The Empty Spreadsheet and the Lesson of Honesty in Football Analysis

On the night of June 27, 2026, in Kazan, my spreadsheet for the Germany versus South Korea match lit up with a number that made me certain of the outcome: Germany at 1.9 xG, South Korea at just 0.4. The model gave the reigning champions a 78% chance of winning. When the final whistle blew, the score was 0-2. My spreadsheet became a pile of meaningless figures, and I sat in silence in my rented room, watching all 64 matches replay in my head as a chain of repeated miscalculations.

What troubled me was not the failed prediction. In this profession, being wrong is routine. What troubled me was how close I came to fabricating a story to defend my data. When my editor asked why the model had missed, my first reflex was to reach for plausible explanations: Germany lacked luck, the referee was biased, South Korea parked the bus. All of them might be true, but none was proven by data. I was doing exactly what I condemned every day: filling a gap with speculation.

Six years later, sitting in Nha Trang, I realise that the memory of that night in Kazan is not a story about a failed model. It is a story about a profession facing its greatest temptation: the temptation to speak even when there is nothing worth saying.

Context: When football is written with blank cells

Modern football analysis runs on a paradox. Data has never been more abundant. Every match in a major league generates millions of data points: the coordinates of every pass, the running speed of every player, the curve of every shot. But at the same time, the pressure to produce content has never been greater. Newsrooms need articles before the match ends, need verdicts before the data can be processed, need a story before any truth exists to tell.

That gap is where fabrication thrives. When there is no data, a writer has three choices: wait, admit the information is missing, or invent a plausible-sounding story. The first two demand courage and may cost a front-page slot. The third always pays faster.

I have seen this in my own field. Before the 2026 World Cup semi-final, while I was working for a leading data company, the editorial desk demanded an analysis of the France versus Morocco match to air before kick-off. The problem was that our model was still running and Morocco's pressing data was incomplete. My manager suggested I write first and fill in the numbers later. I refused. An hour later, I had the full dataset and discovered what every possession-based model had missed.

That is why I have always believed the honesty of this profession is measured not by the quantity of articles, but by the number of times a writer dares to say: "I do not have enough data to conclude." Yet that boundary is more fragile than it appears. Because even with enough data, a larger question remains: am I asking the right question?

Core: The matches that taught me to read the blanks

In the summer of 2026, at the European Championship, I watched Denmark versus Finland in an unusual state of mind. Christian Eriksen collapsed in the middle of the pitch in the 43rd minute. The match was suspended for more than ninety minutes. When it resumed, Denmark lost 0-1 with a missed penalty of their own in the 74th minute. On my spreadsheet, the better team lost an impossible match.

What I observed in the following days was not in any ordinary dataset. Denmark's passing tempo rose from 4.2 to 5.7 metres per second. They pressed harder, ran more, and over the next five matches their PPDA reached 8.9 — the best in the tournament. Denmark did not defend out of fear — they defended to reclaim their breath. An emotional crisis triggered a physical and tactical response that no dry model could predict.

From that moment, I began to understand a principle: emotion is data. Not the kind measurable in metres per second, but the kind that shapes those numbers. When Eriksen fell, I could no longer view the match as a sequence of technical events. I had to view it as a sequence of human reactions to collective grief.

But that lesson also raised a question I had to answer. If emotion is data, how accurate is that data? Denmark played better after the shock, but was that a rule or merely a beautiful anecdote? I decided to test it by comparing them with ten other teams in the group stage. The result showed they were not a lone exception — the phenomenon of a collective response after a shock recurred across many teams, to varying degrees. Emotion is not a story. Emotion is an observable, cross-checkable variable.

By the 2026 World Cup, I applied the principle more thoroughly. Before the semi-final, every model predicted a France win over Morocco. But when I checked Morocco's pressing data, one figure stood out: their five-second counter-press recovery rate was the highest in the tournament, 11.3 per match. They averaged only 35% possession, yet generated four shots from direct turnovers, against an average of 1.2 for other teams. Numbers never lie, but they are very good at telling half the truth — they told me about a team weak in possession, without telling me about the tournament's strongest team at turning turnovers into chances.

I published the analysis of Morocco's counter-pressing model. When Brazil were eliminated in the quarter-finals, the piece went viral, but it also placed me in an awkward position. The company wanted to adjust the data to make it more readable for a mass audience. I refused, and stated my reason clearly: if we bend data to make the story easier to hear, we are doing exactly what we criticise the old models for — simplifying the truth to please the audience.

What I learned from Morocco goes deeper than predicting one match correctly. It taught me that possession is not a measure of power. A team with 35% possession can own a match by another definition — one in which they force the opponent to play at their rhythm. The transfer market does not buy players — it buys the probability of the future. And a team does not buy possession — it buys the chance to win.

But there was another, colder lesson I only truly understood in 2026. When the Bundesliga returned after the pandemic behind closed doors, I analysed 136 matches. The home win rate dropped from 41% to 29%. Penalties awarded to home teams fell by 37%. The empty stands of 2026 taught me: home advantage is not in the grass, it is in the ears. Crowd noise shapes referees' decisions and, in turn, shapes match outcomes. An invisible variable can be stronger than any technical metric a computer can measure.

I tell these stories not to prove I am good. I tell them to expose something uncomfortable: every time I predict correctly, I must admit the probability was on my side, not that I understand more than anyone else. And every time I predict wrongly, I must admit my model is missing a variable. A wrong model does not mean the data is wrong — it just means I have not read the question correctly.

That is why I have grown increasingly suspicious of rankings built on gut feeling. A ranking that claims Team A is stronger than Team B without publishing its methodology is not analysis — it is a statement dressed in the clothing of data. But my suspicion also turns on myself. When I build a new model, I always ask: if this model is right, what on the pitch will prove it? If nothing on the pitch can prove it, then the model is not an analytical tool but a belief.

During my years working in Vietnam, I learned that local context is a variable European models never fully capture. The pressure of the stands in Southeast Asia is not the same as in Europe. The tempo of regional leagues is not the same as in the Premier League. When I import a model built from European data into a Southeast Asian tournament, I am not merely applying a tool. I am carrying a worldview. And that worldview can be wrong.

There is a concept I borrow from architecture: negative space. In music, the silence between notes matters no less than the notes. In football analysis, the silence between two assertions matters too. An analyst who knows when to stay silent is one who knows when their data is not yet ripe. That silence is not an empty space to be filled. It is part of the composition.

The contrarian angle: When football rewards speed over truth

There is something I rarely hear anyone in the industry say out loud: the football analysis industry is designed to reward speed, not truth. Algorithmic platforms prioritise fresh, plentiful, fast content. A meticulous analysis after three days of research can lose to a heated status line written in thirty seconds. That incentive structure produces a psychological consequence: analysts learn that saying more pays better than saying right.

I do not think this is a conspiracy. I think it is a structure. And structures have no malice; they simply operate. But precisely because they operate unconsciously, they are more dangerous. When a model becomes a trend, people begin to seek confirmation rather than refutation. You do not want your model to be wrong, so you start looking at the match through the model, rather than looking at the model through the match.

What I mean is not that data is useless. What I mean is that data without context is merely a more formal way of presenting speculation. Numbers never lie, but they are very good at telling half the truth. And half a truth, in a single match, is often more dangerous than a distorted truth.

If there is one thing I want everyone in the profession to admit, it is this: the fact that an analyst dares to say "I don't know" is not a sign of weakness. It is a sign of maturity. A model that cannot be wrong is a model that cannot learn. And in football, where every match is an experiment we can never repeat, the ability to learn from error matters more than the ability to predict accurately.

Takeaway: Signals from the next round

I trust process more than inspiration, because process can repeat and inspiration cannot. My process now begins with a question, not a conclusion: what is this data hiding? What story is this match telling that my dataset cannot? If the future of modern football lies in combining models with intuition, then the value of an analyst lies not in owning the best model, but in knowing when to discard it.

The 2026 World Cup taught me one thing: the best data is only ever a map, never the terrain. The map tells me which team is better in probability. The terrain tells me which team actually won. The distance between the two — between map and terrain, between model and match — is where our profession truly lives. And it is also where I, whenever I stand before an empty spreadsheet, must learn to stay silent.