A Property Release Inside a Football Dataset: Why V.League Needs a Source-Validation Gate

On 9 October 2026, Faisal Town Group issued a release about the inauguration of Jamia Masjid Hazrat Ali (R.A.) inside the Faisal Hills housing scheme, on Main GT Road near Taxila, Pakistan. The release carried a date, an hour, the name of the group's chairman, the technical specifications of the building, and a reference to an Asia Pacific property award. It entered a football data pipeline labelled "football".

Twenty-three information points were extracted. Not one referred to a club, a player, a coach, a competition, a transfer window, a governing body, or any commercial football activity. The entities actually present were Faisal Town Group, Faisal Hills, Jamia Masjid Hazrat Ali, Chaudhry Abdul Majeed, Faisal Mosque in Islamabad, the Asia Pacific Property Awards 2026–2027, and six place names.

Nine football-specific analytical instruments — tactical systems, xG and PPDA, FFP and PSR, transfer structures, league positioning, dressing-room ecology, value-chain transmission — were applied to the text in turn. All nine returned null across every substantive field. Not because the instruments were broken. Because there was no football in the text.

The only thing that survived every layer of verification was a calendar check. 9 October 2026 falls on a Friday, matching the detail that the mosque "opened ahead of Friday prayers". A small signal, but enough to conclude that the factual skeleton of the release — a building exists, it opened on that date, at that site — is almost certainly sound. The evaluative layer is not. And the label is entirely wrong.

I read that report three times, not because it was difficult, but because it was familiar. Thirty years of tactical analysis in Japan taught me that the most dangerous error in this trade is not a fabricated number. It is a true number, placed in the wrong drawer, that nobody bothers to check because it sounds plausible.

When Vietnamese football data passes through the same door

V.League has no centralised data pipeline of the kind run by the Premier League or J.League. Information about the competition arrives from four directions: VPF as the organising body, the broadcaster holding the rights, the media departments of individual clubs, and an intermediary layer of aggregator sites, supporter groups, and accounts that specialise in transfer news.

Those four sources do not share a standard. But once they flow into the same spreadsheet, they become the same kind of data. A club publishes a statement about an injured key player, an aggregator copies it, an account translates it into English, a model reads it and tags it "injury". Nobody in that chain asks who the original source was, or what interest that source has in telling the story that way.

I saw exactly this mechanism in the Japanese championship in the 1980s, when I stayed two hours after a 1-1 draw at Mitsuzawa purely to redraw Furukawa Electric's pressing scheme by hand. There was no pipeline then. I was the pipeline. And because I was the pipeline, I had to verify every source manually.

In 2026, when editors born in 2026 at a Japanese sports site kept citing Expected Goals, I objected. I said paper data could not capture real space. Then Kawasaki Frontale beat Urawa Reds 4-3 with an xG of just 2.8, thanks to three shots from outside the box. My hypothesis collapsed. I quietly learned Python and modelled 1,200 matches from 2026 to 2026, and found that xG is only accurate when paired with the position where the attack begins. At fifty-eight, I typed every line of Python to prove the young ones wrong — and ended up proving myself wrong first.

The lesson that year was not "trust the data". It was: data is only trustworthy once you know which door it came through.

The mechanics of a classification error

The incident of 9 October is not an analytical failure. It is a taxonomy failure. The two differ in nature and in how they are fixed.

An analytical failure happens when the tool is right but the input is wrong, or the input is right but the reasoning is wrong. It is loud and easy to detect, because the output usually contradicts observable reality. A classification failure is silent. The document enters the correct pipeline, is processed by the correct procedure, produces a null result, and that null is read by the system as "no signal" rather than "wrong category".

The mechanism here was almost certainly keyword matching. The release contains the words "venue", "capacity", "inauguration", "award", "infrastructure". An automated filter saw that string and filed it under sport. No gate checked whether the text contained a football entity.

Such a gate is cheap. It needs one condition: the text must contain at least one entity from the set {club, player, coach, competition, football governing body, football stadium}. The release of 9 October fails that condition at the first step. It gets blocked, reclassified as Real Estate, and never touches the model.

V.League needs that gate, but in a manual and stricter form. When a club announces an injury, the gate must ask: is there a player name, a position, recent minutes played, a count of matches in the last 21 days. If three of those four are missing, the item does not enter the injury table. It goes into a separate drawer called "club statement" and is never used for calculation.

Source tiers: the layer Vietnamese football has not built

The release about Jamia Masjid offers five figures: capacity for more than 22,000 worshippers, a claim to be "second-largest after Faisal Mosque" in the Islamabad–Rawalpindi region, 300 parking spaces, 200 ablution places, and an Asia Pacific Property Awards citation for 2026–2027. All five trace to a single origin: Faisal Town Group, or the article's own framing. No second source is cited.

Football has the same source tiers and routinely violates them. A transfer story published by the club itself carries the highest credibility. A story leaked by an agent to a newspaper carries medium credibility, because the agent has a direct interest in creating negotiating pressure. A story copied by an aggregator from a social account carries the lowest credibility, which in practice is zero.

But when all three sit in the same table, they look alike. And when a squad-value model reads all three, it aggregates numbers of identical shape and entirely different nature.

This is why I keep one hard rule in every piece of analysis I write: every figure carries a source label, and the label sits next to the number, not in a footnote at the end. Readers have the right to see who is speaking before they decide whether to believe it.

The problem of the marketing comparative

The claim "second-largest in the region after Faisal Mosque" is the most dangerous kind of statement in any dataset: a ranking assertion that never discloses its metric.

Ranking depends entirely on which quantity you pick. Covered prayer-hall area is one quantity. Total site capacity is another. Actual congregation served on a peak day is a third. Three measures produce three different rankings, and the release does not say which one it used.

Vietnamese football has an entire collection of the same statement type. "Second-largest stadium in Southeast Asia" — measured by capacity or by pitch area? "The biggest-spending club in V.League" — transfer fees, total wage bill, or total outlay including signing-on fees? Three calculations, three answers, and the club always picks the one that flatters it.

The problem is not that clubs lie. The problem is that such a statement, a few months later, gets copied into reference material and becomes a default truth. Once it is default, nobody re-checks it, because re-checking is the work of the sceptic, and scepticism takes time.

Two gaps Vietnamese football is leaving open

I hold two professional convictions, and both are cases of missing data rather than wrong data.

First: signing-on fees for free agents. When a player reaches the end of his contract and joins a new club, the money paid to him and his agent appears in no transfer database, because technically no transfer fee was paid. Yet the real cost of that deal still exists, and it is frequently larger than the transfer fee for an equivalent player still under contract. The result is that every spending table understates clubs that build squads this way. A supervision gap, not an accounting error.

Second: fixture density. When a player suffers a muscle injury, the published information usually covers the injury type and the expected absence. Nobody records that he played seven matches in 23 days, three of them involving long-distance travel. No medical department survives a squad playing two games a week for six consecutive weeks. But that density never appears in the injury table, so it never appears in the conclusion.

Both cases resemble the release of 9 October in one respect: the information published is not false. It simply omits the part that matters most.

The contrarian view: the real risk is true but useless content

When most people discuss data contamination, they think of fake news. Fake news is loud. It provokes argument, it gets fact-checked, it gets refuted. A system develops an immune response to it, however slowly.

But the release about Jamia Masjid contains not one false word. The building exists. The inauguration date exists. The chairman exists. The award exists. The entire text is accurate, and the entire text is useless to a football dataset. Precisely because it is accurate, nobody checks it. Precisely because it is useless, nobody notices.

That is the paradox of every data pipeline: false content is stopped at the door, while accurate but miscategorised content walks straight into the engine room.

There is a second paradox, and I think it is more serious in Vietnamese football. We are sceptical of institutions and generous with non-institutional data. When VPF publishes a figure, supporters immediately question it. When an aggregator account posts a figure, supporters copy it and share it. It should be the other way round, or at minimum the same protocol should apply to both. I learned this in 2026, when the stewards at Mitsuzawa asked me three times for my credentials and phoned the organisers to verify. They suspected me because I was a stranger in the press area. And because I was suspected, I had to redraw Furukawa Electric's pressing scheme by hand — and that analysis earned a phone call of praise from head coach Saburō Kawabuchi himself.

The person blocked at the J.League gate in 2026 now writes about how data changes tactics. That gate never disappeared. It simply moved from outside the stadium into the machine.

The gate must be set by people, not algorithms

One detail in the report is worth keeping: this was a taxonomy error, not an algorithm error. No algorithm invented the label "football". Someone configured that category. Someone approved that pipeline. Someone decided keyword matching was sufficient qualification.

That matters because it changes the fix. If the fault belongs to the machine, you improve the model. If it belongs to people, you change the approval procedure — and changing procedure is always harder.

In Vietnamese football, an approval procedure for information barely exists at club level. A post on an official club channel is treated as a primary source, yet it passes through no internal verification step. The person writing it may be a communications officer or a freelancer, and their motive is to please the board, not to supply accurate data.

I do not blame them. Club media is not a statistical agency. But if we want to build a data platform for the league — and V.League needs one if it is ever to price its rights correctly — there must be a layer between the club and the table, doing exactly one job: labelling the source.

Three questions before believing a figure

From this incident I draw three checks, and I will apply them to every piece of analysis I write next season.

Step one: does the text contain entities. If an article about football names no club, no player, no coach and no competition, it is not an article about football, whatever vocabulary it uses.

Step two: who does the figure come from. If it comes from a party with a direct interest and there is no second source, it is a marketing claim, not data. The correct handling is not deletion but labelling — and never using it for ranking.

Step three: is the metric disclosed. If a ranking claim does not state what it measures, it cannot be verified, and the unverifiable does not enter the aggregate table.

These three steps are cheap, manual, and require no algorithm. They require one person willing to spend two hours after a match sitting down, as I did at Mitsuzawa in 2026.

What to test at the next round of fixtures

I propose a small experiment, doable at the next V.League matchday. Pick any figure published by a club — attendance, injury count, a player's minutes, the value of a contract. Then find a second source outside that club. If you find one, the figure deserves a place in your dataset. If you do not, leave it in the "club statement" drawer and never use it to compare against anyone.

Do that with one figure and you will understand how a release about a mosque in Taxila could pass through nine layers of tactical analysis without being stopped.

I have watched the ball roll my whole life, but only after stepping away from it did I truly understand: what determines the quality of an analysis is not the model, but the list of things the writer refused to let in. A healthy football dataset is measured not by the records it holds, but by the records it threw out.