When the Data Falls Silent: The Line Between Analysis and Fabrication in Football

That night I sat in front of a screen with an empty data table. Fourteen RB Leipzig matches were already sitting on my hard drive, but my pressing-count software returned zero in every cell. No coordinates, no timestamps, not a single duel recorded. Two paths appeared in my head: admit the collection system had failed, or type in numbers that sounded plausible so the piece would look complete. I chose a third path. I recounted by eye, one action at a time, marking every timestamp. Two hundred and twelve pressing actions across fourteen matches. That number is not pretty, not as smooth as an automated report, but it is real. And in the profession I chose, real is the only non-negotiable standard. Years later, I still keep that habit. Every time a data table comes back empty, the memory of that Leipzig night returns, along with a question anyone doing analysis has to answer before typing the first word: when there is no data, what do you write? The answer sounds simple, but the football-content industry is finding it harder and harder to accept. There is a paradox at the centre of how we consume football information today. The more data is generated — every Premier League match now produces millions of positional data points — the greater the expectation that the analyst must always have an answer. Audiences do not accept silence. Algorithms do not reward emptiness. And editors need a headline, a number, a conclusion to fill the space on the page. I once sat in a meeting room in Manchester where an analytics deck was presented with full xG, PPDA and heat-map figures, but nobody in the room checked whether the raw data source actually existed. The deck was beautiful. Every cell was filled. No cell was left blank. And precisely because of that, nobody asked questions. When a table is too perfect, it grants a kind of immunity from scepticism. That is the trap I call "a template filled before the data arrives". The frame is ready. The subheadings are printed. The chart captions are written. Only the number is missing. And when the number does not come, the first thing sacrificed is usually honesty, not word count. I learned this not from a grand failure, but from a piece that was spiked in 2026. That summer, at nineteen, I sat in a newsroom preparing for the World Cup quarter-final between France and Uruguay. I had spent two weeks logging every France set-piece from the group stage. Forty-seven situations. I predicted France would win through dead-ball situations, and I had a long list to prove it. The editor spiked the piece. His reason was that "women's analysis is too emotional". I did not argue. I sent all forty-seven set-piece situations back through the internal email, with timestamps and positional descriptions. France won 2-0. The goal came from a corner, exactly the scenario I had drawn. The editor published it and put my name on the byline. What I took from it was not "I was right". What I took was a professional rule: when you are asked to say something you have no evidence for, the only way to protect yourself is to go back to the raw data, however small, however ugly, however unflattering. Forty-seven set-pieces is a small sample. But a small real sample is always worth more than a large invented model. There is a line I often repeat in analytics workshops: every number is a testimony, and the analyst's job is to make sure it cannot lie. But there is another version of that line few people are willing to hear: when a number is absent, its absence is also a testimony. That was the lesson of the summer of 2026, when the Premier League returned after the pandemic with empty stands and the five-substitution rule. I annoyed a lot of people by saying that no-crowd football is not "easier to analyse". The opposite. When you strip away the noise of the stands — the thing that usually masks small changes in match rhythm — you see more clearly that many teams no longer had anywhere to hide behind the emotion of a crowd. I tracked twenty Liverpool matches in that period. I found something I still remember clearly: from minute 60 to minute 75, Liverpool increased their pressing intensity exactly when opponents usually made three substitutions at once. Their xG rose by 0.23 after substitutions. The figure 0.23 is not large, but it appeared repeatedly in one specific time window, and that turned it into a signal rather than a coincidence. My supervisor undervalued the research direction. But the piece was published in a student journal, and it reached an analyst at Burnley. From there, I changed my writing completely. I began splitting a match into fifteen-minute blocks. Not because fifteen is a pretty number, but because it is long enough for fitness and tactics to shift, and short enough that a goal has not yet blurred its own cause. In a match, the score can change in the 89th minute. But the breaking point of a match usually comes much earlier — in the 58th minute, when the midfield loses its shape, when a full-back starts half a step slow, when a team starts abandoning its second line on counters. The score is the result. The fifteen-minute block is the cause. And if you only look at the score, you are reading the ending of a book you never opened. This way of reading is also how I face empty data tables. When a fifteen-minute block has no figures, I do not fill it with the match average. I leave it blank, and I note clearly: "no data in this block". An acknowledged gap is a gap that can be filled. A gap disguised with fake numbers is a gap that will ruin every model behind it. This is where the football analytics industry needs to face itself. When an analytics table is generated with a blank title, blank source, blank author stance and an empty list of information points, the only correct response is to declare that there is nothing to analyse. No match to dissect. No tactic to compare. No number to cite. Any content produced from such an input — however plausible it sounds — is fabrication, and worse, fabrication wearing the costume of a professional report. I know this sounds like a dry technical argument. But it has real consequences. Imagine a club analyst reading a report that looks complete, with tidy figures, and making a personnel decision based on it. Imagine a sports editor using that report to write a column, and that column being read by tens of thousands of fans. Imagine all of it resting on an empty source. Fabrication does not stop with its creator. It spreads exponentially, because every reader assumes the person before them already checked. Prejudice is only noise data the market has not learned to process. But fabrication is not noise. Fabrication is fake data processed too well, to the point where it looks more credible than the truth. This is where I think we are misreading a lot of matches. The public loves decisiveness. A headline that asserts with certainty spreads faster than one that admits uncertainty. A piece predicting Team A will win gets shared more than one saying the data is insufficient to conclude. So economic pressure pushes practitioners toward strong claims, even when the foundation for those claims is paper-thin. I fell into that trap once. In my first year of work, I wrote a fully confident analysis of a Championship team based on three matches I rewatched on tape. Three matches. I called it an "analytical sample". It was not an analytical sample. It was three matches, and I had turned them into a trend. A senior colleague pointed at the number three on the screen and asked one question: "And if the fourth match goes the other way?". I had no answer. I rewrote the piece the next day, with a humbler headline and a line admitting the sample was small. Since then, I set a rule for myself: never use a small sample to produce a large conclusion. And never use a large conclusion to hide a small sample. This brings me to a view many in the industry will not like. The value of an analyst is not that he always has an answer. It is that he knows when the answer does not yet exist. In a market where information is produced at industrial speed, the person who can say "I don't know yet" decisively is the most trustworthy. Disciplined silence is far harder than filling a page. And this is the line I want people to see clearly. On one side is honest uncertainty — "my data is not strong enough to conclude". On the other is fake certainty — "this team will definitely win because I have the numbers". From the outside, the two can look alike. Readers do not see the raw data table. Readers see only the words. But their consequences are opposite. One builds a foundation for a belief that can be corrected. The other builds a foundation for a belief that will collapse. I do not predict. I only read data one beat faster than everyone else. And when the data falls silent, I choose to stay silent too — until it speaks. Because football, in the end, is a system that can be verified, not a poem to be interpreted. There is one thing I have observed over years in this job: teams that succeed tactically are not the ones with the most data, but the ones that know which data to trust and which to discard. RB Leipzig under Ralph Hasenhüttl was an example. Their 4-2-2-2 was not a miraculous invention. It was a way of organising space so that Timo Werner could run into the gap behind the opposition defence. What made the difference was not the shape on paper, but the coaching staff's understanding of which data mattered — Werner's starting position, the timing of the pass, the direction of the opposing centre-back — and which data was just noise. When I counted two hundred and twelve pressing actions across fourteen matches, I was not counting only to prove a point. I was counting to find a pattern. And the pattern lay in this: Leipzig's pressing did not happen randomly, it happened on a specific schedule tied to Werner's position. That schedule is exactly what an automated data table, if filled with invented values, would never reveal. This is why I always return to a basic principle: figures must have a source. Figures without a source are not figures. They are pseudo-scientific literature. A rule changes one line, a football philosophy changes a whole generation. But this is only true when those changes are tracked with real data. The five-substitution rule did not make every team a better pressing side. It only transformed teams with enough squad depth to exploit the new window the rule opened. And the same is true of how we write. A new tool does not make content better. A new process does not make analysis more honest. Only the discipline that comes with it does. I once read an internal analytics report — not from a club, but from a media outlet — in which every fifteen-minute block was filled with figures, until I noticed that the total shots in the match were fewer than the total shots of the blocks added together. A basic arithmetic error. But it had passed three layers of review before reaching me, because everyone reads the conclusion, nobody reads the method. That is a lesson in verification. When a number appears in a fifteen-minute block that does not match the rest, you stop. Not to make it fit, but to find out where it came from. If a fifteen-minute block has no source, it has no value. If a conclusion has no timestamp, it has no value. This is why I despise pieces that recount a match as one continuous ninety-minute block. A match is not a block. It is six fifteen-minute blocks, and each block has its own logic, its own pressure, and its own gaps. When you write about a match without splitting it, you are not describing the match. You are describing your impression of the match. And impression is noise data. Back to the opening question: when there is no data, what do you write? My answer is: you write that there is no data. You describe the gap, you mark it, and you leave it intact until someone finds a way to fill it with truth. That is not a failure. That is professional honesty. I know this does not produce attractive headlines. But it produces a foundation for the next headlines to be built on rock, not sand. The next match comes this weekend. I will sit in front of the screen, open the data table, and check whether it speaks this time. If it does, I will read it. If it does not, I will wait. And I will not fill any blank cell with numbers I never counted. That is the line I choose to stand on. Not because it is safe, but because it is true.