When Football's Analytics Pipeline Returns a Blank Page

At 2 a.m., the phone buzzed, and a truth cracked open.

It was March, in Chengdu. I was rewatching the tape of the 2026 Shanghai derby, an old habit for sleepless nights, because that was the match after which a supporters' group called for my boycott, following a piece in which I wrote that a domestic striker's goals came only against weak opponents. The screen lit up. A young editor sent me a data file produced by an outsourced system. Title: none. Source: none. Genre: unclassified. List of information points: empty, not a single line. But the domain label stayed lit green: football.

He wrote: "Could you check, the system says this article has no content to analyse."

I read it three times. Then I stayed awake until five in the morning because of a smaller detail: the system never raised an error. It stayed silent, then forwarded that blank page to the next step as if everything were normal.

Most crises in football data do not come from wrong numbers, but from blank pages treated as though they were full of content.

Football analytics in the region has lived on faith in the data pipeline for a decade. Few people still chart every phase by hand. Newsrooms rent automated systems to extract text, recognise entities, tag topics, and hand the writer a refined package of raw material. The promise was beautiful: faster, cheaper, covering leagues no reporter could afford to fly to. A second-division match in Portugal kicking off at 3 a.m. would still get written up.

In Vietnam, most sports outlets do not build their own tools. They buy them, or share infrastructure. That means a failure at the ingestion layer spreads across several newsrooms at once, and none of them sees the failure, because everyone receives a file that looks normal.

The cost lies elsewhere. Such a pipeline has at least two parts doing two different jobs. The classifier answers one question only: what field does this topic belong to. The extractor is the part that pulls out sentences, names, figures, timestamps. The first part worked; the second died. The result is an odd hybrid state: the football label glows while the interior is hollow.

I call it an empty report. It is more dangerous than an obvious fault, because an obvious fault gets fixed.

There are four causes, ordered by how often I have met them. First, the source page is blocked, or sits behind a paywall, or is rendered in JavaScript so the crawler receives only a skeleton with no flesh. Second, the article did download but was stripped clean by ad filters, leaving only a line asking the reader to log in. Third, the extraction model timed out and returned the default schema, meaning a table of empty cells. Fourth, and most irritating, someone submitted a placeholder record and forgot to delete it.

None of those four is the fault of the original article. All of them are the fault of the pipeline.

The problem begins at the next step. An analyst is asked to write nine dimensions: tactics, club finance, results, league landscape, regulatory compliance, dressing room, risk, media, industry transmission. The brief does not say the input is empty. So people write. And the register of an analytical report is always confident, even when it is describing nothing.

More concretely: the extractor must recognise names, club names, competitions, timestamps and figures. When it dies, those things vanish together. A piece about a transfer loses the fee, the selling club, the buying club. A piece about an injury loses the number of games missed, and turns a player with a chronic record into a meaningless name.

I once received such a record. It discussed a player, but not which player. It mentioned an illegal approach to a man still under contract, but not who approached whom. It mentioned third-party ownership, long banned by FIFA, but not which party. It mentioned multi-club ownership networks of the City Football Group or Red Bull kind, where players and staff are moved by a central hand, but not who went where. Such a record is useless to a writer, yet entirely valid to the system.

I have seen this with my own eyes. A tactical report with room for expected goals, for passes allowed per defensive action, for pressing structure, for the position of every line. But without a single information point, the writer can only fill the blanks with phrases like insufficient information and keep the frame intact.

At the same time, a financial report can construct an entire story about UEFA's financial fair play rules, about the Premier League's profit and sustainability rules, about sell-on clause percentages, about the solidarity mechanism that pays training clubs. It sounds professional. But if no club can be identified, it is all rhetoric.

Then the dressing room. A report can discuss the power model of the head coach, whether one man controls recruitment or only the training plan, which players carry dense injury histories, where the generational friction lies. All of it is possible. As long as you never ask for a name.

Then the final layer, which I consider the most dangerous: media and expectation. An empty system can still generate a frame about the market's expectation cycle, about the gap between squad value and results, about the credibility of transfer sources. The trouble is it has nothing to compare. Without market expectation and without objective assessment, there is no gap to measure. But the frame still looks good, and a good-looking frame is easy to publish.

I am not writing this to snipe at a piece of software. I am writing because the same logic runs at a higher level: the sports rights economy. Streaming platforms are still paying peak prices for rights packages, losing heavily, then repeating exactly the mistake pay television made twenty years ago. Buying volume, hoping quality will generate itself. My data pipeline does the same: it buys volume of articles and trusts that meaning will arrive on its own.

It does not arrive on its own.

People call me a heretic, but I only see what they refuse to look at. Since I began writing at twenty, I learned one simple thing: watch the tape first, conclude afterwards. Based on my experience following matches across many seasons, what separates an analyst from a talking machine is not the volume of data, but whether a person is willing to open a figure and check whether it has any interior.

In 2026, I built an argument about a striker from exactly one fact: five consecutive derbies without a goal. The piece reached 2.3 million views in 48 hours. That club's supporters' group called for my boycott. Three days later, an assistant coach of the national team messaged me privately, saying my analysis was sharp and that his young charge was mentally fragile under pressure. I do not tell that story to praise myself. I tell it because it shows something else: an argument only stands when the facts are real.

In 2026, at the Luzhniki stadium, I commentated a semi-final live and mispronounced a Belgian midfielder's name three times in the first half. The internet laughed at me for a week. That night I stammered, but history did not. I shut the door, turned off the phone, and rewatched that national team's technical footage for thirty days. The result was a piece called insane, in which I gambled that a nineteen-year-old would change European football within five years. Time answered for me.

Both times, what saved me was going back to check the raw material.

Here I could be wrong in at least three ways.

First: perhaps that empty report reflected the truth. Perhaps the source article genuinely had no content, was a placeholder record, and the system behaved correctly by returning an empty list. If so, the problem is not the pipeline but the person who submitted it.

Second: the minimum gate I am about to demand could kill the very thing I love. In 2026, mid-pandemic, I built a late-night streaming series out of old contacts. A reserve goalkeeper in Guangzhou, nicknamed Fat Cat, called at 2 a.m. to describe his fear of being forgotten by his club while the stands stood empty. That conversation became a long feature, shared more than forty thousand times. Had an automated gate blocked every unverified source that night, that piece would not exist.

Third, and perhaps most accurate: the risk is not in the machine. It is in the person who reads the output and refuses to look. A system returns a blank page, and an editor is too tired to click the file open. I have been that person. I have published simply because the frame was already there.

So my proposal is small, and needs no regulation. Before any analytical step runs, the pipeline must ask itself one question: is there at least one real information point in hand. If not, it must shout, rather than silently returning empty cells.

And I offer a testable prediction: before the end of 2027, at least one major outlet will publish a machine-written match report for a fixture that was postponed, and will have to correct it within seventy-two hours. When that happens, do not blame artificial intelligence. Ask who looked at the blank page and nodded.

This industry does not lack data. It lacks people willing to open the file.