model: "Our data tells us how far the team runs. It does not tell us what the team believes in."

I learned at Anfield that belief is also a variable. And it is the only variable no camera can capture.

Now let us return to VAR, a subject on which I have always held a different view from the mainstream. Many believe VAR was designed to eliminate refereeing errors. But look at the data from the past three Premier League seasons and you will see something else. The number of VAR interventions in penalty-area incidents has not decreased over time — it has merely shifted. Decisions have not become clearer; they have become more complex, more recorded, and therefore more contested.

The clause "clear and obvious error" sounds like an objective standard. But "clear and obvious" is a subjective judgement packaged in legal language. A VAR referee looks at the same frame as ten million viewers and reaches a different decision. Not because he is blind, but because he is being asked to answer a question that has no single answer: when does a contact become a foul?

I once spent a full season logging every VAR decision in the Premier League and classifying them. The result did not surprise me: the same type of contact, in the same position, at the same speed, could lead to two entirely different outcomes. The difference was not in the incident. It was in the person sitting in front of the monitor.

This is why I believe data cannot solve every problem in football. We can know exactly where the ball went, at what speed, with what force. But we cannot measure intent. And football is a game of intent before it is a game of movement.

Let me return once more to those nine empty cells in that report. They are not the failure of the writer. They are another form of data: data about the limits of knowledge itself. A report that says "I know this" has value equal to the information it contains. A report that says "I do not know this" has value equal to its honesty — and in an industry full of overconfidence, honesty may be the scarcest data of all.

Vietnamese football, where I was born and which I carry in my heart every day in Liverpool, is at the early stage of its data revolution. V.League clubs are beginning to hire analysts. Academies are beginning to use GPS to track young players. But I hope for one thing in that future: that we do not repeat Europe's mistake — that is, believing data can replace intuition, that models can replace the eye, that numbers can replace stories.

Because after thirty-five years, after thousands of spreadsheets, after tens of thousands of hours reviewing footage, I still learn at Anfield that belief is also a variable. And I have still not found a way to include that variable in my model.

In a world of long seasons, the awakened can only rely on their own spreadsheet. But I have learned that sometimes the awakened must also sit quietly before a blank page and accept that silence is also an answer.

So what is the signal for the next round? I think it lies in a place few are looking: the ability of analysts to say "I don't know."

In an industry that runs on certainty — certain about this player, certain about that tactic, certain about the result of the match — the ability to admit uncertainty is a form of intellectual courage. It is not as appealing as a bold prediction, not as eye-catching as a decisive statement. But it is honest. And in an industry where every mistake costs clubs tens of millions of euros and costs players their careers, honesty is not a virtue — it is a requirement.

I returned to the nine-section report last weekend. I printed it out, placed it on the table, and read it one last time. Then I wrote in the margin, in pencil: "Insufficient data to conclude." That is perhaps the most honest sentence I have written in years.

Those who are right before their time always pay the price of solitude. But that solitude is not a punishment. It is the price of a conversation that has no interlocutor yet. And I still believe that one day, someone will hear the whisper of data — even when data leaves nothing behind but blank pages.