The Swallowed Tone Mark: The Data Gap That Makes Vietnamese Football Invisible

September, 22:40 Barcelona time. I pasted a 1,180-word V.League 1 match report into the entity-extraction model I use for every piece of analysis I write. The report had player names, club names, the scoreline, bookings, substitutions. The model returned exactly one line: entities: 0.

I ran it again. Still entities: 0. I switched to a report from a different outlet, same matchweek, same language. This time the model recognised three names. Two of them were truncated at the tone mark: a Vietnamese surname lost its tail and was replaced by a black square; a club name split into two separate entities simply because the newsroom was using a legacy encoding.

The model was not broken. It was telling the truth. And what it said was this: Vietnamese football is losing its data at the character layer, long before anyone gets to ask a tactical question.

A LEAGUE THAT IS GOVERNED BUT NEVER MEASURED

The Vietnam Football Federation holds the sporting-governance mandate; the Vietnam Professional Football Joint Stock Company operates V.League 1 and V.League 2. That is a clear two-tier structure, not far removed from the model used by neighbouring countries in the region.

But a governance structure and a data structure are two different things. In European leagues, a second-tier match and a top-tier match can sit inside the same Opta or StatsBomb database, sharing one event format, one coordinate system, one labelling convention. In Vietnam, match data is produced in fragments: some from the broadcast production company, some from the organiser, some from domestic statistics sites that collect it themselves, and a very large share typed by hand by the reporters themselves.

The result is a league with scorelines but no event history. With a table but no time series. With players but no stable identity record.

When the stadiums fell silent in 2026, I understood something: football never died, it simply took off its clothing and revealed its skeleton. In Europe, that skeleton is event data — the thing that kept running through the months without crowds. In Vietnam, the clothing came off and what lay beneath was a far larger void than anyone had assumed.

FORTY REPORTS, AND THE DEATH OF A NAME

I built a small test set: 40 Vietnamese-language match reports drawn from three outlets, spread across one V.League 1 season. I ran the default extraction pipeline — UTF-8, no normalisation, no alias dictionary. The result was enough to make me stop and take careful notes.

The problem is Unicode. Vietnamese uses Latin script but with an unusually dense load of tone marks. A name can be stored in composed form, where a base letter carries combining marks, or in precomposed form, where it is a single code point. The two strings render identically on screen, but a computer treats them as entirely different strings. On that foundation, a system without normalisation will treat two spellings of the same name as two different people.

Then come the legacy encodings. Many publishing systems in Vietnam still carry the inheritance of code pages that predate Unicode. When text passes through several conversion layers, the tone marks get swallowed or turn into garbage characters. And so, inside a single database, a national-team player can exist as four or five different string variants. There is no way to join them without a hand-built alias dictionary.

Take three concrete names from the test set. Nguyen Quang Hai, Nguyen Tien Linh, Nguyen Hoang Duc. All three are names that have appeared densely in Vietnamese newspapers for years, tied to national-team windows and club fixtures. In my raw data, every one of them existed as at least three different strings, and none of those strings matched any other without an intermediate normalisation step.

I tried to patch it. I normalised all text to a single Unicode form, added a two-column alias table — accented and unaccented — and ran it again. The number of recognised entities rose sharply, enough for me to reconstruct the starting line-ups for most of those reports.

The conclusion was not tactical at all. The biggest loss in Vietnamese football data does not sit at the analytical layer. It sits at the writing layer. And it happens before any expected-goals model is ever trained.

In the summer of 2026, I saw the Opta ghost — and from that moment, my eyes stopped believing what they saw. But it took running this test set at 68 for me to see something more alarming than any ghost: a league erased from the database at the very first character.

THE FOREIGN-PLAYER QUOTA AND A DEBATE WITHOUT A RULER

Without stable identity data, every policy debate has to run on anecdote. The clearest example is the foreign-player quota.

V.League applies a limit on the number of foreign players who can be registered and fielded, a figure that has changed across seasons, plus additional slots for continental competition under Asian Football Confederation rules. The principle is clear. The measurement is not.

To know whether the quota helps or harms, three questions need answers: what share of available minutes foreign players occupy; whether young domestic players in the positions being displaced progress faster or slower than their peers; and whether clubs that use more foreign players generate more points per unit of wage than those that use fewer. All three require minute-level data by player, by season, under stable identity. In a European league, that can be exported in minutes. Here, it is a multi-month project.

The transfer market is a monastery where the numbers chant; I only write down what they pray. But when the monastery has no ledgers, people stop praying — they start quoting prices out loud.

CLUB LICENSING: THE LAW EXISTS, THE NUMBERS DO NOT

In Europe, the financial story of a club is told through financial fair play rules and, more recently, profit-and-sustainability rules. Their principles — loss limits, disclosure requirements, a binding relationship between spending and revenue — all have counterparts in the club-licensing system of Vietnamese football.

The difference is not in the wording of the law. It is in the availability of the data used to enforce it. A regulation only carries weight when breaking it leaves a trace someone can look up. If club financial statements are not published on a regular cadence, in a uniform template, and retained as a multi-year series, then whether enforcement is correct or incorrect cannot be verified from the outside. Supporters are left with belief or disbelief.

This turns injury, contract and transfer into black boxes. A player misses four matchweeks, and nobody knows whether it is injury, an internal sanction, or a contract clause under negotiation. Clubs say only what benefits them. That is the instinct of every club in the world, but where there is no underlying data, that instinct has no counterweight.

THE CALENDAR AND AN INVOICE NOBODY ISSUES

There is a cost Vietnamese football pays every year that has never been booked: the scheduling conflict between clubs and the national team.

FIFA international windows, national-team camps, and regional and continental tournaments all take players away from their clubs during the most important weeks of the season. In Europe, the phenomenon has its own name, and it is measured in points dropped and in injury days that appear after players return.

In V.League, that measurement barely exists. Nobody knows precisely how many points a club loses during a national-team window. Nobody knows how much higher the injury rate is on return. Nobody knows which club absorbs the heaviest cost, or whether compensation is warranted. FIFA's solidarity mechanism — the system that distributes a share of transfer fees to clubs that trained a player — is a real money flow, but it passes through a system in which the receiving side often lacks records clean enough to prove its own claim.

I am 68 years old, and yet data is younger than I have ever seen it — every season it grows another layer of teeth. The problem for Vietnamese football is not that its data is immature. The problem is that the data has never been born correctly in the first place.

THE VOID IS NOT A TECHNICAL FAULT. IT IS DATA.

This is where I want to argue against most of what is usually said.

The natural reflex when an extraction pipeline returns an empty result is to treat it as a fault and patch it. An engineer will check the logs, look for page-fetch failures, look for paywall blocks, look for JavaScript-rendered pages. All of that is reasonable.

But there is another possibility few are willing to consider: an empty result can be a finding rather than an incident. When a dense, lengthy body of text yields almost no entities at all, that says the infrastructure beneath it was not built to carry analytical load. That is information about the maturity of a market, not about the quality of a report.

The real danger sits in the next step. An empty pipeline will always tempt someone to fill it. If the process does not permit the conclusion insufficient information, the model will generate a conclusion that sounds entirely plausible and rests on nothing. In my line of work, a null result must be published as a null result, together with a specification of what would be needed to run the job again. Inventing a complete analysis out of an empty input is the most serious error a data writer can commit.

The same principle applies to Vietnamese football. When there are no numbers, the correct move is to say there are no numbers. A football nation that tolerates heroic prose in place of measurement will gradually lose the ability to tell progress from the feeling of progress.

There is a subtler confusion still: trusting the label instead of the content. A classification tag reading Vietnamese football is a routing label, not evidence. A league that is labelled but not measured is a league that is described but not known. And a football nation that is described but not known cannot price itself correctly.

Women's football is the clearest illustration. The results of the Vietnam women's national team are real and countable. But the data surrounding the domestic women's championship is far thinner than the men's game, while commercialisation announcements still appear on schedule as a corporate social responsibility line item. When the presentation outruns the data, what is being sold is not women's football. What is being sold is the image of the organisation selling it.

THE NEXT STEP IS NOT TO BUY A DATA PROVIDER

I do not think Vietnamese football needs a large contract with a global data provider. That is the most expensive way to solve a cheap problem.

What is needed first is character discipline. A single Unicode standard applied to every piece of published text. A player alias dictionary, accented and unaccented, published openly so every party uses the same one. A minimum event format, simple enough that a second-tier club can comply with it. A season-by-season archive that is never overwritten. None of this requires the budget of a major league. It requires one accountable person and a decision that does not change for ten years.

Once that foundation exists, everything else becomes feasible: valuation models for domestic players, impact assessment of the foreign-player quota, measurement of the cost of international windows, and transparency in club licensing.

A beautiful number is like a perfect pass: it does not need explaining, it only needs to be seen. But before there can be a perfect pass, someone has to hold the ball correctly. Vietnamese football has the ball at its feet. The question for the next round of fixtures is not who will score, but whether the game can keep hold of its own name inside a database — tone marks intact.