Broken Data Pipelines: The Ugly Truth Behind the Global Football Analytics Industry

Hook: A Match That Never Existed

On the evening of March 12, 2026, a football analysis channel based in São Paulo posted a 45-second infographic on social media, flatly asserting that a major Brazilian club had dominated possession with 68%, generated 1.4 xG, and delivered a perfect high-pressing performance. The only problem, and the fatal one: that match was never played. No team took the field that day. No referee blew a whistle. No spectators occupied the stands, not even virtual ones. The entire infographic was generated from a template mistakenly published by a staff member after the internal data pipeline collapsed and returned an empty result. The terrifying part was not the technical failure. The terrifying part was that the post accumulated 900,000 views, 12,000 shares, and four tactical comments over 500 words long analyzing "the team's pressing system" before anyone discovered the truth. When I ran the same test across six major football data platforms operating in Europe and South America, four of them still displayed xG figures for a match that did not exist in any official fixture list, with no referee and no match report. That was the moment I realized the problem was not one broken link. The problem was that the entire machine had been programmed to celebrate even when there was nothing to celebrate.

Context: Fifteen Years of Transformation and Its Hidden Cost

Over the past fifteen years, the global football analytics industry has undergone a transformation that anyone in my line of work must acknowledge: from intuition to data, from binoculars in the stands to dashboards on laptops, from handwritten notes on paper to machine-learning models predicting outcomes. When I was a master's student at the University of São Paulo in 2026 and released my video on Corinthians' "pressing fallacy," all I had was three Excel sheets and free software. Today, every top-flight Brazilian club operates at least two player-tracking systems, each match generates over 1.4 million data points, and any football analysis account on X can access a London-based provider's API before the referee blows the final whistle.

In Vietnam, this wave arrived later but crashed in harder. Since 2026, domestic football forums began translating and reposting indices from international providers. Television commentary programs gradually introduced xG, PPDA, and "progressive pass rate" into prime-time slots. V.League 1 news groups on Facebook and Zalo shared data graphics as if they were irrefutable indictments. Fans started debating "a central midfielder's progression index" the way they might debate the weather. This is precisely my point: when the data infrastructure is more fragile than the pitch infrastructure, the football industry has built an extraordinarily sophisticated interpretive machine, while the data-collection machine is hollow from within.

I am not saying this to dismiss progress. I am saying this because I have spent nearly a decade inside this industry, standing on both sides of the battle—the side that produces data and the side that consumes it—and I have come to see that the gap between surface glamour and actual quality only widens. This is a systemic problem, not an individual error. And it directly affects how we understand football, from the single-match level to the national-strategy level.

Core Insight: The Reality Gap

What is striking is not that a data pipeline broke. What is striking is the speed at which the football analytics industry learned to fill that void with expert-sounding judgments without ever checking whether there was anything to judge.

I spent six weeks tracing this phenomenon. I started from a simple and perhaps naive hypothesis: if I feed an empty data pipeline and ask it to return an analysis of a match, which system returns what? The results forced me to write this piece. Of the eight major data providers I tested—three European companies, two Brazilian, two Southeast Asian, and one American—five "filled the blanks" by assigning season-average values to a match that did not exist. No error message was sent out. No red flag was raised. Only one system out of eight returned null values and blank spaces, and that system was built by a group of students at a university in Paraná with a budget under twenty thousand reais.

What does this say about our industry? It says that behind every beautiful graphic, behind every tweet with an attached radar chart, behind every three-thousand-dollar scouting report, there is a very real possibility that we are presenting a shadow. I do not say this humorously. I say it after nearly a decade in the trade. I was once fiercely criticized for a 2026 Corinthians analysis video, and I have moved past the phase of being proud of my data foundation. But the moment I saw a major broadcaster's dashboard displaying PPDA for a match postponed two hours by heavy rain, I understood that this industry's problem is not a lack of data. The problem is that we have turned data into a god we worship, without ever checking whether that god exists.

The Three-Tier Architecture of Modern Football Data

Before continuing, I need to be clear about the architecture of the modern football data industry, because I believe understanding this structure is a prerequisite for understanding everything else.

The professional football data industry has three tiers. The first is the collection tier, where companies like Second Spectrum, Hawk-Eye, Stats Perform, and Wyscout deploy camera systems and human event taggers at stadiums. In Brazil, this tier is paper-thin. The league contracts with certain providers, but only for televised matches, and data quality drops sharply for second-division or cup games. In Southeast Asia, the situation is no better. Many V.League 1 matches are still tagged by one or two part-time staffers, with no optical tracking, no in-ball sensors. This means the indices we see on dashboards are built on a dataset with far lower resolution than Europe's top leagues.

The second is the processing tier, where raw data is cleaned, standardized, and transformed into indices. This is also where my problem begins. During my testing, I discovered that many processing systems behave in an extremely dangerous way: when input data is missing or empty, they do not report an error. They substitute the mean value. This is a deliberate design decision, because the system's goal is to "deliver a complete dashboard to the end user." If an index is missing, the dashboard looks bad. If the dashboard looks bad, the client is unhappy. If the client is unhappy, the contract is cancelled. So the system fills the blanks—and it fills them with the most potentially misleading value available.

The third is the interpretation tier. This is where we—analysts, journalists, content creators, influencers—work. We take data from tier two, add tactical context, narrative, and opinion, and turn it into consumable content. The problem is that we often have no way of knowing whether a specific index in our dashboard is real data or a mean value inserted to fill a gap. And when we write about it as experts, we transmit that uncertainty to thousands, tens of thousands of readers.

A Concrete Example of the Collapse

Let me illustrate with a specific example from my own experience in the 2026 Brasileirão season. I followed a match between a mid-table club and a relegation battler. The match was played in horrendous weather—heavy rain, flooded pitch, poor visibility. The referee allowed play to continue after a thirty-minute stoppage. Fifteen minutes after the final whistle, an international data provider published xG figures for both teams: away side 0.87, home side 2.14. Final score: 0-0.

I was at that stadium. I took handwritten notes on every chance. The home side created only three chances that could be called "dangerous," and none had a scoring probability above 0.15 under any model I know. So where did the 2.14 xG come from? I contacted the provider. The answer, after three days of waiting: the optical tracking cameras were disrupted by light reflecting off the flooded pitch, ball-position data was lost for roughly eighteen minutes, and the processing software inserted root-mean-square averages to "stabilize the data." The result: a match with no goals suddenly became a match with a total xG close to that of a ferocious attacking display.

If I had only read the dashboard without attending the match, I would have written an analysis about "poor finishing" or "the goalkeeper's best performance of the round." And I would not have been mathematically wrong—I would simply have analyzed a different match from the one that took place. This is the essence of the problem: flawed data does not produce obviously wrong conclusions. It produces plausible-sounding conclusions that describe a different reality.

The Parallel With the 2026 Empty-Stadium Press

Here is where I want to dig deeper. In 2026, when the pandemic emptied every stadium on earth, I was fired from a Brazilian media outlet for an analysis headlined—in the editorial team's words—as "reactionary": "Football without spectators is a mirror reflecting the industry's manufactured emotion." At the time, I thought I was writing about audiences. I now understand I was writing about something deeper: when the stadium empties, the football industry loses the emotional shell that hides its hollow internal structure.

The empty stadium of 2026 was the most honest stress test of what we call "the football experience" without spectators, and it remains the most honest stress test of what I am describing in this piece: when you strip away the coating of emotion, of chanting, of media, what remains is a system whose true value depends on whether it can operate even when empty. But this time, the phenomenon is not an empty stadium. It is an empty data dashboard. And unlike an empty stadium, an empty dashboard cannot be detected by the naked eye.

Imagine an analyst in Hanoi opening their dashboard on a Monday morning, having been unable to stay up all night for a European match because of the time difference. They read the numbers. They write a twelve-line tweet. They do not know that three indices in that table—xG, PPDA, and "touches in the box"—were filled with season-average values due to a processing failure at tier two. They are analyzing a fictional match in the language of a real one. And their audience, who believe they are reading deep analysis, are consuming a product worth zero. This is what I call the reality gap in modern football analytics: the industry has built an interpretive infrastructure of extremely high resolution, but the data infrastructure beneath it remains at the level of a leaking pipe.

The Case of Emerging Markets and Global Asymmetry

I want to devote this section to an aspect European analysts often overlook: global asymmetry in football data infrastructure. This is not a purely technical matter. It is a political matter of knowledge.

In Europe, a Premier League match is tracked by at least three independent optical systems, plus on-site human taggers, plus in-ball sensors. In Brazil, a Brasileirão match is tracked by one system, and third-division matches may be tagged by a single part-time staffer. In Vietnam, the situation sits somewhere between these two poles. What does this mean? It means that when an analyst in Vietnam uses data from a European provider to evaluate a Vietnamese player in V.League 1, they are comparing two fundamentally different-quality products. They are applying a template built for Eredivisie data to a dataset with far lower reliability. And they will never know, because the processing tier will never tell them that the source data is missing.

I had the chance to speak with a former event tagger at a data provider operating in Southeast Asia. This person, who asked to remain anonymous, told me that in many matches with bad weather or poor pitch quality, they were asked to "fill the blanks" by inferring ball positions from prior events. That means: if a pass exchange was not recorded, the tagger would automatically attribute it to the most likely player based on the ball's last position. The result is a dataset that looks complete, but in reality, roughly 8 to 12 percent of events are the product of inference, not observation.

Imagine what that means for an individual stat sheet. If I am a central midfielder with 42 touches in a match, but 5 of them were inferred, my stat sheet does not reflect what I actually did. It reflects what an algorithm thought I was likely to have done. And if a Vietnamese club uses that stat sheet to decide to sign me, the club is making a decision worth hundreds of thousands of dollars based on a product less accurate than advertised. This is not speculation. This is operational reality. And it is why I believe any serious discussion of applying data analytics to Vietnamese football must start with the infrastructure question, not the model question. You cannot build a beautiful house on a foundation that does not exist.

The Contrarian Angle: Where I Could Be Wrong

Now comes the part where I must check myself. Because I have built my reputation on controversial opinions, I know well that a contrarian angle is not the same as a correct one. So let me examine the arguments against my own position seriously.

First, perhaps the strongest: the inserted mean values may not be as harmful as I describe. Perhaps in a match where data is lost for only a few minutes, filling means does not change a careful analyst's conclusion. That is a fair point. If data was lost in the first three minutes of a decided match, filling the gaps may not affect the final judgment. But the problem is that the analyst does not know which data was lost. They do not know whether it was the first three minutes or thirty minutes mid-match. They do not know whether the xG figure filled in is the affected index. That ignorance is the problem, not the filling itself. And the solution is not to stop analyzing. The solution is to demand transparency in the infrastructure.

Second: the whole football data industry is young, and these problems will resolve as technology advances. I have heard this argument many times, and I concede there is some truth to it. Better automation, more sensors, and more advanced error-handling algorithms will improve reliability. But this is not merely a technical problem. It is a problem of economic incentives. As long as clients pay for a dashboard that looks complete, providers will have every incentive to deliver a dashboard that looks complete. Technology can mitigate the problem, but it cannot resolve the underlying conflict of interest without a change in how the industry shapes its product.

Third: perhaps I am exaggerating the importance of this problem because I look at it through the lens of someone obsessed with accuracy. That is a fair point. Most football audiences do not need xG precision to the thousandth. They need a story. And if the story is told with an index carrying a 15-percent error, the audience is still satisfied. From a consumption-experience perspective, the problem I describe may not matter. But from the perspective of a club spending millions based on data, or a nation building a youth-development program on data, the problem is life-or-death. The exaggeration is not on the side where I describe the problem. The exaggeration is on the side of an industry advertising its own precision.

The Human Cost

Before I move to the progressive conclusion, I need one section on what technical analyses usually ignore: the human cost. Behind every broken data pipeline, behind every mean value inserted to fill a gap, there is a person.

There is an event tagger working twelve hours at a stadium on the outskirts of São Paulo, paid seventy reais per match, told to be "flexible" when the data is insufficient. There is a young analyst in Hanoi working until two in the morning to prepare a scouting report, only to discover their data source was auto-generated without warning. There is a second-division Brazilian player undervalued because their index was filled with another player's mean. There is a coach fired because a fabricated xG figure showed his team "finished poorly" when in fact his team created nothing. These people do not appear in infographics. They do not appear in radar-chart tweets. But they are the ones who pay for the glamour of the analytics industry.

I do not write this to generate emotion. I write it because I believe any serious discussion of innovation in football must include the question of who benefits and who is harmed. In this case, those harmed are the least-voiced people in the industry: low-level data staffers, young analysts, players in minor leagues.

Infrastructure as Power

I want to extend this analysis into a broader dimension, because I believe it has meaning beyond the purely technical. When I talk about data infrastructure, I am not just talking about cameras and software. I am talking about power. In modern football, whoever controls the data infrastructure controls how the story is told. This is why Europe's biggest clubs invest tens of millions of euros in internal data systems: not only to understand their teams better, but to control how their teams are understood.

In Brazil, this asymmetry is stark. Clubs like Flamengo and Palmeiras have analytics departments with dozens of full-time staff. Clubs like Bragantino or Juventude have one or two part-timers. In a match between these two types of clubs, data is not neutral. Data reflects the standpoint of the party with more resources to collect and interpret it. For Vietnamese football, I believe this is an important lesson. When V.League 1 clubs begin investing in data analytics—and they should—they must understand that infrastructure is not a neutral investment. It is a strategic investment that determines what kind of story will be told about Vietnamese football in the future. If clubs depend on foreign providers for every index, they are allowing a third party to shape how their teams are understood. This is not a theoretical issue. It is an issue of data sovereignty.

Positive Signals

I have spent most of this piece criticizing. But as I have learned in my career, a critic who only criticizes without proposing solutions is intellectually lazy. So let me speak about what is being done right.

During my testing, I encountered several encouraging initiatives. In Brazil, a group at the Federal University of Rio Grande do Sul is developing an open-source data system in which every missing value is clearly flagged rather than filled in. The system has not been adopted by commercial providers, but it represents a different approach. In Southeast Asia, some clubs in Thailand and Malaysia have begun building internal data departments to reduce dependence on international providers. That is the right direction. And in Vietnam, I am told some universities have begun training in sports data analytics, though at a small scale. That is a start. But I want to see more.

What Needs to Change

If I were given one chance to change how the football analytics industry operates, I would focus on three things. First, data transparency. Every published index must carry an indicator of source-data quality. If an index was filled with a mean value, that must be clearly shown. This is not a difficult technical requirement. It is a professional-ethics requirement. Second, investment in grassroots infrastructure. Minor leagues, youth teams, and non-televised matches must be tracked with quality at least equivalent to major matches. Otherwise, we will keep creating a two-tier system where rich teams are analyzed with real data and poor teams with fake data. Third, analyst training. Everyone working with football data must understand how data is generated, from camera to dashboard. An analyst who does not understand the limits of their data is not a careful analyst. They are a translator of a text they cannot read.

Progressive Conclusion: The Mirror

I write this at a moment when the football analytics industry sits at the peak of its technical arrogance. Data providers compete to publish newer, more complex, "more accurate" indices. Analysts build personal brands on their ability to read those indices. And audiences, with no way to assess data quality, trust the numbers they see. But I believe this moment will end. Not because of a revolution, but because internal contradictions will force the industry to change. A club will spend millions based on a flawed index. A player will lose a contract because of fake data. An important match will be fatally misunderstood. And when that happens, the industry will be forced to confront the question it has been avoiding: if we cannot trust the data, what do we trust?

The answer, I believe, is not to abandon analytics. The answer is to rebuild data infrastructure with honesty about its limits. This is hard work. This is work that does not generate viral tweets. But it is necessary work. Because the truth is this: an empty data pipeline is not a technical failure. It is a mirror. And when you look into that mirror, you see our industry reflecting itself: a system that can invent a fictional match, host a debate about it, and award a prize to whoever analyzes it best.

The empty stadium of 2026 was the most honest stress test of what we call the football experience. The empty dashboard of 2026 is the most honest stress test of what we call deep analysis. Both show us the same thing: when you strip away the coating, what remains may be empty or may be full. And only honesty in the infrastructure can tell us the difference. The final question I leave to the reader, and to myself: when did you last check a number before citing it?