When a 4,000-word football analysis contains zero numbers: lessons from an empty data pipeline

I opened the document at six in the morning, before heading to the training ground. Nine sections of analysis, forty-five lines of conclusions, a risk matrix arranged by severity. The report ran over four thousand words, exactly the format of a deep-dive piece any football desk would publish. But from the first line to the last, no match was named, no contract was described, no player appeared. Every data field carried the same tag: 'N/A — insufficient information.' Tactical assessment impossible. Financial assessment impossible. Risk assessment impossible. A machine had produced a football analysis with no football inside it.

I am used to receiving false reports during the transfer window. A player 'about to join,' a fee 'about to be agreed,' a negotiation 'about to collapse.' It is a season in which every source has a reason to talk and every word has a reason to be wrong. But this was the first time I received an analysis with nothing football-related in it. It did not lie. It also did not say anything. It repeated one answer: not enough data to reach a conclusion. And that empty honesty is what made me keep reading.

This report came from an automated article-processing system. It ingests input, extracts entities, labels the domain, then pushes the result down to a deep-analysis layer. In modern sports media, such pipelines handle hundreds of pieces each day. They decide which articles get tagged, which get recommended, which get promoted to the homepage. The noisier the transfer window, the more editors rely on machines to separate signal from noise. Yet machines create noise of their own — not the noise from the stands, but the noise of a system operating without understanding what it processes. An empty stadium still has noise. That noise is the noise of bad data.

In the report I read this morning, the extraction layer returned a peculiar result. The article title field read N/A. The source field read N/A. The article-type field read 'Unclassified.' Worse, the 'entities involved' field — the place where club names, player names and competition names should have been — contained an instruction: 'identify from the information points above.' This is a classic defect that engineers call template leakage: the question frame was pushed into the data field, and the actual data vanished. In dressing-room language: the starting lineup was published, but all eleven players on it were injured.

The only thing that survived the entire process was a label: 'football.' Lowercase, as if the system registered that whatever it was processing belonged to the sport just before everything else fell apart. A single signal is enough to route a document into the sports stream, but not enough to support a conclusion. For me, the boundary between an analytical tool and a story-fabrication machine is this: without a minimum data sample, no algorithm is allowed to speak.

I learned that lesson the hard way. In 2026, still a high-school student in Beijing, I published a preview predicting Germany would beat Mexico 2-0 at the World Cup. I relied on head-to-head records, on champion pedigree, on the name 'Germany.' The result was a 0-1 defeat to a Lozano goal, but the real disaster was that I had ignored Mexico's high-block pressing data: nineteen pressing actions in the opponent's final third in the first half, double Germany's average. From that match I set an unwritten rule: no commentary before pressing, xG and line-distance data. A big name must never replace a small data sample.

Three years later, I faced the opposite problem. When the Bundesliga returned during the pandemic in front of empty stadiums, I used nine rounds of data to argue that home advantage had dropped sharply. A professor pushed back: the sample was too small. I had to add five seasons of historical data to prove the decline was outside the margin of error. That experience taught me that every number must sit next to a sample limitation; otherwise it is just a lucky number. Data does not lie, but the people who choose the data do. This morning's report sits at the opposite extreme: it has no sample to defend, and it does not try to invent one.

At Euro 2026, half of my readers mocked my article as 'unromantic.' I wrote that Italy had only 48% possession against Wales and left space behind the full-backs whenever opponents transitioned quickly. Three weeks later, in the semi-final against Spain, Italy conceded sixteen shots and won only on penalties. The 'Mancini revolution' story remained beautiful, but my data was not wrong. I bring this up because it shows something uncomfortable: people prefer a good story to an accurate spreadsheet. And that is exactly why an empty pipeline like this report — if dropped into the publishing machine — would not be challenged by readers. It would be shared.

The report ranked five risk levels. The highest was not an injury or a financial-fair-play breach; it was a process risk: an empty result was forwarded to a downstream system without validation, and if that system has no rejection mechanism, it will fabricate conclusions to fill the void. In football, the equivalent is a club doctor asked to sign a fitness certificate for a player who has not trained all season. If he signs, the danger is not only to that player — the entire club's medical judgement collapses. For me, a successful contract is written in January, not June. A trustworthy analysis works the same way: it is built from clean data, not from a beautiful template.

People might think the problem here is missing data. The real problem is the habit of publishing empty data as if it meant something. Every day, on football sites, there are analyses written not from data but from an editor's desire for a good narrative. They pick a name, pick a deal, then search for a few numbers to illustrate it. This report, by contrast, refused to conclude when data was missing. A system that says 'not enough data' deserves more trust than a writer who invents a story without evidence. In the dressing room, I learned to read people through small details: how a player ties his boots before a big match, how he puts his phone away before stepping onto the pitch. Those details never appear in a stats sheet, but they are the raw material of every report. This morning's analysis had none of those details — but it did not pretend to have them. A document that admits 'insufficient information' is rarer than a document full of false information.

During the transfer window, what drowns fans is not news; it is fake confirmation. A social-media account posts a fee; another website reposts it; a reputable journalist repeats it, and the whole market treats it as fact. My two-independent-sources rule was built against exactly this: if only one agent speaks, I do not publish; if a reporter in another city, working for another newspaper, confirms the same fee, the story begins to stand. But when a newsroom's own automated system produces an empty analysis and nobody notices the emptiness, the risk is no longer outside — it is inside the newsroom itself.

The report also leaves signals worth tracking. The rate of empty data items in a batch keeps rising every few days. Instruction templates leaking into value fields suggest the extraction layer never reached the reading stage. And when a 'time sensitivity' field is never assessed, it means the pipeline stopped too early: it recognised the topic but never arrived at interpretation. I have seen similar signs in football: a team that runs a lot but runs to the wrong places, a player who is sent on but has never done the right drill. Machines are like players — running the wrong way is worse than not running at all.

The report also touched on what I consider the most important thing in this trade: source-quality grading. During the transfer window, I usually work with four tiers of sources — a nameless social-media account, an aggregator site, a reporter with a track record, and official club confirmation. Only the last two belong in a published piece. But a machine system, unless programmed to rank sources, treats an agent's rumor exactly like a press release. It mixes rumor with fact and pushes out what I would call a 'fake dressing-room note.' That is why the two-independent-sources rule is not an option: it is the final net that prevents noise from turning into fact.

A football analysis, whether written by a person or a machine, needs at least four things: a name to identify the subject, a few discrete events to anchor the story, an author's stance to define the frame, and at least one metric — a transfer fee, an xG figure, a possession rate, a contract length — to tie the conclusion to reality. This morning's report had none of those four. It was too empty to be called an article; it was only a skeleton of formatting. But that skeleton is the most valuable thing I have read this week, because it exposes a disease spreading through sports journalism: people sell dreams before checking whether the dream has any numbers to support it. The media sells dreams; I sell the dressing-room notes.

The two-source verification rule is not stubbornness. It is a safety door. Automated analysis systems need the same safety door: they must know how to refuse when the input is empty. But most systems on the market are designed to produce output, not to control quality. They are like a striker who always shoots even when he has never seen the goal. The report suggests a clear fix: place a hard validation gate, reject any payload with fewer than a minimum number of data points, and return an 'INSUFFICIENT_INPUT' status instead of an analysis skeleton. It sounds technical, but it is actually a youth-development lesson: do not send a player into the match when he has not done a single training session.

After a morning spent reading a report with no numbers, I do not expect an article to be written from it. I want to know whether football newsrooms will have the courage to build a source filter for their own systems. Because if a pipeline without a null check is allowed to publish four-thousand-word analyses, the football dream the media sells every day will soon be written in a language without a single number. And in this transfer window, when every source is shouting about hundred-million-euro deals, let me repeat one thing: fans see the performance; I see the Tuesday-morning training session. An analysis without data is as meaningless as a contract without a signature. The empty stadium still has noise — but the scariest noise does not come from the stands. It comes from a printer running without ink.