Empty Data Is Not Clean Data: When the Perfect Report Contains No Evidence

On November 12, a post-match analysis went live with everything a modern analysis is expected to carry. Expected-goals curves plotted minute by minute. A pressing map divided into three zones. Passing networks for both midfields. A tidy conclusion that the away side lost the centre of the pitch during the first fifteen minutes of the second half. Eighteen hundred words, published forty minutes after the final whistle. Nobody in the newsroom checked the data source. The next morning, an engineer opened the system log and found that the raw feed had been empty for six hours. What we published that night was an empty frame, coloured in the right places. The axes stayed where they were. The labels were complete. Only the data was missing. I sat in front of that screen for a long time, not because of the article, but because of a different question: if nobody had opened the log, how long would that report have survived? In nine years of data journalism, I have learned that a modern football analysis system has four layers. Ingestion pulls raw data from a provider. Extraction pulls out discrete information points. Summarisation turns them into a story. Only the final layer produces tactical, financial, and personnel judgements. The interesting part is where it breaks. There are two kinds of failure. The loud kind: the system throws an error, nothing gets published, and the matter ends there. The silent kind: the domain label still reads football, the template still runs to completion, every field still gets filled — except the content. The silent kind is far more dangerous, because it produces something that looks finished. And in football, things that look finished are rarely questioned. Now look at the pitch. The same four-layer architecture sits behind scouting, behind medical departments, behind transfer valuations, behind contract renewals. A club does not sign a player because a dashboard looks good. It signs because it trusts that the dashboard was computed from real data. If extraction goes silent and nobody opens the log, what the club receives is an empty frame coloured in correctly — and they will pay for it. In 2026, in my first month on the desk, I wrote that FC Seoul won the K-League on the back of 12 goals from 38 set-piece situations, 31.6 percent, against a league average of 18.4 percent. An editor threw the draft back at me with a line I still remember verbatim. I did not argue. I re-watched every minute of footage, annotated each dead-ball moment, and attached a methodology appendix so that anyone could check the work. From then on, every piece I filed carried a sources-and-method section. A year later, before South Korea met Germany at the 2026 World Cup, I filtered the Bundesliga player data and found Germany's average PPDA stood at 15.2 — meaning opponents completed 15 passes before every active defensive action — alongside enormous variance in defensive-line height. I wrote that a counter-attacking side with Son Heung-min's pace was the perfect fit, and that Manuel Neuer pushing high as a sweeper would leave space behind him. On 27 June 2026, the score was 2-0. Germany did not collapse for lack of talent. They collapsed because nobody read the whisper of the numbers. But 2026 taught me only half the lesson. The other half arrived in 2026, when stadiums emptied, my company lost seventy percent of its revenue, and editors were laid off in waves. I refused to write speculation about what might have happened without COVID-19. Instead I quietly built a ghost-match database, collecting more than six hundred closed-door fixtures as a control group. The world stopped turning, but my ghost football database kept breathing. My first battle had no audience. Just me, a spreadsheet, and a sinking club. That dataset taught me a principle worth carving into a keyboard: empty data and clean data are entirely different things, and the gap between them must not blur during presentation. When a metrics table has no row in it, the valid check result cannot be no issue found. The correct result is not yet assessed. It sounds like semantics. In football it is money, points, and survival. Take a financial example. A player signs for a 40 million euro fee on a five-year contract. Simple division gives 8 million euro a year on the books — the number that decides whether the club has room for its next signing. If contract data is missing at the extraction layer — term unknown, fee unknown, wages unknown — then every compliance analysis written afterwards is literature. And it will still read beautifully. This is also why I oppose the loan-with-obligation-to-buy structure spreading through recent windows. On paper it gives smaller clubs good players immediately. In substance it shifts financial risk onto them and turns them into finishing schools for the wealthy. But to prove that, I need fees, wages, and contract lengths. When those are empty, all I can publish is a list of what must be recovered. That ghost database later saved me an entire transfer window, because real football is not always as real as data. Or a medical example. A midfielder returns from a hamstring injury, but his high-intensity running volume over the last three matches has not been recorded. If the system returns an empty cell and somebody still concludes his fitness has recovered, starting him is a gamble wearing a scientific label. Same mechanism every time: an empty frame, coloured in correctly, and nobody checks. Their point of failure was not in the dressing room. It was in the third column of the table I filtered. Which leads to the question I consider central to every sports newsroom today: when data is missing, what should be published? My answer since 2026 has not changed: publish the list of what must be recovered. A title and a source. At least three discrete information points. Named entities — clubs, players, coaches, competitions. The stance and purpose of the original author. A time-sensitivity assessment, to know whether the story is still hot. And any quantitative figure: transfer fee, wage bill, xG, league position. That list is not attractive. It has no clickbait headline. It is eight lines long. But it is the only thing that puts an analysis desk back on track, and it reveals something more important than any content: where the system broke. Here I have to argue against my own tribe for a moment. People often conclude that data can outrun the crowd, and therefore data always beats intuition. That claim fails on the word always. In 2026 the metrics showed me something the eye could not. But there are also matches where a veteran coach's eye identifies a structural problem for which the model has no variable at all. If the metrics have not answered the question, the correct conclusion is that the metrics have not answered it — not that the metrics have refuted it. Distinguishing those two statements is my entire profession: data that proves and data that has not yet answered. Merging them is blinding yourself. At 33, I believe every number is a witness that never perjures itself. But a witness only testifies when asked the right question. The bigger trap lies with readers, and I have to admit this. A report that writes not yet assessed in all nine analytical dimensions will earn a few hundred reads. A confident report about a collapse will earn a few hundred thousand. That is a structural gap, not an individual's mistake. Any system that rewards confidence more than accuracy will eventually manufacture false confidence. In this industry we built exactly such a machine, and it runs very smoothly. I also have to be explicit that not yet assessed does not mean no risk. This is where most readers slip. An empty cell does not say the club is safe. It says nobody has measured. In a relegation month, that difference can be the whole season. The 2026 ghost database taught me this another way. When European stadiums emptied, home advantage dropped noticeably across many leagues — widely documented, though the size varies by country and by definition. Nobody could have drawn that curve from an empty table. It appeared only because someone collected data for months and accepted that the answer would have to wait. An empty dataset, read correctly, is a measurement of the measuring system itself. It tells you which layer your scouting department is blind at, which layer the newsroom is blind at, and how much of a club's decision-making rests on assumption. For a coach, the more practical question is simpler: am I missing data on fitness, on the opponent, or on my own team? Three different answers lead to three different match plans. Data practice is not fortune-telling. It is making sure the same lie never fools you twice. The next matchweek will open again. There will be post-match analyses with elegant curves and tidy conclusions. The question I want you to carry into it is not who will win, and not who will go down. It is this: which part of this report was computed from real data, and which part was simply coloured in to fit the frame?