From Vietnam U20 in 2026 to Brazil in 2026: Three Times the Data Forced Me to Rewrite

One Point, Zero Goals, and a Save Key I Never Pressed

2:47 a.m., May 28, 2026. Saigon was raining. My Excel file had 41 rows, each one a midfield sequence from Vietnam U20's group-stage matches at the FIFA U-20 World Cup in South Korea. My index finger rested on the Save key and did not press down. The article I had published three days earlier was contradicting the data I had just counted by hand.

Three days before, I had written a piece titled "Stop Praising U20's Spirit, Dissect Their Tactical Brains." I called head coach Hoang Anh Tuan's deep defensive block cowardly. I demanded a high press. I was writing as a 22-year-old journalism student who believed that his own intense feelings were evidence. More than 200 comments arrived within two days. Most of them were not about tactics. They called me a traitor to Vietnamese football.

Then I re-watched all three matches, rewound every sequence, counted every misplaced pass, counted every time Vietnam's midfield let the opponent circulate the ball with no one closing them down. The midfield's pass accuracy in the group stage was 38 percent. That number did not excuse the coach. It was an indictment of someone else entirely: me.

What I wrote about U20 was not wrong — the way I proved it was.

Six years later, in a small studio on Nguyen Thi Minh Khai Street, I retold this story on episode 214 of my podcast. By then I had made the exact same kind of mistake twice more: once with Germany at the 2026 World Cup, once with Brazil at the 2026 World Cup. All three times, my headline was right. All three times, my stated reason was wrong. This article is the audit of those three failures — and of a football culture that still analyzes with its eyes.

When an Entire Football Culture Analyzes With Its Eyes

Vietnamese football has a long-standing habit: argue from the result, then go looking for a reason. When a team loses, the entire media apparatus goes hunting for someone to blame. When a team wins, it goes hunting for a hero to celebrate. In both cases, the operational question — what did this team actually do on the pitch, how many times, at what efficiency — is rarely asked.

I grew up inside that environment. In 2026 I joined the sports department of a television station and learned the discipline of writing from early-career observation. But the discipline of observation is not the discipline of measurement. It took 200 people insulting me and my own data betraying me before I could tell those two things apart.

The problem with Vietnamese football is not a shortage of emotion. This football culture has an excess of emotion. The problem is that emotion is used as a substitute for method. When a national team enters a major-tournament cycle, that emotion gets compressed until almost every judgment becomes a judgment about identity: who we are, how we play. Very few judgments become judgments about mechanism: which line got rotated, where, and how many times per half.

During the period when I wrote the U20 piece, detailed domestic data barely existed in public. Vietnamese stats sites stopped at goals, cards, and possession. Possession is the most harmless metric of all — it tells you who held the ball and nothing about who controlled the match. A team can hold 65 percent of the ball and lose 3-0 while nobody in the press room asks what that ball was used for.

That was the ground I had to work on. And that is exactly why my three failures were useful. They showed that the problem is not a lack of numbers. The problem is reading numbers on the wrong layer.

Failure One: Vietnam U20 and the 38 Percent Pass Accuracy

Back to the 41-row spreadsheet. When I finished counting, the picture looked nothing like what I had assumed.

Vietnam U20 finished the 2026 U-20 World Cup group stage with a single point and no goals scored, against France, Honduras, and New Zealand. The popular read at the time was: weak team, deep block, defeat is reasonable, and if anything is worth mentioning it is spirit. I objected to that read. But I objected with a wrong-layer read of my own.

What I counted was not "the team did not dare to press." What I counted was that the midfield could not sustain a pass under pressure. Thirty-eight percent accuracy means two out of every three passes went to an opponent's foot or into empty space. That is a technical problem at the individual and structural-line level, not a courage problem.

And when a midfield passes at 38 percent, choosing a deep defensive block is a logical consequence, not a cowardly choice. You cannot press high if you cannot keep the ball for more than three passes. A high press requires something that U20 side did not have: the ability to convert a ball recovery into a moment of control. Demanding a high press from a team passing at 38 percent is like demanding that a non-swimmer jump in to save a drowning person. The act looks brave in my article and looks like a disaster on the pitch.

This is what I learned, and it shaped everything I have done since: the symptom on the pitch is what the team does; the diagnosis is what the system that produced that team does. A deep block is a symptom. The technical foundation for passing under pressure at youth level is the diagnosis. Demanding a change in the symptom without touching the diagnosis is demanding something that cannot exist.

My second article admitted that my earlier criticism had been shallow. I attached a chart I drew myself in Excel, with the vertical axis showing misplaced passes per minute. None of the 200 people who had attacked me came back to praise it. But three months later, a young coach at an academy in the north messaged me asking for the raw file. That was the moment I understood that data has something a hot take does not: it exists outside the argument.

Failure Two: Germany 2026 and the Trap of the Single Cause

In 2026 I was interning at a digital sports outlet. Germany were eliminated from the World Cup in Russia in the group stage: a 0-1 loss to Mexico, a 2-1 win over Sweden, a 0-2 loss to South Korea, two goals in three matches. I started writing immediately.

My thesis: Joachim Low was wrong to use Thomas Muller as a false nine. My evidence came from the South Korea match — Muller scored no goals, provided no assists, and touched the ball only 21 times. The piece was shared more than 1,000 times in the first two hours.

Then I opened the detailed event data and looked again. And I realized I had just committed exactly the error I had committed with U20: pick one cause, then go find the number that defends it.

Germany's missing No. 9 was a symptom, not a diagnosis.

Germany's operational problem in Russia lived in the pressing block behind the ball. Opponents were allowed an average of 14.2 passes per possession before Germany closed them down. According to the data I cross-referenced, that was the highest figure among teams eliminated in the group stage that year. In other words, Germany did not apply pressure. They let opponents circulate freely in the middle third, and by the time the ball approached Manuel Neuer's goal, every positional error had already accumulated twenty seconds earlier.

Muller touched the ball 21 times not because Muller played badly. Muller touched the ball 21 times because the ball never reached Muller's zone. And the ball never reached Muller's zone because Germany's midfield could not win it back quickly enough. I had read a metric from the final layer of the causal chain and labeled it the first cause.

I agonized for a week, read ten more analyses, and had to publish a correction. Being wrong at the 2026 World Cup taught me more than being right for an entire season.

From then on I adopted a rule I cannot break: before locking in a thesis, establish clearly which layer is the symptom and which is the mechanism. I started using metrics like xG, passes allowed per defensive action, and high-line pressure as a way to check myself, not as a way to decorate an article. Readers do not need to know the name of the metric. Readers need to know that my conclusion can be verified by someone else.

Failure Three: Brazil 2026 and the Humility of a Model

In March 2026, all of European football stopped. My podcast fell into crisis: listens dropped from 8,000 to 1,200 per episode within a month. I did not panic. I retreated into research, the way people in this profession do when they lose the stage and go back to the desk.

I downloaded Serie A, Bundesliga, and Premier League data and simulated classic matches using passing maps and positional heat maps. I made a special episode with a premise-driven title: if you abolished the offside law, football would become basketball. I pulled 27 disallowed goals from the 2026-20 Premier League season as raw material. That episode reached 42,000 listens overnight, five times my previous record.

When the stadium empties, the noise disappears and the data starts talking.

Colleagues told me I got lucky. I nodded and quietly built a spreadsheet projecting the impact of rule changes. But what I actually learned was bigger than one successful episode: I discovered that the fear of being forgotten is the best fuel for analytical work. It forced me to generate topics from historical data instead of waiting for news to arrive.

By the 2026 World Cup in Qatar I was the lead host. I declared on air that Brazil would stop at the quarter-finals because Richarlison was not a pure No. 9. I cited the fact that he had scored three goals but generated only 0.8 shots per match when playing behind the defensive line.

Brazil met Croatia in the quarter-finals, held 58 percent of the ball, and lost on penalties 2-4. My headline was right. The joy of a correct prediction evaporated fast when I reopened the match data.

Richarlison created two chances. The real problem was in midfield, where Casemiro won only 3 of 9 duels. When a midfield loses duels at that rate, any analysis of the centre-forward position becomes an analysis of a number with no operational meaning. I was wrong again — for the third time, in the same way.

I shut myself in a room for three days, built a logistic regression model from xG, pressing intensity, and duel win rates across all 32 teams, and published a survival-coefficient table before the knockout round. That model was later cited by many listeners. What I remember most is not its accuracy rate. What I remember most is that it forced me to write down, in public language, what would make the model collapse.

The Layer Data Cannot Reach

Three failures taught me something very specific: a conclusion can be right about the result and wrong about the cause, and that kind of error is more dangerous than being flatly wrong, because it rewards the writer with the feeling of having been correct.

But if this article stopped there it would be selling you exactly the product I just warned you about: a tidy belief in the power of data. I have to say the rest.

Football does not need you to believe it; it needs you to verify it. And there are things in Vietnam that cannot be verified with the data currently available.

The V.League has no publicly accessible positional tracking data. There is no standard xG model for the domestic league, because xG requires event data with coordinates, and coordinate event data requires a collection system the league has not operated at sufficient scale. That means every tactical judgment about Vietnamese football at the time I write this is still largely a judgment made from video and the human eye. Including my own.

That is the fatal weakness in the position I pursue. I built my credibility on the principle of "cite your sources," but the source here is a foreign source, and the evidence is foreign-league data. When I apply a European model to a V.League match, I am performing an analogy, not a measurement. An analogy can be right in direction and wrong in magnitude — and in football, being wrong in magnitude is usually enough to ruin the conclusion.

There is one more layer models cannot touch. In 2026, when I wrote about U20, I did not know that some of those players read what I wrote. By the time I knew, I had already published. I used to think analysis was a neutral act. It is not neutral. It acts on real people, at an age when a sentence can stay in your head longer than a training session.

That is why I no longer carry the couldn't-care-less attitude of a professional provocateur. I still write sharp headlines. But the body of the piece has to bear their weight.

Where I Might Still Be Wrong

I have to interrogate myself here, because otherwise this article is just a performance of humility.

Possibility one: I may be overcompensating. Three failures caused by emotion could push me to the opposite extreme — believing that if there is a number, there is a truth. But numbers do not generate meaning by themselves. Which metric I pick, what weight I assign, where I cut the sample — all of these are human decisions. A logistic regression is not neutral just because it uses Greek letters.

Possibility two: sample size. A major-tournament cycle gives a national team three to seven matches. At that sample size, almost every difference in a metric sits inside the noise band. The metric I used to draw conclusions about Germany in 2026 may have been one cold night for a 33-year-old midfield. I built a systemic story on a sample too small to support a system. I still do it, with more controls, but I still do it.

Possibility three, and this is the one that worries me most: the chain of evidence can be right while the conclusion is still wrong, because I asked the wrong question. For years I asked "who is responsible." A better question might be "what mechanism allowed this to continue." The first question leads to blame. The second leads to redesign. Players create moments; systems create players. And a system has no head coach to sack.

Every argument has one layer of data that has not yet been turned over. My question is whether I have the patience to turn it over before going on air, or whether I turn it over after the piece has already been shared a thousand times.

What I Will Verify Over the Next Three Seasons

I am not finishing this with a summary. I am finishing with a falsifiable claim.

Within the next three seasons, I predict at least one open-source xG model for the V.League will appear, built by Vietnamese people, using event data collected by the community itself. The condition for this prediction to be right: a group of three or more people agrees to record data manually for at least one full season, unpaid, with no compensation for their time. The condition for it to be wrong: the league buys a commercial data package before the community gets there, at which point the domestic data question becomes a budget question.

And I predict one more thing, verifiable within the current season. The national team in the next major-tournament cycle will be judged mainly on results, as usual. But within 48 hours of the decisive match, at least one domestic analysis will cite operational metrics — passes allowed per defensive action, chances created from set pieces, midfield duel win rate — instead of only possession and shot counts. If that happens, domestic analysis has shifted one layer. If it does not, we are still where we were, and I still have work to do.

Each of my three failures took me about a week to digest. A transfer deal is only truly cheap when you look at it three seasons later. So is a thesis. I no longer demand to be right immediately. I only demand to be verified.