The Data Keepers: What Remains When Football's Analytics Pipeline Goes Silent

A Morning With a Blank Screen

6:40 a.m. A training ground on the outskirts of Guangzhou. The gatekeeper opens the iron gate on time, as he does every day, not a minute earlier, and vapor still clings to the grass in a way only someone standing on the touchline would notice. I sit down on the familiar concrete bench, open my laptop, and see a blank screen.

Not blank because the power is out. Blank because the data pipeline I open every morning — the metrics dashboard of a team I have followed all season — returns an empty table: no pass counts, no heat maps, no pressure figures, not a single line. The dashboard's lights are still green. It simply says nothing at all.

What surprises me is that I am not panicking. Eighteen years of carrying a recorder and a notebook have taught me something no classroom ever did: a blank space is also data. The question is not "where did the system break" but "when the system goes silent, what can I still see."

I close the laptop. And I start looking with my eyes.

On the pitch, a central midfielder arrives three minutes after the first group. He does not run. He walks one lap around the halfway line, touches his hand to the post of the small goal, and only then begins warming up. A detail so small that if a television camera were here, it would be cut from the frame. But it is exactly the thing my data pipeline would never record, even if it were working perfectly.

This article was born from a morning like that. No goals, no scoreline, no brace to open with. Only a blank screen, a training ground, and a question I have carried through many seasons: are we building data pipelines to understand football, or to replace the act of understanding football?

From Notebook to Pipeline

In 2026, when I walked into the sports department of a television station in Belgrade, a sports reporter's toolkit contained exactly three things: a notebook, a pen, and a pair of legs. You went to the ground, you counted with your eyes, you remembered with your head, you wrote with your hand. If you wanted to know how many passes a team made, you had to sit back, watch the tape, and count them yourself. No one did it for you, and no one sold it to you.

Twenty years later, the young reporter sitting next to me in Guangzhou can open a browser tab and immediately see completed passes, ball recoveries in the opposition's final third, the distance covered by each player, and even the goal probability of every shot. He does not need to count. He needs to know how to read.

That shift is not only a matter of technology. It is a matter of power. When data becomes a commodity, whoever holds the data holds a piece of the story. And when the football story is told through numbers supplied by a handful of companies, the question "who is telling it" becomes no less important than "what is being told."

Look at today's football data market. There are corporations that collect event data — every pass, every tackle, every foul — and sell it to broadcasters, bookmakers, clubs, and sports news sites. There are companies that specialize in positional data, recording the coordinates of twenty-two players and the ball twenty-five times per second. There are scouting platforms that let a lower-division English club watch thousands of hours of footage of a defender playing in the Norwegian third tier. Together they form a network I call a "pipeline": raw material enters from the pitch, flows through processing plants, and emerges as charts, reports, and decisions.

But there is one thing these pipelines never record: how a player plants his foot on the grass after being substituted in the sixtieth minute. And sometimes, that detail is the most important thing of all.

Anatomy of a Data Pipeline

To understand how a screen can go blank, we need to know what is inside the pipe. A modern football data pipeline usually has four layers.

The first layer is collection. Here there are two main families of raw material. One is event data: every action by a player is coded manually or semi-automatically by people sitting in closed rooms, watching the tape and typing into a system. The other is positional data: a camera system or a chip in the shirt records the position of each person and the ball. This layer is where humans and machines meet, and also where error is first born.

The second layer is cleaning. Raw data is always dirty. A touch is mistaken for a pass. A player is assigned the wrong shirt number for three seconds. A ball goes out of play and comes back, making an algorithm think a combination has occurred. The cleaning layer is where small decisions — this algorithm or that one, this threshold or another — quietly shape everything that flows downstream.

The third layer is modeling. This is where raw data becomes a metric. One model computes goal probability from position, angle, and the number of defenders in front. Another computes a team's pressure level. A third assigns value to a player based on his contribution to the collective's ability to create goals. Every model is a philosophical view of football rewritten in mathematics.

The fourth layer is presentation. This is where data meets the reader. A chart, a table, a metric printed in bold on a news page. And it is also the most deceptive layer, because a number presented neatly always looks more credible than a complicated truth.

What few people notice: most of the power of football data lies in layers two and three, the layers the reader never sees. When you read "Team A completed 89 percent of its passes," you are trusting hundreds of small decisions made earlier by people you have never met.

And when the pipeline goes silent, what you lose is not the truth. What you lose is one version of the truth.

xG: When a Goal Is Measured in Probability

Over the past two decades, no metric has changed the way football is read as much as goal probability, usually written as xG.

The core idea is simple. Every shot does not produce a certain goal; it produces a probability. A shot from close range, at a wide angle, with no defender blocking, has a high probability. A shot from outside the box, at a narrow angle, through three bodies, has a low probability. Add up all those probabilities and you get a number saying: if this match were played a thousand times, Team A would score an average of this many goals.

The power of xG is that it separates outcome from process. A team can lose 0-1 while generating three times the xG of its opponent. In a single match, that says little — football is cruel and random by nature. But across ten, twenty, thirty matches, xG begins to reveal the truth: which team is playing better than its results, and which team is living on luck.

This is where I learned my biggest lesson about data. In a season when I followed a club in Guangzhou, there was a run of five matches in which the team won four. On the scoreboard, everything looked beautiful. But the xG table showed that in those four wins, the team created fewer chances than its opponents and won thanks to extraordinary finishing efficiency and a few miraculous saves. I wrote a short piece saying this winning run was fragile. Three weeks later, the team lost four in a row.

I tell that story not to praise myself. I tell it to point out something data can do that the naked eye struggles to do: it measures the gap between what is happening and what should be happening.

But xG also has a dark side. When a metric becomes popular, it gets abused. People start using xG to draw conclusions about matches they have never watched. People cite xG as a verdict, forgetting that xG is only a model — and every model has assumptions. An xG model trained on data from the big European leagues can misread a match in a league with a completely different tempo and space.

And here is what I always tell young reporters: a number does not speak for itself. It only speaks when you know the conditions under which it was born.

PPDA and the Trap of Pressing

If xG is about the quality of chances, another family of metrics is about how a team wins the ball back. Among analysts, one of the most cited metrics is PPDA — the number of passes an opponent is allowed to make before your team performs a defensive action.

The lower the PPDA, the higher the press. The higher the PPDA, the deeper the team sits and waits. It sounds simple, but this is one of the most misunderstood metrics in the game.

The first problem is context. A team can have a very low PPDA simply because it is losing and has to push up. Another team can have a high PPDA because it is leading and deliberately cedes territory. The same number, two entirely different stories.

The second problem is location. Where you press matters. A team that presses immediately after losing the ball in the opponent's final third creates very different chances than a team that presses in midfield. PPDA cannot distinguish the two.

I remember a training session where I was allowed to sit in a corner and observe. The coach split the team into two blocks and kept shouting one word: "position." Not "faster." Not "harder." Just "position." After the session, his assistant explained: pressing is not about running a lot, but about standing correctly. A player who runs thirty meters to chase the ball may be breaking the whole system, if the position he vacates is exactly where the opponent wants to exploit.

That is what a metric like PPDA can never tell you. It measures the number of actions, not the quality of waiting.

Some revolutions have no slogans, only training sessions no one films. And in those sessions, what is taught is not to run faster, but to understand that every step has a price in space.

When Data Goes Silent

Back to the blank screen that morning.

In the data industry, there is a term for this phenomenon: silent failure. Unlike a system crash — where everything collapses loudly and everyone knows — silent failure is when the system keeps running, keeps showing green lights, but the data it returns has long been wrong, incomplete, or empty, and no one has noticed.

This is the most dangerous kind of error in any data pipeline, and also the kind least discussed in sports journalism.

Imagine the consequences. A metrics dashboard is broken for three weeks. A reporter writes ten articles based on that dashboard. An editor approves those ten articles. Readers read, believe, and share. No one knows that the foundation of all that analysis was a blank space presented as a fact.

I once witnessed a milder version of this story. One time, my data dashboard showed a team's pass count higher than reality, because the system counted passes from the extra time of a different match that had been misassigned to this one. If I had not gone back to the tape, I would have written a piece praising that team's ball control based on a number that did not exist.

The lesson is clear: data does not protect itself. There must be a human standing between the pipeline and the page, and that human must have the right to say "hold on."

In the newsrooms where I have worked, the person with that right was usually the least noticed. Not the famous editor, not the writer with the biggest following, but the fact-checker — the one sitting at the back of the room, pen in hand, eyes fixed on a spreadsheet, whose job is to doubt everything.

I call those people the data keepers. They do not go on air. No one knows their names. But when the pipeline goes silent, they are the only ones still standing in the right position.

Money, Rules, and Financial Pipelines

Football data does not live only on the pitch. A large part of it lives in the books.

For more than a decade, European football operated under a financial control system the industry calls financial fair play, created to stop clubs from spending far beyond their means. In recent years, England's top league replaced it with a new set of rules on financial sustainability, with loss limits calculated against a club's revenue.

What interests a notebook-carrying observer like me is not the numbers themselves, but how they are produced. One club is docked points for exceeding its loss limit. Another escapes punishment because of how it classified an expense — for example, money spent building a stadium is treated as infrastructure cost, not counted against the loss limit. The difference between being punished and not is sometimes an accounting decision, not the amount actually spent.

In modern football, an accountant can be as important as a striker. That is a line I wrote a few years ago, and I still believe it is true.

But this is where the story gets complicated. Financial pipelines can also fail silently. A loss is misclassified. A revenue item is recognized a year early. Those things make no noise. They surface only when someone opens the books and reads line by line.

And when a club is docked points, what is lost is not only a position in the table. It is trust. Fans begin to ask: if the numbers in the books can be wrong, can the numbers on the pitch be trusted?

I have always thought this is why a sports reporter should not read only the scoreboard. The scoreboard is layer four — the presentation layer. The truth lives in layers one and two: in small decisions, in unnoticed expenses, in overlooked accounting lines.

The French–Chinese Lens

Born in France and working in China, I occupy a strange position: I see the same match through two cultures.

In France, football is read through theory. People talk about structure, about the gaps between lines, about how a midfielder drops to create a passing lane. French academies are famous for teaching young players to think about space before teaching them to run. A French coach can spend thirty minutes of a press conference talking about a half-meter shift.

In China, football is often read through emotion and results. Fans talk about spirit, about moments, about impossible shots. Patience for a long process is often shorter, and pressure on a coach often arrives faster.

Both ways of reading are half right. And both have blind spots.

The blind spot of the French reading is that it sometimes forgets football is played by humans, not diagrams. A system that is perfect on the board can collapse because one player has lost confidence. The blind spot of the Chinese reading is that it sometimes forgets emotion cannot replace structure. A team playing on spirit but without organization will lose to a team playing on organization but with less spirit, if the match lasts long enough.

I once stood in the mixed zone after a World Cup semifinal. I interviewed no one. I just stood there, overheard conversations, and noted one thing: a coach raised his hand to signal a positional adjustment to a midfielder, and immediately afterward the whole block shifted half a meter toward the flank. No words were spoken. No data was printed. Only a raised hand.

Later I confirmed with a French data analyst that the team had changed formation to limit a creative midfielder on the opposing side. The data confirmed what I had seen. But if I had only had the data, I would not have known when that moment occurred.

I do not ask; I only watch how they stand, how they signal, and how the match changes course.

Vietnamese Football and the Data Problem

In recent years, I have had the chance to follow Vietnamese football from a distance, and I see a story there far more interesting than anything the European dashboards can tell.

Vietnamese football has a young, knowledgeable, information-hungry generation of fans. They do not just want to know the result. They want to know why. They read about tactics, about transfers, about metrics that no one in the region cared about ten years ago.

But the data infrastructure has not kept up. A domestic league may not have positional data for every match. A reporter who wants to analyze tactics deeply may have to rewatch the tape and count himself, just as I did in Belgrade twenty years ago.

That sounds like a disadvantage. But I believe it is also an opportunity.

When you do not have data, you are forced to look. And when you are forced to look, you can find things that someone who only reads a dashboard will miss. I once told a young colleague in Hanoi: do not wait until you have a perfect data pipeline to start analyzing. Start with what you have — your own eyes.

But I also do not want to romanticize scarcity. Data, used correctly, is a liberating tool. It lets a small club find a player the big clubs overlooked. It lets a reporter in an under-covered league prove a point with evidence instead of feeling. It lets fans see the beauty of an off-the-ball run.

Vietnam's football problem is not whether to use data. The problem is building a data pipeline of its own — calibrated for the tempo, space, and playing culture of the region — instead of importing wholesale the models built for European football.

An xG model trained on European leagues may undervalue a shot from outside the box in a league where goalkeepers tend to play higher. A pressing metric built for high-tempo football may not accurately reflect a team that plays slow, controlling possession. Using foreign data without calibration is as dangerous as using no data at all.

That is why I believe the region needs people who do not just know how to read data, but know how to question it. People who understand that a number is an answer to a specific question, and that if you ask the wrong question, the answer will be technically correct but football-wrong.

The Counter-Intuitive Part Lives in the Blank Space

At this point, I want to say something many people in the industry will not like to hear.

The common assumption is that more data means more understanding. I am not sure that is true.

I have seen analysis rooms stuffed with data produce bad decisions. I have seen clubs with a whole team of data scientists still buy the wrong players, because they measured what was easy to measure instead of what mattered. And I have seen reporters with access to hundreds of metrics write analyses shallower than someone who simply watched one match and took careful notes.

The counter-intuitive part is this: data does not make you understand football more. It only makes you understand faster — including when you are understanding wrong.

A blank screen, therefore, is not only a malfunction. It is a reminder. When the pipeline goes silent, you are forced back to the thing you had set aside: observation. And in that observation, you sometimes find what data never touches.

I think of the groundskeeper at the Guangzhou training ground. He does not read dashboards. He knows the moisture of the grass by placing his hand on it. He knows the wind direction by looking at the flag in the corner. He knows whether the pitch will play fast or slow in the afternoon by watching how water pooled after the overnight rain.

No data model simulates that knowledge. And if your data pipeline goes silent on an afternoon like that, the groundskeeper is the only person on the pitch who knows what is about to happen.

They still tend the grass in an empty stadium, because they know that one day the lights will come back on.

The Data Keepers

Back to the morning with the blank screen.

After closing the laptop, I sat watching the training ground for nearly an hour. I wrote twelve small details in my notebook. A player adjusted his gloves three times before extra time in a practice match. A defender always stood slightly off-center toward the left flank when his team took a goal kick. A young striker did not take part in the final drill and sat down by the touchline, his eyes following the ball.

Those twelve details are recorded by no metric. But one of them — the way the defender stood off-center — became the backbone of an article that afterward received more reader mail than anything else I wrote that season. Not because I had better data than anyone else. But because I looked into a blank space and found a story.

This is what I want to leave to anyone reading this — young reporters, analysts, people trying to understand a sport that refuses to stand still.

Data pipelines will grow stronger. They will record more, faster, more accurately. They will tell you how many meters a player ran, what percentage of passes he completed, and how much goal probability he contributed to his team.

But they will never tell you why a captain changed the way he planted his foot on the grass in the second half of a match whose scoreline no one remembers. They will never tell you why a coach stayed silent at a press conference, and what that silence meant.

The captain does not shout; he just changes the way he plants his foot on the grass.

And if there is one thing I have learned after twenty years on the touchline, it is this: the biggest change often begins with a run no one notices.

What Remains After the Blank Space

So when the pipeline goes silent, what remains?

A training ground at 6:40 a.m. remains. A gatekeeper opening the gate on time remains. A midfielder walking one lap around the halfway line before warming up remains. A groundskeeper placing his hand on the grass to measure moisture remains. A fact-checker sitting at the back of the room with a red pen, doubting every number before it reaches the page, remains.

Those people do not appear on any dashboard. But they are football's real infrastructure — infrastructure that needs no electricity, no algorithm, no software update. They are the data keepers in the most original sense of the word: they keep what the machines would drop.

I am not against data. I use it every day. I believe it has made football more transparent, fairer, and sometimes more beautiful. But I believe a good sports reporter must be someone who can live both in the world of numbers and in the world of blank spaces — and know when to be in which.

The blank screen that morning taught me something no dashboard could. When everything you rely on disappears, you find out what you really have. And what I have, after twenty years, is not a perfect data pipeline. It is a pair of eyes trained to see the runs no one notices.

Tomorrow, the pipeline may come back to life. The dashboard will light up, the numbers will flow, and I will sit here again, reading them. But I will not forget to close the laptop once more, walk out onto the pitch, and look.

Because football, for more than a hundred years, has always been decided where the data has not yet arrived.