The Major Tournament and the Limits of Data: Lessons From Penalties That Cannot Be Measured

On the night of December 18, 2026, at Lusail Stadium, minute 118 of the World Cup final. Kylian Mbappé fired the ball into Argentina's net, equalizing at 3-3, completing the first hat-trick in a World Cup final since Geoff Hurst in 2026. The stands erupted. I looked down at my notebook, where I had jotted down the columns of numbers I had tracked all tournament: pass completion under pressure, turnovers in my own half, and one scrawled line I had written to myself — "a shaking hand is not in the data."

Three kicks later, Kingsley Coman stepped up. I knew his numbers. I knew Coman had completed passes at roughly an 84 percent rate in the Bundesliga that season, succeeded on more than half his dribbles, and according to the penalty sessions the French press reported, he was among the most reliable takers on the squad. But I also knew what the data did not say: Coman was just back from injury, and he had never stood before nearly ninety thousand people with an entire nation placed on his feet. Emiliano Martínez dived and saved. France lost. Argentina won. I was not surprised — but I was not unsurprised because of the data, but because I had learned that data has an edge, and that edge is called human beings.

That is the greatest lesson every major tournament teaches me again, in a manner both cruel and beautiful. Numbers do not lie, but the person reading them deceives himself for a lifetime.

Context: why a major tournament is a closed system

I work as a data consultant for a football club. It sounds grand, but the job is mostly sitting alone, taking indicators apart into questions, then answering them with documents and video. I turned 56 in a profession often judged by inspiration, and at this age I have learned that what is valuable is not predicting correctly, but knowing where I might be wrong.

A major tournament is nothing like a regular season. Across a domestic season you have dozens of rounds, a sample large enough for small errors to even out. At a major tournament you have seven matches, sometimes fewer. A small denominator makes every conclusion three times more fragile. A team can win a title with a low average xG, and a team can be eliminated with an enormous one. There is no contradiction here — it is simply that seven matches are not enough for probability to speak, so human beings speak instead.

I remember the 2026 World Cup. Before the tournament, when a newspaper asked me for a prediction, I wrote a line that made people laugh: Germany would be eliminated in the group stage. My basis was not a feeling but the PPDA — passes per defensive action — which for Germany in the pre-tournament window sat around 12.5, notably higher than their own figure four years earlier. Their midfield was allowing opponents to pass far too freely through the middle. On June 27, 2026, in Kazan, Germany lost 0-2 to South Korea and left the tournament in the group stage, despite generating roughly 2.0 xG. The data spoke. Germany left Russia before the group stage ended — I read it in March.

But I do not tell this story to boast that I was right. I tell it to say that even when I was right, I knew I had touched only a very small part of the truth.

The core: what major-tournament data actually says

When a major tournament unfolds, I divide the data into four layers, and I believe the order matters more than the numbers.

The first layer is results data: goals, scores, wins and losses. Everyone who watches football has this layer, and everyone thinks they understand it. But results are the noisiest, because they depend on a moment of luck that the preceding ninety minutes never promised. A match can be decided by a controversial penalty in the 88th minute, and in that case the result says more about the referee than about the two teams.

The second layer is process data: xG, dangerous chances, possession across the three thirds. This is where I work most. At a major tournament I often ignore first-half goals and focus on the structure of chances in the second half, because big teams tend to change at half-time while small teams collapse physically from the 60th minute onward. I once watched a team hold 62 percent of the ball yet generate only 0.7 xG, and that said everything: the ball was at their feet but going nowhere.

The third layer is physical and distance data: ground covered, accelerations, high-speed sprints. A major tournament compresses three weeks into seven matches, and I learned through my own failure that fitness is the most romanticized factor of all. During a European Championship I fell in love with a high-pressing model and wrote a whole paper urging the club to replicate the tireless full-back. The result: my wide players were exhausted after 60 minutes and the team lost four straight. I sat in a meeting with a fitness data sheet no one bothered to ask about.

The fourth layer is what I call silent data: the things not recorded, not measured, yet decisive. A shrug mid-match. A glance averted when a teammate demands the ball. A defender afraid to step up for fear of making a mistake. This layer lives in no table, and that is precisely why it is the most frightening.

Forty pages of a report go silent in a stadium with no applause.

I have lived that. During the pandemic, when leagues were suspended indefinitely, I withdrew to study nearly two hundred major matches played before empty stands. My finding was clear enough: home advantage fell sharply, with the host team's win rate dropping from around 46 percent to around 39 percent. I submitted a forty-page report to the board. They read one page and asked exactly one question: "So how do we win?" I could not answer, and the report was set aside. That season, after the domestic league resumed, my club won exactly one home match.

The lesson is not that data is useless. The lesson is that data must live inside a story. If you hand a coaching staff a page saying home advantage has vanished, they will do nothing. But if you say: "At home this season we win less because the empty stands stop defenders from daring to contest at the 80th minute, and that breaks our defensive structure," they will listen. A number must come with a person.

What the data says about the shifting balance among the great powers

At the most recent major tournament, I noticed a striking trend: teams no longer win by dominating possession, but by changing the state of the match in fifteen-minute blocks. I call it "fragmented football."

Instead of holding one shape from start to finish, big teams split the match into small blocks, each with a clear objective: the first to stabilize, the middle to press for a goal, the last to preserve or deliver the decisive blow. This leads to an indicator I track closely: goals in the first fifteen minutes versus the last fifteen. At several recent tournaments the gap has narrowed considerably, meaning teams score more evenly and thus collapse less.

But there is a downside. When both teams play fragmented football, the match becomes a chess duel in midfield and the average goal count falls. I once watched a quarter-final where the two teams' combined xG did not reach 1.5. Was that defensive quality, or the timidity of both? This is a question the data cannot answer, and the human eye must take responsibility.

Why the giants collapse in the knockout rounds

I have spent years answering this question. And my answer, after many revisions, lies not in tactics but in two words: expectation.

A big team enters the knockouts under pressure that cannot be numbered. An entire nation places itself on their shoulders. The media counts how many years they have gone without a title. Fans take advancement for granted. This pressure turns simple touches into complicated ones, and borderline decisions into hesitation.

Conversely, a smaller team enters the match with the mentality of someone who has nothing to lose. They accept sitting deep, accept ceding the initiative, and wait for the moment. In Spain's case, they met such a team in the round of sixteen, dominated possession, completed more than a thousand passes, and were eliminated on penalties. All tournament they were the team with the most possession, and that was precisely why they were dragged into a match where what they could not control was their composure from the spot.

This is where I want to dedicate a paragraph to questioning myself. As a "cautious doctor of tactical models," I always want to speak first, speak clearly, speak with evidence. But I know that excessive caution turns me into someone who never dares to judge. So I will say it plainly: the probability of a possession-dominating team winning a major tournament, by my observation, is falling. I estimate such a team has roughly below a 20 percent chance of reaching the final, depending on the quality of its defense. This is a wide confidence interval, and I accept I may be wrong. But I will not stay silent merely for fear of being wrong.

The counter-intuitive angle: correlation is not causation

This is the part I write for myself, as a reminder.

There is a mistake I once made and watch many colleagues make: reading a correlation and assigning it a cause. For example, I observed that champions tend to have a low average of goals conceded. I want to conclude at once that good defense leads to titles. But the truth is more complex. A champion concedes little partly because they are the champion — meaning they play fewer difficult matches, meet weaker opponents in the group stage, and rest more. A low goals-conceded count is a consequence, not entirely a cause.

I once did the same with xG. In one match my team had 0.4 xG while the opponent had 2.1. I published that my team deserved a heavy loss and was called a traitor by home fans. Three rounds later, my team truly collapsed, losing three straight in exactly the messy defensive style I had described. But that does not mean my analysis was right about everything. It only means that in that case, process data correctly forecast the future result. Had my team won the next three, I would have been a liar in the public eye, with the very same numbers on the page.

Three traps I always remind myself to avoid.

First, mistaking the silence of data for the absence of truth. When an indicator says nothing, it may be because the match was truly balanced, but it may also be because the indicator is not sensitive enough. A midfielder can play brilliantly without a single assist. If I look only at assists, I will wrongly conclude he was invisible. I must return to video, to context, to the question: is this data silent because of a gap, or because I am asking the wrong question?

Second, worshiping numbers enough to forget people. Before each analysis I ask myself: which variable am I ignoring? Is it his headache after a sleepless week? Family trouble? These do not sit in the model, and for that reason they break the model.

Third, being so alone in probability that I lose connection. I write for those who believe in probability, not for the crowd of fans. But that belief has a price. It makes me doubt every collective joy. And I have learned I must end each piece with a translation — rendering numbers into human language — so as not to lock myself inside spreadsheets.

On penalties in the 88th minute

There is a line I use often: a missed penalty in the 88th minute has less to do with technique than with the three seconds of silence before the player sets the ball down.

I have watched many shootouts. And what I have realized is that players who score successfully are often not the strongest physically but those with a fixed routine. They walk to the spot with the same number of steps, breathe at the same rhythm, place the ball the same way. Repetition creates calm, and calm is what beats the goalkeeper. This is data I call "habit data," and it appears in almost no statistical table.

Conversely, a big team full of stars is prone to failing in a shootout, because too many stars want to shoot their own way. The order of takers becomes a silent struggle, and the ego defeats collective discipline. In that final, when I reviewed the shootout, I did not see technique. I saw the loneliness of each man before the spot.

This is why I never predict shootout results with a data table. I speak only of trends, of probability, of what happens in roughly three-quarters of cases. The rest I leave to human beings.

The layer of data the media does not see

At every major tournament there is a layer of data the media barely touches: the transfer and valuation layer. A player who performs well at a major tournament can triple his value within three weeks. A player who performs poorly can lose half of it. At a recent tournament, a handful of young talents emerging from less-watched football nations forced big clubs to pay record-breaking fees. I track this as I track a market index, because behind each contract lies a chain of data decisions.

But this is also where I must say what I believe: live data supplied to betting companies is the darkest side effect of the digitization of sport. Every indicator I publish can, unknowingly, become raw material for a market where the data does not belong to the fans. Numbers do not score goals for anyone, but neither do they make anyone better. That is the burden those of us in this profession must carry.

On schedule pressure and the thing called load management

I want to speak plainly about a subject I consider massively romanticized: load management and return from injury.

In theory, load management is science — reducing minutes to protect players. But in practice, I have witnessed many times when load management is turned into an excuse to make room for commercial tours and friendlies. A player is "rested" in an unimportant match, then pushed onto a transcontinental flight to pose for photos and sign advertising deals. Injuries come not from playing too much, but from playing at the wrong time, resting at the wrong time, and being treated as a revenue line.

At a major tournament, teams often play seven matches in about four weeks, with constant travel. This is a perfect formula for soft-tissue injury. And injured players tend to fall in the positions that run the most: full-backs and central midfielders. I have repeatedly published distance and sprint-speed indicators to warn, but a warning only helps when people want to hear it.

Two concrete examples of where data stands

Example one is the case of an African national team at a recent major tournament. They reached the semi-finals — something with no precedent at that level — thanks to a low defensive block, discipline, and maximum use of set pieces. Looking at attacking indicators, they had nothing remarkable. But looking at defensive indicators inside the box, they were among the tournament leaders. This proves what I always believe: a small team can go far if it accepts playing in a way suited to its resources, instead of imitating the giants.

Example two is the case of an Asian team that shocked the group stage by beating two former world champions in succession. I followed this team for months before the tournament, and the indicator that caught my eye was not goals but the number of balls won in the opponent's half. They pressed immediately after losing the ball, within the first five seconds, and that broke the passing structure of big teams. This is a physically demanding model, and I asked myself whether they could sustain it into the next round. They could not, and that does not diminish the value of the wins before.

Why I write less over time

At 56, I write less than before. Not because I am lazy, but because I have understood that every word I set down carries weight, and most of that weight is responsibility.

Five years ago I could write three pieces a week. Now I write one, sometimes taking two weeks. I read it again, I check data sources, I ask myself whether I am describing a real discovery or merely retelling what everyone knows. Caution has not made me less sharp; it has made each judgment worth more.

I choose silence before data when the data is not yet ripe. This makes some colleagues call me slow, but I do not care. A prediction is not seeing the future, but reading the dislocation of the present. And reading dislocation demands patience.

The precondition for applying any model

After the failure of my full-back paper, I added a section to every analysis of mine, one the media sometimes skips: the preconditions for application.

A tactical model from a major tournament cannot be transplanted wholesale onto a domestic club. It needs four conditions: matching fitness, suitable personnel, a suitable development system, and a coaching staff patient enough. Missing one of the four, the model becomes a pretty but useless sheet of paper.

I once forgot this and paid the price. Now I repeat it to myself every time I read a data table about a distant team. Data speaks about its own context. What context am I in? This is the question I must answer before saying anything.

The moment I understood that data never steps onto the pitch

Lach Tray taught me that xG never steps onto the grass. I have said this line many times, and each time I say it, I remember a rainy afternoon in Hai Phong, sitting in the stands watching players warm up. No table told me a player was tired. No model told me the captain doubted himself. I saw it only in how they looked at each other.

This is the final point of every analysis. Every number, however accurate, stops at the edge of the pitch. A match without spectators is a mirror — look into it, and every model is distorted. And I have learned that the distortion is not a flaw of the model, but the nature of human beings.

At 56, I have stopped believing in numbers — but I believe in the way numbers are betrayed.

A judgment looking forward

When another major tournament comes, I will again open my data table, again track each indicator, again write scrawled lines in the margin. But I will spend more time on the silent layer — the layer that appears in no report.

I will notice how a young player breathes before stepping onto the pitch. I will notice how a captain looks at his teammates after conceding. And I will remind myself that the truth of a match, in the end, lies between the numbers, not inside them.

Football is probability, the script is human, and human beings never read the script.

So the question I carry into next season is not which team will win. The question is: when the data table has said everything, what remains on the pitch — and will I be quiet enough to see it?