The Discipline of Silence: When Football Analysts Learn to Say 'Insufficient Data'

The Discipline of Silence: When Football Analysts Learn to Say 'Insufficient Data'

Rostov-on-Don, the evening of 2 July 2026. The temporary analysis room of a Brazilian television channel where I was sitting had three screens: one showing the live feed from the stadium, one showing a heat map updated minute by minute, and one showing a metrics dashboard I had built myself. Before kickoff I had committed to a prediction for the crew: Japan would collapse under Belgium's physical pressure, and they would not have enough depth to withstand the tempo of Kevin De Bruyne and Eden Hazard in the second half.

By the 52nd minute Japan led 2-0. Genki Haraguchi opened the scoring in the 48th minute, and Takashi Inui doubled the lead four minutes later with a long-range strike from outside the box. My dashboard was still blinking with pre-built probability values. They were completely wrong. Belgium came back to win 3-2 through goals from Jan Vertonghen, Marouane Fellaini and Nacer Chadli, the winner arriving in the 90+4th minute.

What I remember is not the scoreline. It is the emptiness of a carefully prepared model suddenly falling silent, and my own inability to say what came next.

When data becomes a new faith

Over the past fifteen years, the analysis rooms of major clubs have changed beyond recognition. Colorful dashboards have gradually replaced the assistant coach's notebook. Expected goals, expected assists, tackles per defensive action, running distance by fifteen-minute block — all of it now appears in pre-match press conferences. Data has travelled from the server room to the bench, and from the bench to the mouths of supporters.

What is striking is that this spread has moved faster than the understanding of it. Many people cite a metric the way they would cite an established truth, forgetting that every metric is born in a specific context, calculated with a specific formula, and subject to specific limits. A twelve-match sample cannot speak for a season. A three-match sample cannot speak for a cycle. Yet in everyday conversations people still talk about trends based on such short windows.

I will always remember the line that has followed me throughout my career: "Numbers tell the first part of the story; the rest is flesh and sweat." It does not deny data. It places data in its proper position — as the opener, not the final judge.

The summer of 2026 and the first question

In 2026, when I was thirty-nine and working as an assistant tactical analyst for Fluminense, the coaching staff brought a passionate proposal: to adopt a high-pressing model based on GPS positioning data collected from the last twelve matches. The metrics looked beautiful. Midfield running distance was up, ball recoveries in the opponent's half were up, chances created after winning the ball in the opponent's half were up.

I was the only person in the room to ask: is twelve matches enough to conclude anything, and how stable is the sample across the three previous seasons?

That question stretched the meeting by another two hours. But it saved us from a hasty decision.

When we re-examined the full data set, I found something interesting: the team's defensive system only truly succeeded when the opponent's lateral passing rate exceeded 62%. Below that threshold, when opponents passed more vertically and directly, our back line exposed deadly gaps between the two centre-backs and the two full-backs. The twelve matches the staff had presented happened to fall within the group of opponents who passed laterally a lot — a chance coincidence, a skewed sample, a textbook data trap.

Based on an analysis of forty-seven matches, I proposed keeping the 4-2-3-1 shape and simply intensifying pressure on the right flank, where opponents routinely exposed weaknesses during ball circulation. That season Fluminense finished sixth, four places better than the previous campaign.

A number needs a precondition

The first lesson I drew was not about which model was right or wrong, but about the fact that every tactical conclusion needs an accompanying precondition. When I say a defence is effective, I am obliged to add: under the condition that the opponent's lateral passing exceeds 62%. When I say a high press is effective, I am obliged to add: under the condition that the midfield keeps its spacing below fifteen metres during transitions.

This may sound obvious, but in reality the majority of football analyses forget it. People cite a metric as if it were always true in every context, then are surprised when it fails in a specific match. That surprise does not come from the data. It comes from having assigned the data a power it never had.

From the 2026 season onward, I began writing my analysis reports in a fixed structure: data, context, limits of application. I never issue a tactical judgment without its precondition attached. This structure makes the reports longer, forces the coaching staff to read more, but it also makes the decisions more certain.

World Cup 2026 and the forgotten space

Back to Rostov-on-Don. After the match I watched the footage five times. The first time to find faults in Belgium's defence, the second to study Japan's transitions, the third to count how often Belgium lost the ball in midfield, the fourth to measure the distance between lines, and the fifth simply to sit in silence.

What I missed was not in any traditional metric on my dashboard. It was in the space between the lines — something my model could not measure. Japan did not win through physicality. They won by exploiting the gap between Belgium's midfield and defence in fast counter-attacks, where running distance and tackle counts reflect nothing at all.

"The model is not wrong — it just has not yet learned how to speak." I wrote that line in my notebook after the match.

It took me three months to rebuild my analytical framework. Those three months were not for learning new tools, but for learning to recognize the zones my old tools could not reach. I added a mandatory section to the end of every report: "Overlooked factors." This is where I force myself to confess what my model does not see.

World Cup 2026 taught me that every model needs a humble seat.

2026 and the empty stadium

When the pandemic suspended leagues in 2026, I was tasked with analysing thirty matches played without spectators in the Brasileirão for a sports magazine. This was a rare opportunity — the first time in modern history that football was played under conditions almost free of the crowd variable.

The result staggered me. Home win rates fell from 48% to 39%. A nine-percentage-point drop is not small in a league where every point is precious. More importantly, teams playing a high press lost an average of 12% efficiency. The cause was not fitness but psychology. Without the roar from the stands, players lost part of an invisible drive in decisive duels.

"A match without spectators is the flattest mirror football has ever held up to itself."

Across those thirty matches I saw something normally hidden: how much of so-called "home advantage" actually comes from the stands rather than from the pitch or travel distance. "Home advantage does not live on the scoreboard; it lives in the players' eardrums." That was the conclusion I drew after cross-referencing pressing data, ball recoveries in the opponent's half, and passing-error counts across three different zones of the pitch.

I wrote a forty-page report proposing an adjustment to the "home pressure index" for all future analyses. The editorial board initially objected, saying it was too long and too technical for a general readership. They later split it into three instalments. The first on home win rates, the second on pressing efficiency, the third on psychological impact. All three received positive feedback, and I understood that readers are not afraid of complexity, only of meaningless complexity.

The limits of the sample and the classic trap

"A year without spectators, and we discovered something new about the game." But that discovery is only valuable if we know where it came from and under what conditions it applies. Thirty matches is a sample large enough to see a trend, but not large enough for an absolute conclusion. I always remind myself of this before making any judgment.

The classic trap of football data analysis is confusing correlation with causation. A team plays a high press and wins many matches — that does not mean the high press caused the wins. It may have won because it had better players, or a favourable fixture list, or because its opponents were in a squad crisis. Data shows us what happened, but it does not automatically tell us why.

At Fluminense in 2026 we nearly fell into this trap. Twelve matches against opponents with high lateral passing rates produced a sample so beautiful it was hard to believe, and without cross-verification across three seasons we would have built a tactic on a skewed foundation.

Cross-verification is not a formality. It is a way of thinking. Before trusting any metric, I always ask: where does this data come from, how large is the sample, is it skewed, and does the conclusion still hold when the context changes.

Tradition and data do not stand on opposite sides

In Brazil, where I live and work, there is a constant tension between those who trust traditional football intuition and those who trust data. Data sceptics say it kills the beauty of the game. Data zealots say intuition is obsolete.

I do not stand with either. "Tradition and data are not opposed; we use the latter to keep the former." What I learned at Fluminense is that data can help preserve traditional values rather than erase them. My proposal to keep the 4-2-3-1 rather than tear everything down for a flashy pressing model is one example. Data helped us understand better why that tradition worked, and under what conditions it needed adjustment.

In Vietnam, where I was born and still follow football from afar, this tension is also taking shape. Domestic leagues are beginning to adopt technology, clubs are beginning to collect player data, but the analytical infrastructure remains thin. Many decisions are still based on the instincts of insiders, and that has its virtues. Lived experience in football is a form of data that is never written down.

What I want to see is a meeting of those two streams, rather than a battle for the right to be correct. A good analyst is not someone who replaces intuition with data, but someone who knows when to listen to which.

The transfer market and the youth-price bubble

There is one field in which data is being badly misused: the transfer market. I have witnessed too many deals in which the price was shaped by a handful of glossy metrics while the context was entirely ignored.

When a twenty-year-old scores eight goals in fifteen matches in a mid-tier league and is valued at one hundred million euros, people usually invoke metrics like expected goals per ninety minutes, chances created, or successful dribbles. But they rarely ask the most basic question: how many matches has this player played at the elite level, and how does he respond when facing well-organised defences at the highest level?

A player who has not played even fifty elite matches cannot justify a hundred-million-euro price. This is a naked gamble, in which risk is transferred from the buyer to the ultimate payer — and the ultimate payer is usually the supporter, who bears rising ticket prices and expectations placed in the wrong place.

I am not opposed to paying high prices for young talent. I am opposed to paying high prices based on small samples without acknowledging that the small sample has limits. A young player can shine for fifteen matches for many reasons unrelated to his true quality. A fitting tactical system, weak opponents, protective team-mates, luck in finishing. Ignoring these factors is ignoring most of the story.

Expected goals and its blind spots

Expected goals is an excellent tool for assessing the quality of chances in a match. But it has its own blind spots. It cannot measure the quality of transitions, the sharpness of combination play, or the psychological pressure in decisive minutes. It does not tell us why a particular chance appeared, only how likely it was to be converted.

So when I read an expected-goals table, I always read it alongside other data: successful transition counts, ball recoveries in the opponent's half, average distance between lines at the moment of losing the ball. No metric stands alone. Each is a piece of the puzzle, and the picture only emerges when the pieces are placed correctly.

This is why I am allergic to commentary that throws out a single number and immediately concludes. A number standing alone is like a sentence with its second half cut off. It may be technically true, but it cannot tell the real story.

Video assistant technology and the question of final value

Video assistant technology has created a new layer of data and a new layer of controversy. It helps correct clear errors, but it also raises a question: what is a clear error, and who decides?

I have seen situations in which technology confirmed a referee's decision technically, yet persuaded no one in the stands. This shows something important: data and technology can deliver truth, but they do not automatically produce consensus. The human sense of fairness is far more complex than a slow-motion frame can convey.

In such cases, the best analyst is the one who knows how to place data in its proper context — including the emotional context of the match and the historical context of the competition.

A counter-intuitive view: when discipline becomes paralysis

Here I must be honest about a paradox I have experienced myself. The discipline of saying "insufficient data" is a virtue, but it can become an excuse for avoiding judgment. I have met analysts who use the principle of humility as a shield, and who then never issue any conclusion that could be contradicted. They are always right because they never say anything specific enough to be wrong.

That is not science. That is avoidance dressed up in academic language.

Football does not wait for data perfection. The match is on Saturday, and the coaching staff need a decision on Friday. They cannot wait another three seasons of data to be certain our model is right. In such moments the analyst must dare to make a judgment based on what is available, while stating clearly the level of confidence in that judgment.

"The best coach knows which numbers to trust when it gets hard." And the best analyst is the one who knows when to stay silent for lack of data, when to speak because there is enough ground, and when to speak with a warning attached that he may be wrong.

At Fluminense in 2026, had I simply said "insufficient data" and stopped, we would have preserved a mistake. What I did was not to refuse a conclusion, but to refuse a conclusion based on a skewed sample, and then to verify again in order to reach a grounded conclusion. The difference between those two attitudes is the entire meaning of the profession.

I call it decisive humility. Humble about the certainty of the model, decisive in issuing a recommendation. These two qualities do not exclude each other. On the contrary, they complement each other.

Source discipline and cross-verification

Another important part of the profession is source discipline. I grew up in journalism from 2026, when I was a reporter for a sports newspaper in Madrid, and there I learned that an unverified piece of information is not information. It is only a rumour shaped like a truth.

Today, with the spread of social media, rumour can travel faster than truth and be far more persuasive. An article based on an unclear source can shape public opinion about a player within hours. And once opinion has formed, it is hard to dislodge even when the truth is brought to light.

So in every analysis I write, I always make clear where the data comes from, how large the sample is, and whether I have verified it myself. When I cannot verify it, I say so. This is not perfectionism. It is the minimum respect owed to the reader.

I have also learned that a source can be right about the event and wrong about the viewpoint. A reputable newspaper can report a press conference accurately yet offer a distorted interpretation of its meaning. The analyst's responsibility is to extract the core event, discard the original viewpoint, and retell the story through his own understanding. That process demands a strict honesty with oneself.

What the eye misses when the lights go off

There are details in football that only appear when we take the trouble to look into the quiet spaces. The off-ball movement of a holding midfielder during pressing sequences. The wasted spaces on the flanks when a team transitions from defence to attack. The small decisions that only reveal themselves after many replays.

I spend a great deal of time on those details. Not because they are more prominent than the goals, but because they are what create most of the value of a match. The spotlight tends to fall on goals, but the truth of a match usually lies in the unlit zones.

This is why I believe football analysis should not be the business of goals alone. A team can win through an own goal, but its real story is written in forty other passages of play whose names nobody remembers.

The between-lines corridor and the dead zone

In modern football, the most important area on the pitch is the space between the lines — where the ball is passed through and where the match is truly decided. In Rostov, that space was the crux. At Fluminense in 2026, that space was the weakness I found after analysing forty-seven matches. In the spectator-free 2026 season, that space became harder to close because of the loss of psychological pressure from the stands.

I have built a habit: before judging any match, I always check the environmental context — attendance, weather conditions, pitch condition. This context is not an accessory. It is part of the data. A model that ignores environmental context is like a map that ignores the actual terrain of the land.

What I still verify in the next match

When every analysis is complete and every metric checked, I always leave an open question. That question is not meant to blur the conclusion, but to remind me that football can always outrun a model. Belgium and Japan in 2026 taught me that in a painful way.

For the next match I follow, I will check two things first: the stability of the data samples across the last three seasons, and the contextual factors my model cannot measure. If there is one thing I have learned over my career, it is that the final verification never happens on a spreadsheet. It happens on the grass, under the lights, in eighty minutes that no model can run ahead of.