The Empty Cell and the Trap of Modern Football Analytics

Kazan, June 27, 2026. In the press tribune of Kazan Arena I held a sheet of A4 printed in the hotel business centre. On it was a single data column: Germany's PPDA across the three group matches, averaging 15.2. The number meant that for every active defensive action Joachim Löw's side made, they allowed the opponent fifteen passes first. Underneath, in handwriting, I noted: high defensive line, enormous variance in line height, Son Heung-min enjoys this kind of space.

The veteran agency reporter beside me glanced at it and laughed. Germany only needed one goal, he said, and those trendy metrics were decoration. I did not argue. I left the sheet on my knee and kept taking notes.

In the 93rd minute Kim Young-gwon put the ball in the net. The referee consulted VAR and awarded it. Three minutes later Manuel Neuer abandoned his goal to join the attack, Son Heung-min collected the ball near the halfway line and walked it into an empty net. South Korea won 2-0. Germany left the 2026 World Cup at the group stage for the first time since 2026.

I did not cheer. I folded the sheet and thought about something far more frightening than being right: if that page had been blank, would I have dared to write anything at all?

I have carried that question for eight years. This week it returned in a context that had nothing to do with grass.

My career began with a spreadsheet

In 2026 I was the only female intern at a young sports media company in Seoul. I had been a swimmer, and moved into data journalism once my knee stopped allowing two sessions a day. In my first month I wrote that FC Seoul won the K League with 12 of their 38 goals coming from set pieces, 31.6 percent, against a league average of 18.4 percent. An editor threw the draft back on my desk and said women knew nothing about tactics.

I did not argue. I sat down, re-watched every recording, annotated each dead-ball phase by minute, and attached a methodology appendix explaining the expected-goals concept I had used. The piece ran and caused a debate, because it was the first time the Korean league had seen expected goals in a mainstream analysis. Since then I keep one habit: every article ends with a section on sources and calculation, so that anyone can verify it themselves.

My first battle had no audience. Just me, a spreadsheet and a sinking club.

In the summer of 2026 the stadiums closed. My company lost 70 percent of its revenue and editors left one by one. Those who remained were asked to write about football after the pandemic. I refused. I had no data with which to describe a world that did not yet exist. Instead I built my own database, collecting match data from hundreds of old fixtures that the big platforms had merged or deleted from public view. The whole world stopped turning, but my ghost football database kept breathing.

That ghost database later saved me a transfer window, because real football is not always as real as data.

An analysis that returned zero

This week a colleague forwarded an internal document with a short note: take a look, something seems off. It was labelled a Stage-2 deep professional analysis for the football domain. By design it was built on a Stage-1 deconstruction, the record of an original article's information points, entities, author stance and time sensitivity.

That Stage-1 record was empty. Not partly empty. Completely. Article title: none. Source: none. Article type: unclassified. One-sentence summary: blank. Author stance: not available. Article purpose: not available. Information points: an empty list. Time sensitivity: not assessed. Source quality: unresolvable. The only surviving field was the domain label, and it read one word: football.

Which means the entire second-stage analysis was erected on a void.

What caught my attention was not the failure but how the document handled it. It kept the full nine-dimension frame: tactical and technical, club finance and transfer market, results and public-opinion cycle, league landscape and team positioning, rules and governance, management and dressing room, risk profile, media narrative, and industry transmission. Nine dimensions, nine tables. All nine were filled with the same sentence: insufficient information, cannot assess.

That is the correct answer. It is also the answer almost nobody dares to give.

Why an empty cell is more frightening than a wrong one

I have worked with spreadsheets of thousands of rows. In my trade, a wrong number gets caught. It contradicts another column, contradicts the video, contradicts the memory of the people who watched. A wrong cell is a traceable error. An empty cell cannot be traced, because it says nothing at all. And precisely because it says nothing, it invites someone to fill it.

Take the simplest example. PPDA measures pressing intensity: the passes an opponent is allowed before each active defensive action. Lower means more aggressive. But PPDA alone is meaningless. A team that concedes possession by design and defends in a low block will show a very high PPDA without being lazy. A team that presses ferociously for sixty minutes and drops off for thirty produces a flattering average. Anyone who takes that single number and draws a conclusion about fighting spirit has filled an empty cell.

The same holds for expected goals. An expected-goals model takes shot location, angle, body part, the pass that preceded it, the number of defenders blocking and the goalkeeper facing it. When event data for a match is incomplete, the model still returns a number. It does not raise an error. It returns a value that looks entirely ordinary, and that value travels into an article, into a metric ranking, into the recruitment decision of a small club.

At financial level the stakes are higher. UEFA's financial fair play rules and the Premier League's profitability and sustainability regime can only be judged on audited accounts. Broadcasting revenue, commercial revenue, wage bill, net debt: those four columns decide whether a club loses points. The actual points deductions seen in the Premier League in recent seasons, ranging from four to eight points, all trace back to those four columns. Without them, an analyst can only say the data is insufficient. Anything else is invention.

And at transfer level I hold a firm professional position: loans with an obligation to buy are eroding the financial planning of small clubs. A big club parks a player at a smaller one on loan, but the obligation clause converts it into a certain purchase in the next accounting period, and the small club carries the money without controlling the timing. Yet if all I have is a line about a loan, with no contract structure, no fee, no trigger conditions, no sell-on percentage, I cannot say anything about its financial consequences. Contract detail is a data cell, and when it is empty every inference is a false probability.

Nine dimensions, one usable conclusion

The tactical dimension had no subject: no team, no player, no coach, no match. No formation can be reconstructed, no sophistication judged. The financial dimension had no entity, so revenue and expenditure could not be decomposed. The results dimension had no league, no season stage, no form run, so the gap between expectation and reality was unmeasurable. Team positioning requires at least two named clubs to compare; there were none. Governance requires a body and an event; there were neither. The dressing room requires a person; there was nobody.

The media dimension was supposed to follow a story's heat cycle: emergence, acceleration, climax, backlash. To place a story in that cycle you must know when it began. The document had no date. Without a date, every conclusion risks being anachronistic: reading a pre-window story as if the window had already shut.

The industry transmission dimension collapsed the same way. The chain runs from academies upstream, through clubs and competitions midstream, down to broadcasting, commercial and derivative markets. No event, no chain. A major transfer can ripple into shirt sales, sponsorship activation, video-game licensing. A sanction can ripple into squad valuation and multi-year planning. But every one of those ripples starts from a real event with a date and a name.

Then came the last dimension, risk. Across all nine, exactly one risk was identified, and it did not belong to football: the risk that a blank table would be filled with plausible-sounding conclusions. Its level was rated high. Its likelihood was rated high, because it had already happened.

That is what makes the document valuable despite being empty. It shows that the greatest danger in my trade is not a wrong number. It is the false confidence built around a number that does not exist.

The weak point sits in the third column

People assume analytical failure lives in the dressing room, in form, in a row between manager and captain. I do not. Their weak point is not in the dressing room. It sits in the third column of the spreadsheet I filter.

Germany in 2026 had talent. Neuer, Kroos, Hummels, Müller. Germany did not collapse for lack of talent. They collapsed because nobody read the whisper of the numbers.

More precisely: when the defensive line pushes high and its height varies wildly, the space behind the centre-backs becomes an exploitable zone. A team with a forward who runs straight into that zone generates chances with a very low conversion probability but a very high frequency. Low probability, high frequency. That is the formula for group-stage shocks, where you have only three matches to correct a mistake. Nobody wanted to read a column of variance when that team had won the World Cup four years earlier.

When completeness becomes a false standard

The wider story of football analytics is usually told like this: machine-generated content is flooding everything, and the answer is more data, better models, more metrics. I do not accept that diagnosis.

The problem is not a shortage of data. The problem is that the industry has silently agreed a full table is better than an empty one, regardless of what gets put in it. When the article template has nine pre-built tables, and when editors only count filled cells, young analysts learn a dangerous reflex: leaving a cell blank is failure, filling it is job done.

I have seen that reflex across many metric tables. A team's set-piece goal share runs well above the league, and the immediate conclusion is that their coach is a dead-ball specialist. But a high share can come from a tiny sample, from weak defending opponents, or from happening to employ the two best crossers in the division. Correlation is not causation. I write that sentence three times a week and people still refuse to read it.

There is a distinction I learned over eight years, and it turns on two kinds of numbers. The first is proof data: it exists to confirm or reject a hypothesis you already have. The second is unanswered data: it exists to say the question is still open. Football analytics is excellent at the first and has almost no room for the second. A cell saying the data is insufficient looks like a confession of failure.

And when the second kind is driven out of the newsroom, it does not vanish. It moves somewhere else: the comment section, where a supporter confidently declares his team deserved more, based on exactly the forty-five minutes he caught during his lunch break.

Data practice is not about prophecy. It is about never being fooled twice by the same lie.

The signal for the next cycle

The failure will be fixed. The source will be reloaded, the deconstruction will re-run, the nine dimensions will have data and a real analysis will be written. That is the easy part. The harder part is designing a process in which an empty document counts as a valid result rather than an error to be hidden.

I am tracking three signals next season. First, how many football analyses attach a methodology detailed enough for readers to verify, rather than a decorative table of metrics. Second, how many data platforms openly flag a missing value rather than quietly interpolating it. Third, how many newsrooms dare to keep a blank conclusion when the template demands one, and let that blank become the content.

People watch the goal and cheer. I watch the seventeen-minute probability chain to understand why it happened. When that chain has no data, the only honest option is to say I do not know yet.

At thirty-three I believe every number is a witness that never lies. But an empty cell can lie, if someone is confident enough to fill it.