When the Data File Comes Back Empty: The Discipline of Verification in Football's Analysis Room

In my office in Chengdu there is a data file I have deliberately refused to delete. It has every field a match report is supposed to have: fixture title, competition code, timestamps, a slot for expected goals, a slot for pressing intensity, a slot for expert commentary. The structure is so complete that anyone opening it and skimming would assume it is an ordinary analysis. Only by reading line by line does the emptiness show: every field holds its default value, and not a single figure was ever written in. What made me keep it was not the technical incident. It was the fact that the file passed three automated validation layers. It was structurally valid. It was correctly formatted. It raised no error. Had I been a young writer under pressure to file before going on air, I could have turned that empty shell into a very persuasive tactical piece: a side building a 4-3-3 block, a midfield pressing at medium height, both fullbacks pushing high, the centre-backs shifting into a back three in transition. Nobody could check it, because the file never named a match. An empty data file taught me more than any football analytics conference I have attended. A valid shell with nothing inside Football analysis has changed its skin over fifteen years. From notebooks and videotape, analysis rooms moved to automated feeds: event data on every touch, positional tracking data, models that quantify chance quality. A modern analysis piece typically passes through four stages: fetch, parse, verify, publish. Every stage can break. The easy break is the loud one. A dead connection, a software error, and the writer knows immediately that there is nothing to write. The dangerous break is the second kind: the system returns a perfectly shaped data object with an empty interior. The tables still render. The headline still exists. The fields still sit in their correct places. Only the content has vanished. The difference between those two failures mirrors the difference between a team with no data and a team whose data was never loaded. In the first case, the analyst knows he is blind and behaves accordingly. In the second, the analyst believes he is seeing. I once met that trap on a larger scale. On 30 June 2026, reviewing France's 4-3 win over Argentina in the World Cup round of 16, I counted Lionel Messi's touches in the attacking third: 23, his lowest across five matches at that tournament. I did not publish immediately. I cross-checked three independent statistical systems and re-watched the footage from three camera angles before writing. With a single source, I would have held a beautiful conclusion and a non-trivial probability of being wrong. Three hypotheses before any conclusion When an empty file lands in front of me, my instinct is to build several hypotheses rather than pick an answer. The source may simply contain nothing: the original article sits behind a paywall, is blocked by region, or has been removed. The fault may lie in the fetch stage, in which case one retry over a different route settles it. The remaining possibility is a failure in the parsing stage, where the information array was dropped in transit even though the source system was intact. These three directions lead to three different actions. If the source is empty, change the source. If the fetch failed, retry by another route. If parsing failed, fix the input pipeline. Choose the wrong direction and everything downstream is meaningless. One small detail narrowed the search considerably. The label for the one-sentence summary still existed; only the value beside it was blank. That means the data object was instantiated first, and the content was never attached. Had both label and value disappeared, the story would be different. A trace that small can shrink the debugging surface from an entire pipeline to a single stage. Football leaves similar traces. A team that exposes its midfield line in the 30th minute usually exposed it in the 8th, only nobody was recording yet. People are good at spotting midfield mistakes, but better at spotting them before the ball rolls. A gate, and the cost of not having one The operational lesson is concrete: you need an entry gate. Any data packet without a title or without an information array should be returned as an error state and never pushed downstream. It sounds obvious, yet most analysis rooms at small and mid-sized newsrooms have no such gate, because a gate produces no content. It only prevents false content from being born. Tactics are not a diagram on a whiteboard; they are a habit repeated over 90 minutes. A verification process works the same way. It only has value once it becomes a habit repeated on every batch, not a guideline pinned to the meeting-room wall. The bigger risk is contagion. If one file in a batch is faulty, sibling files in the same batch are likely faulty too, because they travelled the same pipeline and the same conversion step. In football terms, this is a single bad scouting report copied into four other dossiers. The error is not in each dossier; it is in the assembly line. So I read a batch twice. The first pass checks existence: is there a title, does the information array contain elements. Only the second pass checks quality. Mixing the two is the surest way to miss an empty file, because an empty file tends to look extremely tidy. Lessons from an empty stadium In 2026, when the pandemic turned stadiums into silent concrete blocks, I had a rare laboratory in front of me. I selected ten Leicester City matches in the Premier League after the restart and counted the share of safe sideways passes against risky forward passes. The sideways share rose from 24 per cent to 31 per cent. I could have written something loud: the crowd noise vanished, the players grew timid. But ten matches is a small sample, and ten matches inside a singular period is smaller still. I added an explicit methodological-limitations section at the end, stating the sample size, stating the circumstances, and confining the conclusion to its proper scope: within the ten matches observed. An empty stadium is the largest laboratory there is: it shows which team plays through structure and which team only plays through emotion. An empty data file deserves the same treatment. It permits no conclusion about any team. It permits only a conclusion about me: I came close to writing a tactical analysis out of thin air. Summer 2026 and the trap of certainty In August 2026 I spent nearly the whole month tracking Atalanta. The club sold several key players, did not reinvest at a matching level, took only one loan with a purchase option for Duvan Zapata, and kept Gian Piero Gasperini's back-three system intact. I wrote a prediction that the results would not hold, based on the precedent of mid-tier clubs selling players without reinvesting. The rule I set for myself afterwards still stands: never build on transfer rumour, only on signed contracts. Summer 2026 taught me that a mid-table club buys out of fear, not out of a plan. The prediction went the other way. Atalanta entered the 2026-2026 season playing the most ferocious attacking football in Serie A, scored 98 goals, finished third and reached the Champions League quarter-finals. I was wrong. What matters is where I was wrong: the contract facts were sound, the inference drawn from them was not. A historical precedent is a precedent, not a law. That mistake reinforced the rule I had already adopted after the Messi piece. Every article carries a source note for its figures. Every conclusion splits into two parts: the verifiable facts, and the inference I personally own. Readers are entitled to know which part they are reading. The blind spot is speed The familiar story about this profession is that careless writers fabricate numbers and decent writers verify. I do not buy that division. The problem is not personal morality. It is the speed gap. Publishing takes ten minutes. Verification takes two days. In a news cycle where every account reposts within half an hour, the verifier always arrives last. Nobody orders anyone to invent. It is enough that the system rewards whoever gets there first. The second paradox is more uncomfortable. A missing data field is an honest signal. It says plainly: there is nothing here yet. A field filled with a default value, or with a second-rate source, is a trap, because it looks like information. In this trade people fear the empty gaps, but the real damage comes from the gaps that have already been filled with something. On a morning with fifteen matches to cover, the choice is always the same: publish something thin but true, or publish something thick with nothing inside. The empty data file sides with the first option, and it is not remotely persuasive. What to verify next match Space is the only thing that cannot be bought in the transfer market. Data is the same. No budget can purchase an index that never existed. What I will do next match is small and specific. Pick one metric, count it myself, then compare it with the source. If they diverge, I log the discrepancy. If the file comes back empty, I write the word empty and stop, rather than filling it with something that merely sounds reasonable. Readers follow football to learn what happened on the pitch. They do not read to learn how confident the writer feels.