Heat Maps, xG and the New Divination of Modern Football

On the second monitor of a Shenzhen newsroom, the heat map of a right winger glowed into a long streak down the right flank. The system logged 41 touches in that zone, the most in the match. The post-match report called it "a flawless performance controlling the right channel." The next night I rewatched all 90 minutes and counted 38 of those 41 touches as sideways or backward passes, in a game his side had led by two goals since the 55th minute. Not one dribble. Not one pass through two lines. The heat map was not wrong. It simply drew something different from what the report claimed it drew.

I tell this story for a reason other than mocking data. In nearly fifteen years of watching and writing about football, I have never seen a good tool misused with such confidence. People no longer argue with their eyes. They argue with screenshots.

In modern football, data has stopped functioning as evidence and started functioning as decorative language. The danger of this trade is not wrong numbers. The danger is analysis presented too neatly, too completely, too confidently — while what sits under the paint is empty space.

Context: thirty years from notebook to tracking screen

Football entered the data era along a fairly clear path. From the early 1990s, event-data companies began logging every action: who touched the ball, where, when, and with what result. By the 2010s, event data became an industry standard, and in the second half of that decade positional tracking data at hundredths of a second arrived in Europe's top leagues.

In Vietnam the wave arrived later but fast. V.League clubs began putting GPS vests on players in training. The Vietnam Football Federation built analysis units for the national teams. After each round, domestic sports sites began publishing expected goals, heat maps and red-and-blue comparison tables.

What matters is that the speed of adopting the tool far outpaced the speed of understanding it. A coach needs years to read one metric correctly. An editor needs weeks to put that metric on the front page.

That gap produced a concrete consequence: heat maps became illustrations for every article, and expected goals became the verdict in every argument. Whoever had the higher number won the debate. Whoever had the redder map was called better. Nobody asked under what conditions that number was produced.

There is an economic force behind this, and I think it matters more than the technical one. Sports content now runs on a brutal rhythm: the final whistle blows, and within twenty minutes readers expect analysis. Nobody has twenty minutes to rewatch the tape. Everyone has twenty minutes to open a pre-built template with boxes, tables and ready-made conclusions.

That is where the risk appears. When the input data is empty or has not loaded yet, the system does not stop. It keeps running. And it still produces something that looks finished.

Core: what a heat map draws, and what it hides

Start with the mechanism. A heat map takes the coordinates of every touch or recorded position of a player, then smooths them with a distribution function to create coloured zones. Redder means denser. Mathematically it is a spatial interpolation. Technically it is a way of presenting positional data.

The problem is that a heat map measures exactly one dimension: where the player was. It does not measure what he did there, and even less what he opened up for others by standing there.

Think about what falls outside the map. A midfielder dropping deep to drag a defender out of position so a teammate can run into the vacated space leaves no trace on his own heat map. A centre-back stepping up to cut a passing lane and force a change of build-up direction shows no colour at all. A striker repeatedly moving behind the defensive line to create pressure so the opponent dare not push up has an almost empty map.

A heat map is a map of where a player stood, not a map of where the match was decided. Those two things overlap only in a handful of cases, and the media has silently assumed they always overlap.

There is another layer of error rarely discussed. Event data only records what the coder decided counted as an event. An off-ball duel can be skipped. A block where the ball does not visibly change direction can go uncounted. A pass cut off by indirect pressure can still be logged as completed. Every dataset has its own conventions, and conventions never appear on the graphic.

The Giroud case: when the number contradicts the match

I want to return to the 2026 World Cup, because it is the cleanest example for this whole argument.

Olivier Giroud played the entire tournament as the centre-forward of the champion. He finished it without a single shot on target — a statistic widely published across international data platforms and mocked for years afterwards. His heat maps in the knockout rounds showed almost no significant red inside the opposition box. Judged only on positional data, he looked harmless.

But watch the tape and you see something else entirely. Giroud kept standing between the two centre-backs. He pinned them. He contested aerial balls in positions that did not help his own scoring but helped his team keep the ball high up the pitch. Much of the space Antoine Griezmann and Kylian Mbappé exploited throughout the tournament was created by Giroud refusing to leave his spot.

That is why I wrote the piece headlined "France won the 2026 World Cup thanks to a false number 9." The backlash was fierce. A group of young coaches argued online that I was manufacturing scandal for attention. Only after France beat Croatia 4-2 in the final did some international analysts begin talking about the centre-forward as a mobile decoy.

The false number 9 does not exist on the pitch, but it lifts the trophy.

I took a technical lesson from that episode, and it has followed me for years. Writing about Giroud, I could not simply say he was good. I had to supply what the data did not display: which defender he pinned, at which minute, with or without the lead. Without those details, my argument was just a poem with page numbers.

xG: probability is not quality

Expected goals is a model that calculates the probability of a shot becoming a goal based on position, angle, body part and other factors. It is genuinely valuable, especially when judging process over a long enough period.

But it has three limitations most sports writing ignores.

First, every provider uses a different model. Two identical shots can produce two different values on two platforms. Readers have no way of knowing which model they are looking at, because the headline simply says "xG."

Second, the model does not know who is shooting. From the same position and the same angle, a world-class forward and a centre-back out of his depth do not share a probability. The model averages, and averaging always erases the individual.

Third, and most importantly, xG counts only shots that happened. It is completely blind to chances that never became shots — a pass read by a defender, a run anticipated, a decision to hold the ball half a second too long. This is the biggest blind spot in every football data model today: they measure consequences, not what was prevented before a consequence could occur.

In the 2026 World Cup final I filed a piece the moment Lionel Messi opened the scoring in the 23rd minute, arguing the goal came from individual French defensive error rather than Argentine tactical quality. When the match ended 3-3 and Argentina won on penalties, my piece was mocked hard. Only when I rechecked the data did I see I had missed something: Messi had three shots on target and created five chances, the most in the match. I publicly corrected the article.

My error was not doubting Messi. My error was concluding before rechecking the source. That is a writer's failure, not a metric's.

PPDA and the trap called game state

Another metric is constantly quoted without its conditions: PPDA, the passes an opponent is allowed per defensive action. Lower means more aggressive pressing.

The catch is that PPDA depends heavily on game state. A team leading by three sits deep and concedes the ball, sending PPDA soaring — and gets called passive. A team chasing the game pushes up and posts a very low PPDA — and gets called intense, when in reality it is desperately chasing.

One metric, two opposite stories, and both get written without anyone noting the scoreline at the moment of measurement. I have seen PPDA tables comparing two teams from different matches, different rounds, different game states, with a conclusion printed underneath as if it were a controlled experiment.

When the input is empty, the system still produces an answer

This is the section I most want people in the industry to read closely.

I once saw a scouting report designed down to the last cell. It had nine sections: tactical analysis, financial analysis, results analysis, league-position analysis, compliance analysis, management and dressing-room analysis, risk analysis, media analysis, industry transmission analysis. Each section had a table. Each table had rows. Each row had a conclusion.

But reading to the end, I realised something: not a single player's name appeared in the whole document. No club. No match. No date. Every cell was filled with the same phrase in every position.

The report was not formally wrong. It was simply empty of content while retaining the full visual authority of a vetted document.

That is the danger. Empty data carries an illusion of neutrality. It is an invitation to fabricate, and the football industry accepted that invitation long ago.

When no player is named, an analysis system does not stop. It runs all nine sections. It still delivers a document that looks structured, looks methodical, looks verified. And if the reader does not go line by line, they will believe the void is a conclusion.

I call this the inverted false-negative effect. In medicine, a false negative is when disease exists but the test says otherwise. In football analysis, it is when nothing was screened at all, yet the report states no issue was found. Those two states differ fundamentally. One is safety. The other is never having looked.

Transfers: buying a story, not a person

The transfer market is where this mechanism operates most brazenly.

In 2026, while Asian football media fixated on Kylian Mbappé staying at Paris Saint-Germain, I noticed a small detail: Erling Haaland's representatives had hired a law firm based in Manchester to handle his image rights. I contacted a source close to the Borussia Dortmund coaching staff and confirmed that a 60 million euro release clause had been activated. I published an exclusive before Manchester City made it official on 10 May 2026.

What I learned was not that I was clever. What I learned was that I had checked a chain of facts instead of writing on instinct. I built a source file, cross-checked the agent's transaction history, and published with the phrase "according to a source close to the situation" rather than claiming the contract was in my hand.

A transfer does not buy a player — it buys the story people want to believe. A transfer truly fails only when the story sold to the market does not match the truth on the pitch. And in most cases the two are never compared.

Esports: where regulation lags behind data

Meanwhile, another sector is showing what happens when data and money move faster than oversight.

I have followed esports from when it was a small scene to when it became a billion-dollar industry. Esports betting is eroding competitive integrity faster than traditional sport, because there are dozens of times more matches, each is shorter, the audience is younger, and the regulatory machinery still moves at the pace of traditional sports federations.

That pace is exactly the pace of a broken analysis pipeline. Data flows in faster than people can check it. And when the checking speed is lower than the generation speed, the gap is always filled with belief.

Vietnamese football: where the biggest gap lies

Based on my experience watching matches in both the V.League and national team competitions, I think Vietnamese football sits in a peculiar and actually favourable position: we adopted metrics late, so we still have a chance to adopt them properly.

Public data for the V.League remains thin. Basic per-match metrics exist, but detailed event data at the level of individual actions is barely released. The immediate consequence is that domestic analysis still leans on feel — not necessarily bad, but it cannot generate information gain, the value of telling readers something they did not already know.

The long-term consequence is the concern. When a football culture adopts a tool late, it usually imports the bad habits of those who went first. We may soon see articles full of heat maps and indices, written by people who never rewatched a single half, ending in claims that cannot be checked.

The national team's matches at AFF Cup tournaments, in World Cup qualifying and at Asian U23 championships have shown the opposite pattern, and it deserves to be copied: a low-block defensive shape combined with fast transition has produced results beyond expectations for years. Judged on possession alone, Vietnam look weak. Judged on how the lines move during transitions, the picture flips entirely.

That is a lesson for writers, not for coaches.

The contrarian case: where I might be wrong

I have to be explicit here, because a piece sceptical of data that refuses to limit itself is just a lazy piece written at length.

There are places where data has created real, large, verifiable value. Brentford and Midtjylland are famous examples of using models to find players in markets the naked eye never scans, and both lifted their clubs sustainably rather than through luck. Brighton is another: buying low, selling high, and maintaining squad quality across multiple managerial changes.

Applied sports science is the same. Load management built on positional data has reduced muscle injuries at many clubs. Early detection of overload signals before an ACL tear is an achievement nobody can deny.

And I have erred in my own way. In 2026, working as a field reporter at the U20 World Cup in South Korea, during the match between Vietnam U20 and France U20, I mispronounced striker Jean-Kévin Augustin's name three times in the first half. Viewers called in to complain. I had to rewatch the whole tape and take notes on every action to understand how imprecise I had been.

The pitch never lies — only I once misheard a name.

That name I got wrong is the most expensive lesson journalism ever gave me.

The risk I actually worry about is different: scepticism about data can be read as hostility to knowledge. I do not want anyone using this piece to justify refusing to learn a new metric. People are not wrong to use heat maps. They are wrong to use them as a substitute for watching the tape.

And I must leave another possibility open. Perhaps I am exaggerating, because most fans do not read tables and only remember what they saw. Perhaps sports journalism has always been this messy for decades, and only the tools have changed. I have no way to rule that out, and I do not want to.

The only thing I am sure of is this: a beautifully presented conclusion is not the same as a verified one.

Takeaway: a minimum ruleset for readers and writers

If I had to offer a minimum ruleset, it would be three things.

For writers: every time you put a metric in a piece, state the conditions that produced it. Who supplied it, what game state it measured, and over how many matches. Three lines are enough to kill most false conclusions.

For readers: when an analysis looks too perfectly structured, count how many players are named specifically and how many actual passages of play are described. If neither appears, you are reading a scaffold, not an analysis.

For newsrooms: when input data is empty, publish an empty line. A piece that stops at the right moment is always better than a complete piece filled with guesswork. In this trade, having no answer is a valid answer; inventing one is not.

The silence after the whistle is the paragraph I most like to write. It is the only window in a working day when I can hear the match rather than hear myself narrating it.

I do not write to be loved; I write so that others have to stop. This time, what I want people to stop is a habit: the habit of believing a beautiful graphic is a verified fact.

When the stands are empty, I hear the breathing of the match — and I find my own voice. The question I leave for this season is simple: if every heat map vanished from every article tomorrow morning, would we still be able to describe a football match with our own eyes?