When an Empty Data Payload Almost Became a Football Analysis

Late at night in Beijing, the glow of the screen fell across the wall of my study. I opened the final result file before shutting down the computer: an extract from the system I built to turn thousands of football articles each week into verifiable information points. The file was valid. Every data field carried its proper name. The domain label was clear: football. Behind those neat names, though, lay an absolute void — no title, no source, no team, no player, no date. Not a single information point. A file that looked perfect, and inside it, nothing at all.

I sat looking at it for a long while, and realized I was holding the most dangerous object in my trade: an emptiness formatted correctly.

In five years of doing football data journalism inside the Chinese market, I learned to trust the pipeline more than memory. A match passes, the feeling around it fades, but the data settles. The system I call Stage-1 does something that seems simple: it reads a football article and breaks it into countable fragments — information points, entities mentioned, core viewpoints, author stance, time sensitivity. Stage-2 is where I analyze. But Stage-2 only has value when Stage-1 delivers enough raw material. That night, Stage-1 handed me a box of the right size, with the right label, and nothing inside.

What stands out is that the file never raised an error. No red exclamation mark flashed. No exception was thrown. It drifted quietly through the system like any other ordinary file, and had I not opened it by hand, it would have sat in my database as a valid record, waiting for the day it would be fed into some analysis. That is the kind of failure that keeps me awake. Not the loud kind. The silent kind.

At eighteen, when I was a sports management student, I once predicted that Gasperini's Atalanta would hold a top-four Serie A finish, purely because their average PPDA of 9.2 was the lowest in the league — meaning they forced opponents to lose the ball faster than anyone. Back then I believed that if the numbers were right, the conclusion would be right. I was correct, and the piece reached two hundred thousand reads. But what I took from it was not "numbers always win." It was: numbers are only trustworthy when you know how they were gathered.

An empty file tells me the opposite of what it appears to say. It does not say "this article has no football content." It says "I failed to read, and I will not admit it."

The danger is not that data is missing, but that missing data wears the shape of complete data. A structurally correct empty file will not trigger any safeguard downstream. It passes the filters. It sits among real files. And then, on some afternoon, an analyst — or a language model trained to always produce an answer — receives it and begins to write. About what? About nothing, in a very confident tone.

I have seen something similar in another lesson of my life. In 2026, while writing my master's thesis, I compared 142 Bundesliga matches with crowds against 106 post-lockdown matches from the 2026-20 season, and found home win rates fell from 43% to 32%. Dortmund alone, with a PPDA of 8.1, won 67% at home with crowds but only 38% without them. I had a forty-page draft in hand, and I delayed it for weeks just to check more referee variables. Then a German analyst published almost identical results. I understood that perfectionism is not truth's friend — it is only timing's enemy. But that realization taught me something in reverse too: sometimes, knowing you have nothing to say yet is the most honest statement you can make.

Over eleven years observing the industry, I have watched football data journalism change. More newsrooms build automated pipelines, and there is nothing wrong with that. The problem is that we design systems not to miss anything, and forget to design them to know how to refuse. A machine that runs correctly will fill every empty cell. An honest machine will stop at the empty cell and shout.

Looking back at that file, I see at least three layers stacked on each other. The first is technical: most likely the crawler was blocked, or the article sat behind a login wall, or the page returned an empty body because the content was JavaScript-rendered. An article labeled "football" with not a single entity is almost certainly a pipeline fault, not a genuinely empty piece. The second is design: my extraction schema allowed the source and source-quality fields to remain blank. For a football story, especially a transfer story, source credibility is the most important field — and I had let it be left empty. The third, and the most uncomfortable, is cultural: the whole industry is so hungry for volume that we readily treat an empty record as a valid record, simply because it makes no noise.

And this is where I want to pause a little longer.

We are used to the image of a data journalist hunched over spreadsheets. But spreadsheets do not create truth. Spreadsheets only hold truth in place. Truth arrives from a shot in the fortieth minute, from a pass read wrong, from the noise of a full stadium, from the face of a player after missing a penalty. When Croatia reached the 2026 World Cup final with an average xG of just 1.1 per match and won three straight knockout rounds on penalties — with Subasic saving 5 of 12 faced attempts, a 41.7% rate — I wrote that this team did not need possession, it only needed to drag the match into its kingdom. No xG model taught me how to phrase that sentence. I learned it from rewatching those penalties, from counting the breath of players before they stepped up. Numbers came later, to confirm what the eye had already seen.

An empty file cannot walk that path. It has no shot to rewatch. And that is why it becomes an ethical test rather than a technical one.

I still call xG my map. It tells me which way a match flowed, which team created better chances, which team survived on luck. But the map is not the territory. I wrote that line in my first piece and have repeated it in almost every one since, because each passing year proves it more true. A coastline on a map is a thin line. In the world, it is waves, salt, wind, and countless footsteps passing back and forth that no one records. A good analyst knows when the map can speak for the territory, and when it is only staying silent.

With an empty data file, the map is silent. And the only honest thing a practitioner can say when the map is silent is: I know nothing yet.

But that is not what systems usually do. Most pipelines are designed never to return emptiness. They are designed to always have an answer. A model trained on a vast corpus will be very good at producing a plausible-sounding analysis. Hand it an empty box, and it will write out formations, tactics, form, transfer potential, all in a tone so fluent that no one thinks to check what is inside the box. That is not intelligence. That is reflex. And a reflex without honesty attached is just a politer way of lying.

I remember that I used to sell players by minutes run, not by TV fame. That principle sounds cold, but it protected me from overselling a name simply because the name was familiar. Minutes run are an uncontestable truth: either the player was on the pitch, or not. An analysis built on an empty box, meanwhile, rests on nothing at all. It is like selling a player based on minutes no one ever saw them run.

In the days after that incident, I sat down with my system and added a gate. It is simple to the point of being funny: if the list of information points is empty, or if not a single entity is named, the pipeline halts and returns a structured error, instead of quietly passing to the next step. I call it the refusal gate. My trade taught me many ways to see more from a match. It never taught me how to see more from a blank page — and perhaps that is right, because sometimes the work is to admit that a blank page is a blank page.

There is one thing I learned in my younger years, when I threw myself into processing thirty-eight Serie A rounds just to prove a mid-table club belonged in the top group: passion does not save sloppiness. I was right about Atalanta, but if I had been wrong, nothing in how I worked would have protected me from being confidently wrong. Confidence without foundation is only luck wearing a robe.

And a confident system without foundation is the same. It is not lucky. It simply has not been checked.

What troubles me most is not my own pipeline bug. It is the possibility that it happens at far greater scale, in places where no one sits opening each file late at night. When an entire industry races for volume, empty boxes drift through silently. They get counted into publication totals. They get aggregated into coverage reports. They generate analyses in which none of those analyses ever saw a match. And the harm is this: readers cannot tell the difference. They will read a fluent, grammatical, terminologically correct piece and have no reason to doubt it — because doubt only arises when a specific detail makes us pause, and an empty box provides no specific detail to pause on.

So I write this piece not to tell a story about a time I almost went wrong. I write it because what I almost did could become the standard, if no one stops to pause.

Looking ahead, I see a few signals worth tracking in the coming months. First, the frequency of empty records per run. If that number rises above baseline, it is no longer one article's fault — it is a sign that an entire collection layer is degrading. Second, the presence of source metadata: publisher name, author, publication timestamp. As long as those fields are permitted to stay blank, our trade remains vulnerable. And third, the writer's own habits — how many times this past week did you actually sit down and ask yourself whether the data in your hands is a match, or only the skeleton of a match that never took place.

I am grateful to that empty file for one reason. It taught me that editorial discipline is not about publishing fast. Discipline is about knowing when not to publish. In a world where speed is treated as a virtue, stopping is an act of rebellion. But if data is the map and the match is the territory, then honesty lies in recognizing the blank map, instead of drawing roads onto it that do not exist.

There is a line I still keep in my notebook, written from the time I broke my own hands on this love because I wanted it perfect: Tactics are the winners' account, data is the losers' original draft. I have thought about it a great deal this week. Because an empty box is sometimes the most honest original draft — it flatters no one. The only problem is whether we dare to read it as it is, rather than filling in the missing parts with our own hands.