Null Input: The Silent Flaw in Modern Football Analysis

Marseille, a November morning. On my desk sits a fourteen-page document. The cover reads "Deep Tactical Analysis." Inside it is divided into nine sections: tactics, club finance, results cycle, league landscape, rules compliance, dressing room, risk, media, and industry transmission. Every section contains tables, conclusions, and a line marked "confidence level." Yet across all fourteen pages I cannot find a single number, a single player's name, a single scoreline, a single date. No match has ever existed inside it. I sat with that document for a long time. Not because it was difficult to understand, but because it was too familiar. That document is the product of a two-stage process. The first stage is meant to decompose a source article into information points: title, source, events, entities, timestamps. The second stage takes those points and constructs nine dimensions of professional analysis. On this particular run, the first stage returned empty. The title field was blank. The source field was blank. The list of information points contained not a single item. And yet the second stage ran anyway. It filled each field in turn with the phrase "insufficient information," then wrapped those phrases in exactly the framework it had been designed to use. The result was a text with the shape of analysis, the rhythm of analysis, and even a glossary of technical terms at the end. That was the moment I understood that the worrying part is not the machine. The machine did precisely what it was built to do: it observed the format it was given. The worrying part is that we have grown so accustomed to that shape that we no longer check what is inside it. Modern football operates on three stacked data layers. The lowest is raw event data: every pass, every duel, every shot, recorded with coordinates and timestamps. The middle layer is the model layer, where raw data becomes expected goals, pressing indices, space-control metrics. The top layer is the narrative layer, where numbers become the stories audiences actually consume. A pipeline like that is only as strong as its first gate. If the first gate is empty, the two layers behind it can still run — and that is precisely the danger. In statistics, when data is missing, practitioners typically fill with default values, impute from comparable samples, or drop the variable from the model. Every one of those techniques is legitimate, on one condition: the result must carry a label reading "imputed data." In football media, nobody attaches that label. An imputed number walks into the article wearing exactly the same clothes as a measured one. I have watched this industry for thirty-seven years. Over that span, data moved from being decoration at the foot of an article to being the spine of the trade. Clubs hire data scientists. Broadcasters sell advanced analytics packages. Readers learn to read heat maps, learn to distinguish key passes from chances created. That is real progress. But alongside it, a new habit formed: trusting the form of method instead of the method itself. A piece with tables looks more credible than a piece without, even when the tables are empty. That habit was not born from machines. It was born in the press conference. Fate is not decided in the press conference room — but it begins to be written there. A coach answers seven words about the injury of his first-choice centre-back, and in those seven words there is no name, no diagnosis, no recovery timeline. By that evening, those seven words have become a three-thousand-word article with a headline declaring the defence will collapse. Nobody in that chain lied. But nobody checked how much information they actually had either. When people look at Porto 2026 and see a miracle, I see an equation waiting to be solved. That year I rewatched the Champions League final between Porto and Monaco eleven times in three days. Porto held less of the ball, and generated five clear goalscoring chances against Monaco's one. It took me two weeks to write a twelve-thousand-word piece on the geometry of proactive defending. Colleagues called it dry. A lecturer at a university in Lyon used it as teaching material. Both reactions were correct, and both showed me the same thing: when the data is dense, conclusions can afford to be dry; when the data is empty, conclusions are forced to be flashy. There are three routes by which that flashy habit enters football today, and each is identifiable. The first is models running on samples that are far too thin. Expected goals is a good metric, but it is a metric of probability, not of truth. A striker with a high xG across four matches proves nothing beyond the fact that he shot from good positions in four matches. Shots are rare events; in a single game a team may take only eight to twelve. At sample sizes like that, variance exceeds signal. But once that number reaches a newsroom, it is treated as a fixed property of the player. The label "small sample" falls behind, and what falls behind is never read. I have tested this against my own collected data. For a player with ten shots, the confidence interval around his xG figure is wide enough that two players whose numbers differ by a factor of two can still be statistically equivalent. No transfer report prints confidence intervals. No transfer report prints sample size. We have built a culture of citing numbers without building the habit of citing their uncertainty. The second route is the transfer rumour economy. Transfers are the market of hope, and hope rarely respects valuation. A database such as Transfermarkt values players through an algorithm plus community contribution; that is an estimate, not a transaction. Yet in articles, that estimated figure becomes the denominator for calculations that sound rigorous: premium over market value, projected return, deal plausibility. All of it computed from a number nobody has ever confirmed. The Saudi Pro League is the clearest illustration. In the summer of 2026, clubs there spent more than eight hundred million euros bringing in a wave of stars past their European peak. That figure was repeated worldwide as evidence of a rising football nation. But place it beside academy infrastructure, beside the number of certified coaches, beside average domestic attendance, and the picture changes entirely. Those contracts did not develop football. They turned European stars into tourism ambassadors and turned the league into a television product sold beyond its borders. As a commercial strategy that is coherent. It is simply not what the headlines said it was. The third route appeared only in recent years: the automated content layer. Machine-written match reports now exist in lower divisions where no reporter travels to the ground. In principle that is good: more matches recorded, more players named. But automated content is only as trustworthy as its input. When a match has no event data, the model can still generate fluent prose about a match it knows nothing about. And readers, trained over years to trust form, will not notice the difference. In Vietnam, this problem has its own shape. The national league has fewer detailed data sources than Europe's top divisions, while the volume of commentary about it is enormous. That gap gets filled two ways: translating foreign reports, and reasoning from what was visible on television. There is nothing wrong with reasoning from observation. But when that reasoning is presented in the language of data, readers lose the ability to tell observation from measurement. I still remember how the story of Vietnam's 2026 AFF Cup triumph was told. Most pieces spoke of spirit, of desire, of a collective nobody could beat. Those factors were real. They simply do not explain why that team kept clean sheets in the decisive matches, why every goal conceded came from a set piece that had been prepared for, why the midfield always shifted into a four when possession was lost. There was an equation there. It was just never written down, because writing it down sells less. Esports is at the stage football never had a chance to return to: being documented correctly from the start. There, every action is recorded at frame level, every metric is traceable, and there is no room for an analysis built out of nothing. Football has travelled too far in the opposite direction to turn back easily. The counterintuitive point sits here. The standard reaction to a document like those fourteen pages is to blame the algorithm. I think that diagnosis is wrong, and wrong because it skips the most important control gate: editorial judgment. The machine honoured its contract. It was programmed not to fabricate data when data was absent, and it did not fabricate. The fabrication sits at the human layer — in the decision to publish a text with no content, to give it a headline, to place it alongside real work. Deeper still, the genuine blind spot is language. Science has a word for finding nothing: the null result. It is reported, published, and treated as a contribution to knowledge. Football media has no such word. An analysis that reaches no conclusion is treated as a failure rather than a finding. Writers are therefore pushed toward saying something, whatever its basis. Within that frame, the most honest text available — "we do not have enough data to conclude" — is the hardest to publish. The space on the pitch is wider than any great figure who ever stood on it. It is also wider than any table ever drawn to explain it. A table is useful only when it is built from what happened on the grass. When it is built from a void, it becomes an ornament whose persuasive power exceeds that of fact, because nothing specific can be used to refute it. So what should be done? I propose a minimum gate, applied to every tactical claim before it is published. Three questions. First: where did this number come from, who measured it, and by what method. Second: how many observations does it rest on, and how wide is its uncertainty band. Third: what date is it stamped with, and has any variable changed in the interval. Those three questions sound so simple as to be trivial. But apply them to an ordinary evening of watching football, and most of the assertions scrolling across the screen will not survive the first gate. That is why they are rarely asked. I keep those fourteen pages in a drawer. Not as a memento of a machine's error, but as a reminder of the most likely failure in this trade: building a flawless frame around an empty space, then letting the frame convince the reader that the empty space is substance. Next round of fixtures will produce at least a few dozen published analyses, in many languages, with many different tables. It would be useful, while reading one of them, to ask how many information points it was built from, and whether those points actually exist. Collapse is not the end of the tunnel. It is the largest dataset life ever provides — and a null input, recognised in time, is a form of collapse worth recording.