A Complete Template, Zero Data: Pipeline Failure and the Evidence-Substitution Trap in Football Analysis

That evening I opened an analysis file sent over by the data desk. Thirteen major sections. Every one had a heading, a table, a "conclusion" row, an "evidence" cell, and even a glossary of professional terms annotated carefully from xG all the way to sell-on clause. I read top to bottom and underlined four times. By the last line one thing had surfaced: there was not a single piece of football information in the entire file.

Not one club. Not one competition. Not one player. Not one scoreline, fee, or date. The only label surviving ingestion was a single word: football. Every other field read N/A or sat empty. The document itself confessed it had no data, and it was still forwarded as a completed report.

I do not believe in randomness; I believe in repeated passes. A file like that does not reach my desk by chance. It is the trace of a rule-bound failure, and rules repeat until somebody blocks them.

For roughly five years now, most football analysis that readers consume daily passes through an automated chain. The first stage breaks the source article into information points: club names, player names, scorelines, timestamps, the author's stance, time sensitivity. The second stage takes that output and builds a multi-dimensional read: tactics, finance, results, league landscape, rules and governance, dressing room, risk, media narrative, industry transmission.

I once wrote that entire chain by hand. A twelve-part series on Croatia's midfield diamond at the 2026 World Cup cost me a full semester. I logged player coordinates every five minutes, counted twenty-four receptions by Luka Modrić between the lines, and added up to 11.2 kilometres covered, of which only three kilometres were forward movement. To do that I needed real data. When I lacked it, I wrote a single word in my notebook: not yet.

The 2026 shutdown taught me the same lesson at a different layer. One hundred and twelve days without football, an empty Anfield, and I had to remind myself that what I was measuring was no longer normal football. Liverpool's high line made more positional errors, midfielders lacked the auditory cue from the crowd to cover behind, but I could not attach full-season meaning to those numbers. I stated the limits of the model before I stated the conclusion. That is discipline, not modesty.

An automated chain has no habit of writing "not yet". It has a template. A template always has enough cells to fill. That is exactly where the risk is born.

The first thing worth saying plainly: a report that looks structurally complete can contain exactly zero grams of information, and the human eye cannot tell the two apart if it only skims headings. Thirteen sections, tables, bolded terminology, risk flags, star ratings — all of it produces a signal called "professional". That signal is entirely independent of whether data exists.

I call it the evidence-substitution trap. Readers see the markers of method and assume the method was applied to something. In practice, the method was applied to a void.

Look closer at that file and the fault localises neatly. The domain label survived — it returned exactly one word, football. But the article type could not be classified, the title was empty, the source was empty. When a downstream stage runs while an upstream stage dies, the fault sits on the boundary between reading and classifying, which is to say in raw text extraction rather than in football understanding. The machine still knows it is talking about football. It simply has nothing to say.

There is one detail I like because it is uncomfortably honest. In the source-credibility section, the document states that credibility must be judged "from the source fields of the information points". But the information points do not exist. That is a self-cancelling loop: grading the source requires information points, information points require a readable source, and the source is blank. The pipeline design has locked itself out.

As a working analyst, I value that the document chose to say it straight: the high risk rating here is information risk, not sporting risk. That is a distinction most football content online erases. People fold both risk types into one sentence — "the situation looks worrying" — and readers assume a club is in crisis, when the thing in crisis is the machine reading the article.

Put another way, what deserves analysis in that file is not on the pitch. It is in the process that produces what gets written about the pitch. Tactics is the only thing that cannot be faked on the pitch. But writing about tactics can be faked, and it is far easier to fake than most of the industry admits.

The first reflex most colleagues have when they see an empty file is to blame the tool. The machine broke, fix the machine, done. I think that framing is safe and misses the hardest part.

The bigger risk sits on the human side, in the incentive to fill gaps. A template with empty cells generates pressure to fill them. In football, that pressure has a name: it is called punditry. With no data on a deal, people tell a story about "sources close to the situation". With no metric for a defence, they talk about "fighting spirit". Those sentences fill empty cells exactly as the machine does — the only difference is that they are not flagged as errors.

During the 2026 summer window I was once first to report Emile Smith Rowe's loan move. What I remember is not the story. What I remember is the pressure to write more. I had 8.7 passes per ninety in the left half-space, a relationship with a scout, and a double-pivot system to cross-check against. And people still asked why the piece was "thin". Honesty about how much data you hold gets read as a lack of professionalism. That is the root of the problem.

The second paradox: automated chains are usually criticised for being mechanical and dry. In truth, at the final layer, the machine was far more honest than many writers. It wrote N/A in every cell. It pasted a data-integrity warning at the top of the report. It invented no match at all. The document is a mirror: it did precisely what I keep telling young writers to practise — leave the cell empty until there is evidence.

The worrying thing is not an empty machine. The worrying thing is an empty machine still read as a full one. A summariser one layer down can strip the data-integrity warning to make the text tidier, and then the conclusions of a data-free report flow into a news item as an ordinary football judgement. Nobody re-checks, because it has already passed through three layers.

I also do not believe this is an isolated case. Automated failures rarely travel alone; they travel in batches, because of the same feed, the same time window, the same software version. When one article in a batch is empty, its siblings probably are too.

One more thing few notice. That document listed all six football risk categories — sporting, financial, personnel, rules, public opinion, systemic — and marked every one as unassessable. That, not the absence of conclusions, is the genuinely frightening part: a hurried reader can assume those six cells were actually scored.

Every formation is a hypothesis; the match is the experiment. A pipeline is no different. It only earns trust when it has a hard gate: if the input result is empty, stop, do not forward, do not package it as a completed report. That gate costs far less than repairing a wrong news item already published.

But a gate at the machine layer only solves half the problem. The other half lives in the habits of writers and readers. I want football readers to develop a new reflex: when an analysis piece looks too complete, try counting the verifiable facts inside it. If a three-thousand-word piece contains no timestamp, no specific figure, no real name, then the template is speaking instead of the content.

And I want young writers to keep a habit I learned after the 2026 World Cup: after charting Morocco's six matches, I had enough data to conclude on the France game. But across the first three group matches, with fewer than two hundred possessions logged, I wrote exactly two words: not enough.

Football is producing more words than at any point in its history, with less verification capacity than at any point in its history. That gap will not be closed by more models. It will be closed by people willing to leave a cell empty when there is no evidence. Next time you read an analysis with every section filled, try to find the empty cell. If there is none, the whole piece may be the empty one.