Eight Sections, Zero Data Points: The Trap of Sports Analytics

A report of eleven pages sat on my desk in Chengdu at the end of October, when the domestic season had finished and the transfer window had not yet opened. It had all eight sections: technical-tactical matchup analysis, athlete condition, event and organisational context, revenue modelling, rules-compliance framework, health risk matrix, public-narrative analysis, and an industry transmission diagram.

Every section had tables. Every table had three columns: assessment, benchmark, notes. And every cell in those three columns repeated one sentence: insufficient information to assess.

I read all eleven pages. Not one athlete's name. Not one fight, match or bout. Not one date. Not one figure. But the skeleton stood there anyway, aligned, fully sectioned, like a stadium built long ago into which no team had ever walked.

People assume the hard part of this profession is finding the answer. After thirty-eight years I believe the hard part is knowing when to say you have nothing yet to say. I saw this once before, on a training pitch on the edge of Chengdu, and it taught me more than any match I have watched.

When the data industry pays to fool itself

Over roughly eighteen months, the volume of sports analysis produced across Southeast Asia has grown in a way I have not seen in my career. Most of it is not attached to a specific match. It is attached to a topic.

A topic is always available. A match has to be waited for. And waiting is the one thing nobody has patience for.

I have watched how sports desks in Vietnam operate across the last two seasons. A young reporter in Hanoi told me he has to file three analytical pieces a week, while the V-League plays one round. Three pieces for one round. That means two of the three must be about things that have not happened yet, or happened so long ago that nobody remembers.

The most common workaround is to keep the skeleton and swap the contents. The skeleton is never wrong, because a skeleton only asks. The contents can be wrong, so people leave the contents empty. A piece with eight sections, each with a professionally correct heading, each closing on a neutral line like "more data is needed" — that piece looks like a reference document rather than an article, and it is perfectly safe.

Perfect safety is the worst signal in this trade.

Because sports analysis, in its honest form, is an act of risk-taking. You stake your reputation on a judgement that the pitch can overturn in ninety minutes. If you never stake anything, you are never overturned, and you never say anything either.

A neatly presented empty analytical framework is more dangerous than a wrong opinion piece, because it never incriminates itself. A wrong opinion is erased by an 88th-minute goal. An empty framework survives forever, is cited, is archived, and becomes the "background document" for the next empty framework.

Thirty-eight years at the desk have taught me that three technical faults sit behind every empty analysis. None of them concerns talent. All of them concern discipline.

The first fault is failing to extract information points. An information point is the smallest atomic factual unit an article produces: a name, a date, a score, a fee, a quote. One article can run three thousand words and contain two information points. Another can run four hundred words and contain eleven. Which one carries more analytical value needs no debate.

The second fault is failing to extract entities. Without names of people, teams, events and governing bodies, there is no chain to link. A chain is the only thing that lets you say "if — then". Without a chain, every "if — then" statement is a fabrication with good prose.

The third fault, and the one I fear most, is failing to determine the class of the analysis subject. A piece about performance martial arts needs difficulty-score and presentation-score logic. A piece about mixed martial arts needs win-loss and finish-rate logic. A piece about football needs expected goals, running volume and structural logic. Apply one logic to another subject and you are not wrong in the details; you are wrong in the entire coordinate system. Detail errors can be fixed. Coordinate errors turn every correct data point into noise.

I once made exactly that error. It cost me a two-thousand-word article.

Three centre-backs, and the fear of being named

Before I continue with my own mistake, I want to address what I have tracked most closely over the last two seasons: the return of the back three.

In the 2026 V-League season, at one point seven of fourteen clubs fielded a back three in at least half their matches. In 2026 the number did not fall. Analysts called it a tactical advance. I watched the training sessions, and I saw mostly a defence of reputation.

It unfolds almost identically. A team loses three matches, two of them conceding from crosses into the channel between the centre-backs. The coach comes under pressure. The following week the team plays with three centre-backs. Results usually improve for two or three rounds, because opponents have not adjusted. Then the attack weakens noticeably because a forward slot has been sacrificed, and the team starts drawing more than it loses.

Drawing more than losing is the real objective of a great many tactical decisions in Southeast Asian leagues. I do not say this as a criticism. I say it as a risk-management fact.

A back three does not make a team defend better. It disperses media pressure. With a back four, when you concede from a cross, people point at the centre-back who erred and name him. With a back three, people talk about the system, about the space behind the wing-backs, about the wide midfielder who failed to track back. Nobody is named. The system absorbs the punishment on behalf of the individual.

The price is paid elsewhere. The two wing-backs in a back three must cover eleven to thirteen kilometres per match, four to five of them at high speed. In Vietnamese conditions from April to August, at one match every three days, that is a load very few domestic players can sustain across a full season.

I counted. In the 2026 season, at one V-League club with a fixed back three, their two wing-backs missed a combined nineteen matches with hamstring and foot injuries. In the same club's previous season, before the switch, the figure was six.

When analysis presents the back three as progress, it is reading the consequence as the cause. A formation does not create strength. A formation redistributes risk, and here the risk is pushed from the centre to the flanks, from the centre-back to the wing-back, from the man with a voice to the man without one.

What is worth tracking in the second half of the season is not which team plays a back three, but which team has the wing depth to sustain it through July. That is a data question, not a philosophical one.

Transfers: where value is replaced by price

In the transfer market, data is distorted in a different way, and it connects directly to the empty report I read that night.

A transfer window in Southeast Asia has a feature European football journalism rarely meets: most major deals have no disclosed fee. Some have a fee, some have "an undisclosed sum", some are player swaps plus cash, and some are loans with a conditional purchase clause tied to minutes played.

When the input data is opaque, people fill the gap with narrative. And narrative always leans toward the big club.

I followed one specific case in the 2026 season. A small club in central Vietnam sold a twenty-year-old to a team with six times its budget. The media reported it as a "record deal for the small club", complete with a description of the player's potential. Three months later he had started four matches. Total minutes: three hundred and seventeen.

Over the same period, the small club used most of the money to clear unpaid wages and upgrade the stadium floodlights, because they would not be licensed to host home matches without meeting the lighting standard.

The analysis of that deal discussed the player's market value. The analysis should have discussed the cash flow of a club trying to survive into the following season.

Same outcome, insider and outsider watching two different matches.

I have followed small clubs and I believe this: the genuinely valuable deals sit with the teams that cannot afford stars. They are forced to read players more carefully, because they have no budget for error. A big club that buys badly sells or loans. A small club that buys badly loses its foreign-player slot and half a season.

Transfers are where value is replaced by price, and nobody dares say it out loud. That is why transfer analysis reads beautifully and predicts badly.

The offside line and the death of instinct

There is another kind of data changing football in a direction I do not believe is good, and it is also a form of framework overwhelming content.

Semi-automated offside technology has been in use in top European leagues since 2026 and is present in many Asian leagues. It works by tracking thirty body points per player by camera, determining the offside point by the furthest goalscoring body part, and signalling the referee within seconds.

Technically, it is accurate. In terms of the game, it creates something else.

I sat through the tape of a national league match in which four goals were disallowed. Two were disallowed because a toe or a shoulder was ahead of the offside line by less than ten centimetres as recorded by the system.

My point is not the correctness of those two decisions. It is the behaviour of the striker over the following forty minutes. He began starting his runs half a beat later. Not much, but enough for the opposing defender to step up in time.

A good striker does not run at the exact moment the ball leaves his teammate's foot. He runs a quarter-second early, gambling that the defender will not step up in time. That is a calculated gamble, and it is part of the skill. When the system punishes every gamble even when the margin is a knuckle, players learn that gambling is no longer worth it. They stop gambling. And when they stop gambling, they stop running early.

The referee is no longer officiating the match at its own rhythm. The referee is editing the match.

Millimetre offside does not protect fairness; it shifts risk from defender to attacker, and does so permanently. I am not calling for the technology to be removed. I am saying every framework imposed on the game must be measured by what it takes away, not only by what it gives.

The 2026–2026 tape archive: when possession was a warning sign

In 2026, when the pandemic stopped every competition, I entered the worst period of anxiety in my working life. No new matches. No dressing rooms to observe. No trips.

I dug out my tape archive of the national team collected between 2026 and 2026, mostly qualifiers and some international friendlies. Fourteen matches had footage clear enough to count.

I built a table with two columns: first-half possession share, and final result. Nothing else.

The rule I found forced me to rethink how I read results: in ten matches where first-half possession was twelve percentage points or more above that team's own average, they lost the match or surrendered their advantage in the second half in seven. Seven out of ten, or seventy-one percent.

Fourteen matches is far too small a sample to call a scientific conclusion. I know that. But I did not present it as a scientific conclusion. I presented it as an archive mark, a kind of residual curve still visible in the data.

The series was called "The Numbers That Weep" and drew about sixty thousand reads, the highest figure of my writing career. What I did not expect was the professional consequence: a club in Chengdu approached me to consult informally for its analysis department.

Their first assignment was not match prediction. They asked me to review their own possession data across three seasons against second-half results.

The finding ran against the coaching staff's expectations. There were three players whose higher possession share correlated with fewer wins. Not because they played badly. Because they held the ball long, and when one player holds the ball long, the other ten stand still. Standing still is the worst state for a team without the ball, because when they lose it they are not positioned to win it back.

Old files covered in dust, but the curve of that long-range shot from years ago is still visible in the data.

That is what I want to say about the value of historical data. It does not predict the future. It forces you to ask a different question from the one you were asking.

Croatia 2026 and two hundred and fourteen set pieces

In 2026, thanks to a run of field-verified reporting, a television station in Chengdu sent me to Russia to follow Croatia.

Before the semi-final against England, I wrote a two-thousand-word analysis. The central claim: Croatia would lose on fitness. I had sufficient data to defend that claim. Eight of their eleven regular starters were over thirty. They had already played three matches that went to extra time, two of them to penalty shootouts. Their cumulative minutes were significantly higher than England's.

The match was played on 11 July 2026 at the Luzhniki Stadium in Moscow. Croatia won 2-1, held around fifty-eight percent possession, and scored the winner in the 109th minute.

My analysis was wrong in its entirety.

For a month afterwards I rewatched their knockout tapes, all four matches. I logged two hundred and fourteen set-piece situations, attacking and defensive, and marked the start time of each.

What I found lay in the rhythm. In extra time, Croatia did not accelerate. They deliberately slowed, extended their build-up phases, pushed the ball wide more often, and kept their team shape roughly fifteen metres shorter than in the first half. They were not trying to score in the first period of extra time. They were trying to make the first period of extra time pass at the lowest possible fitness cost.

I read the minutes as a sign of exhaustion. They read the minutes as a resource to allocate.

I trusted the arithmetic before I trusted the pitch; that was the most expensive mistake of my career.

Since then, every prediction piece of mine carries a mandatory section titled "the conditions under which I may be wrong". In it I list the assumptions that, if false, collapse my conclusion. With Croatia, my false assumption was this: fatigue is a linearly cumulative quantity. It is not. Fatigue is a quantity that can be managed tactically.

Between two rolling balls there is a silence that contains the whole truth.

The Longfor training ground in 2026 and one thousand two hundred reads

In 2026 I was forty-five, still writing for a print paper in Chengdu. Online sports media was exploding, and a young blogger published a story that a third-tier club in Sichuan called Longfor was going bankrupt. His evidence was a photograph of an empty training ground taken on an afternoon.

I had attended twenty-three consecutive training sessions at that club over the preceding three weeks. I knew they had just signed a sponsorship contract worth two point eight million yuan. I knew that on the afternoon the photograph was taken, the team had trained in the morning because the pitch had to be given over to a youth friendly.

I wrote a rebuttal. I cited the contract figure, the training schedule, the coach's name, the signing date.

My piece ran three days later. Reads: one thousand two hundred.

The blogger's piece, over those three days, had been shared more than forty thousand times.

Fake news does not die because people stop believing it, but because people stop verifying it.

I tell this story not to complain. I tell it because it explains why empty analyses multiply. An empty analysis can be published within two hours. A verified analysis takes three days. Within those three days the topic has died, and the empty analysis has already been accepted as fact.

Since then I have accepted a compromise: short digital dispatches before the long piece is finished, always carrying a source confirmation line. It makes me slower. It does not make me wrong.

The counter-intuitive point about an empty report

Back to the eleven pages on my desk in Chengdu.

The first reaction of most people in this trade on reading such a report is to conclude it is worthless. I think that conclusion is wrong, and this is where I go against the majority.

A report that states "insufficient information" in every cell is an honest report on the state of the data. It does not say the subject is unimportant, risk-free, or inactive. It says we do not yet know.

In this profession people constantly confuse "neutral" with "unknown". Those are entirely different states in terms of what action follows. A team you have no data on is not an average team. It is an unmeasured team. Treat them as average and you will pick the wrong line-up.

I have seen the consequence in recruitment. A player has no data because he plays in a league nobody tracks. He is graded equal to a fully measured player with average metrics. In reality he is an unknown variable, potentially much better and potentially much worse. The club is buying an option, not a mean.

The second counter-intuitive point concerns the writer. I believe a good analytical writer is identified by the number of blank cells he dares to leave, not by the number he dares to fill. A piece filling twenty cells, ten of them speculation presented as fact, is worse than a piece filling seven and flagging thirteen as blank with reasons.

This runs directly against how newsrooms currently score productivity. A blank cell generates no reads. A sourced line of data generates no reads either. Reads come from certainty.

And here is where I doubt myself, because I have lived in both worlds and I know both have a case. I lost three days rebutting a false story for one thousand two hundred reads. I cannot ask a newsroom to fund a business model like that.

So I am not calling for an end to certainty. I am proposing something more specific: a clear separation of three tiers in every piece. Tier one is sourced fact. Tier two is reasonable inference, flagged as inference. Tier three is constructive speculation, flagged as speculation. Readers have the right to know which tier they are reading.

In some sports journalism cultures this has been standard for decades without anyone naming it. They simply never present speculation as fact. Elsewhere the three tiers are blended into one mass and readers lose the means to tell them apart.

The blended three tiers are the structure of that eleven-page report. It looks like tier one, because it has tables. It behaves like tier three, because every cell is empty. And it never declares itself tier three.

Signals to track in the period ahead

What I will be tracking over the next six months is not in the league table.

I will watch whether any outlet in Vietnam begins publishing the number of blank cells in its own analysis, the way public health bodies publish the certainty level of a recommendation. It sounds small. It is not small.

I will watch whether any V-League club begins recording its own training data instead of buying data from outside. A club's own training data is the only kind that cannot be bought with money, and the only kind that cannot be copied.

I will watch the wing-backs at clubs playing a back three, by counting their minutes through June and July. If the number falls sharply against March, the system is eroding itself.

And I will watch for analyses with no player names, no dates, no scorelines. If they keep appearing at the current rate, it means the industry has agreed to trade reliability for volume. When an industry agrees to that, it does not lose readers. It loses the ability to notice that it is wrong.

I am old now, but my count is still twenty years old. And that count tells me the only thing still worth writing in the period ahead is what remains verifiable: a training session attended in person, a contract with a date on it, a striker starting his run a quarter-second before the ball leaves his teammate's foot.

The rest is skeleton. Anyone can build a skeleton.