Football Does Not Lack Data, Football Lacks Auditors

On the evening of 16 June 2026 I was sitting in a rented flat in Lyon's 7th arrondissement, in front of two screens and a ruled notebook. France against Australia in Kazan had just finished. The left screen replayed the footage; the right screen held an Asian handicap odds sheet I had downloaded from five different bookmakers. By the third penalty of the second half I stopped and started counting.

That day, twelve penalties were awarded across stadiums in Russia. Nine of them went to the side rated weaker. I did not trust my instinct, so I did what a first-year journalism student can do: I wrote it down by hand. A month later my notebook held 1,247 refereeing decisions, cross-referenced against open Opta data and the odds of five Asian bookmakers. One referee emerged with 78 per cent of his flagged fouls going in favour of the weaker team, 2.5 standard deviations off the tournament mean.

The piece ran on a faculty blog and was taken down within 48 hours. The spreadsheet is still on my hard drive.

That was the first time I understood something that remains true a decade later: in football, bad data is almost never punished. Good data can be taken down.

Context: an industry running on information nobody cross-checks

Professional football worldwide is a vast revenue ecosystem built on broadcasting rights, shirt sponsorship, ticketing, merchandising and player transfers. Every summer, enormous sums cross borders as transfer fees, agent commissions, signing bonuses and multi-year instalments. Every one of those numbers is born from a signed document with a date and clauses.

The paradox sits elsewhere. While transactions are recorded in extraordinary detail at club level, the information the public receives travels through a chain of intermediaries with no auditing mechanism at all. An anonymous source tells a journalist. The journalist publishes. Four other outlets cite that outlet. By the fiftieth report the story has a specific fee, a specific wage, a specific release clause, and nobody in the chain remembers where the original figure came from.

I call this the amplification effect with no floor: an unverified claim can be duplicated infinitely without gaining a single unit of evidence. That is why every transfer window Vietnamese fans read hundreds of articles about the same deal with dozens of different fees, and none of them is wrong enough to be reprimanded.

The core: dismantling the toolkit the industry actually uses

To discuss football data seriously you have to separate three layers. The first is match data: passes, shots, duels. The second is model data: xG, xGA, PPDA, expectation metrics. The third is financial data: transfer fees, amortisation, wage bills, revenue. All three layers have problems, and their problems are not the same.

Start with xG, Expected Goals. It estimates the probability that a given shot becomes a goal, based on location, angle, shot type, bodies in front of the ball and other variables. Its value is that it separates chance quality from finishing ability. A team with 20 shots worth 0.9 xG played a far worse match than a team with six shots worth 1.8 xG, even if both finished 1-0.

The problem appears when xG shifts from measuring tool to ideology. In Vietnam, post-match reports often cite xG as a final verdict: a winning team that lost the xG battle is called lucky, a losing team that won it is called deserving. That reading ignores two things. First, every provider defines xG differently, and the same match can differ by up to 30 per cent between two sources. Second, xG cannot measure the most important thing in football: decision-making. A shot from an xG-0.04 position can be the right choice if the striker is surrounded by three defenders while a teammate is free on the flank.

The second metric worth naming is xGA, Expected Goals Against. It is the xG a team concedes and is a process measure at defensive level. A defence with low xGA across ten consecutive rounds is usually a well-organised defence. But xGA has a large blind spot: it cannot distinguish between a defence that actively smothers opponents and a goalkeeper repeatedly saving after the defence has been torn open. Both produce low xGA, but one is sustainable and the other collapses the moment the keeper dips.

Then PPDA, Passes allowed Per Defensive Action. Put simply, it is the number of passes an opponent completes before your side makes a defensive action. Lower PPDA means more aggressive pressing. This is the metric I use most when analysing Southeast Asian teams, because it reflects tactical intent far faster than possession share.

PPDA also carries traps. A side that deliberately cedes territory and sits deep will record very high PPDA, which does not mean it is playing passively or badly. Conversely, a team that presses furiously for 20 minutes and then runs out of legs will show a flattering full-match PPDA average while actually losing control after the break. Averages always hide segment values, and short-form reports almost never publish split figures.

Now the third layer, the one Vietnamese football journalists touch least: transfer amortisation.

Transfer amortisation: the machine that beautifies the books

When a club paid £106.8m for Enzo Fernández in January 2026, that sum did not appear on the accounts as a single cost. It was booked as an asset, spread evenly across the contract's years. Enzo Fernández signed to 2032, roughly 8.5 years. Spread evenly, the club books around £12.5m of amortisation a year.

The same happened with Moisés Caicedo and Mykhailo Mudryk, on eight-year and 8.5-year deals. Together those three transfers pushed the club's annual amortisation charge to a level its revenue could not easily match in the short term.

The mechanism is entirely lawful. It is an accounting standard, not a trick. But its effect on the transfer market is troubling. The longer the contract, the thinner the fee is stretched, and the more expensive players a club can buy in a single window while keeping its financial ratios inside the permitted threshold. That structural incentive is what made seven-, eight- and nine-year contracts common in England between 2026 and 2026.

Alongside amortisation sits the sell-on clause. This is a former club's right to a percentage of a player's next transfer. A club that sells a young player cheaply while retaining 20 per cent of a future sale can earn more than the original fee if the player makes good. In accounts, this contingent value is sometimes recognised in different ways, and it is an area readers cannot verify from transfer reports alone.

In the same family sit buy-back clauses, release clauses and performance-related add-ons. A deal announced at 40 million euros may in reality be 25 million upfront, 10 million over three years, five million in add-ons, plus a 10 per cent sell-on. The report prints the first figure. The accounts print all of them.

That is why I tell my journalism students to read the balance sheet before the headline. The transfer market never lies if you bother to read the agent-fee column instead of the player-price column.

FFP, PSR and verdicts written on paper

At European level, UEFA's Financial Fair Play has been replaced by the Financial Sustainability Regulations with a squad cost rule. It caps combined wages, transfer amortisation and agent fees at 70 per cent of revenue, phased over several years: 90 per cent at first, then 80 per cent, then 70 per cent.

Domestically, the Premier League operates the Profit and Sustainability Rules, capping a club's losses at £105m over three years.

What matters is that these rules are not merely on paper. In February 2026 Manchester City was charged with 115 alleged breaches of financial rules. In November 2026 Everton were deducted 10 points, reduced to six on appeal in February 2026, before a further two-point deduction in a separate case. In March 2026 Nottingham Forest were deducted four points.

Looking at that sequence, a pattern emerges that coverage rarely places side by side. All three cases turned on the same technical question: which costs count, which do not, and in which year. The answer depends on classification inside the accounts, not on how well or badly the club plays.

In other words, a club's fate in the table can be decided by an accounting decision made three years earlier, by someone the crowd has never seen.

Three harmless data points combine into a money map leading to a village with no football pitch.

I first wrote that line in an investigation into a sponsorship contract, and it remains the most accurate description of my method. A single number means nothing. A 7.8 million euro agent fee routed to a Luxembourg company is ordinary. A director whose name matches the agent of a reserve player is a coincidence. A company incorporated two months before the contract was signed is common. Put those three facts side by side and it is no longer coincidence.

That is the technique I learned reading French clubs' accounts on Euronext and cross-checking them against the files of French football's financial regulator, the DNCG. That body can demote a club or expel it from the league if the club cannot prove it can balance its books. That power makes DNCG files one of the most reliable data sources in the industry, and one of the least exploited.

Third-party ownership, minors and the grey zone of law

FIFA banned third-party ownership, TPO, in 2026. Before that, investment funds could hold a share of a player's economic rights and profit from each transfer. The ban came because the model created bad incentives: a fund could pressure a player to move early regardless of where his career was heading.

Parallel to it is Article 19 of FIFA's transfer regulations, restricting international transfers of players under 18 to a narrow set of exceptions. The rule exists to protect children from being traded as goods.

One observation is worth making. Both provisions carry clear legal force, and both are hard to enforce because they run against a very strong current of interest. An academy in Africa that develops a 15-year-old may earn more from one move than its entire multi-year operating budget. A European club needs young players to meet quotas and to resell. An agent needs a transfer to live.

When every party benefits from circumventing a rule, the rule is observed only at surface level.

Small-sample illusion: when one match becomes a doctrine

There is a type of data more dangerous than wrong data: correct data drawn from too small a sample.

A club appoints a new coach, wins three in a row, and is instantly written up as a tactical revolution. The industry has a name for it: the new-manager bounce. But three matches is far too few to separate the human effect from the fixture effect. Three weak opponents, two home games, one against a side in an injury crisis: together, those three wins may say nothing about the new coach's quality.

Conversely, a mid-tier side reaching a major semi-final is often analysed as a successful model. But in knockout football a tie can be settled by a penalty in the 89th minute or a red card in the 12th. An amateur side reaching a final usually benefits from a kind draw and one explosive match; it does not prove a system works.

This matters for Vietnamese readers, because regional tournaments and World Cups constantly generate such stories, and the data to test them usually arrives after the story has already been told.

Four years after the 2026 World Cup I drew another lesson from the same dataset: referees read the numbers too. A referee working under crowd pressure, media pressure and scoreline pressure makes different decisions than one working a match whose result cannot be reversed. Recognising that is not the same as concluding misconduct. It means asking questions about how referees are assigned, not just about individuals.

When the data pipeline collapses and nobody notices

This is the part I want most time for, because it is least discussed.

In investigative work I once met a very instructive technical situation. An automated extraction process ran over an article, but a failure at the extraction layer meant it read nothing at all. The output was a full nine-section template with exactly one populated cell: the domain label, containing a single word, football.

If that table had flowed straight into a reporting system unchecked, it would have caused two kinds of damage. The first is false reporting: a document with full headings, full sections and full conclusions, all resting on zero data. The second, more dangerous, is misreading. An empty record can be read as a finding of no issues, and a technical failure becomes a professional conclusion.

In football this happens daily in softer form. One outlet publishes a wrong transfer figure. An aggregator takes it and attaches a source label. A betting app pulls it from the aggregator and feeds it into a pricing model. A fan reads a number that has passed through three layers and believes it is verified fact.

No step in that chain is deliberate deception. Each link does the most reasonable thing with the information it has. Error multiplies because nobody is responsible for tracing back to the source.

Contrarian: why systematic scepticism can also be wrong

Here I have to argue against myself, because this is the part people in my trade skip.

People call me a sceptic. I call myself someone who reads the books behind the pitch. But there is a line I have nearly crossed many times: between the risk of wrongdoing and the evidence of wrongdoing.

An unusual agent fee is a risk of wrongdoing. It is not evidence. A shell company with no physical office is a risk of wrongdoing. It is not evidence. Converting risk into evidence requires at least three independent sources, a money trail reconciled end to end, and a plausible alternative explanation that has been eliminated.

Equally, scepticism about model data easily slides into an opposite but equally wrong position: rejecting the value of data altogether. xG is not truth, but it is not a con either. A team that keeps winning with low xG will usually regress, and that is a statistically grounded forecast, not superstition.

What I refuse is using data to close a debate rather than open it. A good metric should make you ask more questions, not stop asking them.

I should also be blunt about limits. This article has no access to any club's internal data. What I present comes from public sources: accounts filed with exchanges, regulator files, federation regulations, official statements and open data from statistics providers. There are areas where I am entirely blind, for example how a club allocates costs between parent and subsidiary inside the same group. I name those blind spots rather than fill them with plausible-sounding speculation.

What actually needs to change

Back to that empty table. It taught me three things.

First, a mandatory validation gate is needed in every information-production process, whether in a newsroom, a regulator or a club. Any record that still contains template instruction text instead of real data should be blocked before it reaches the public. In journalism that gate has a simpler name: an editor. But an editor can only work if the writer has logged the provenance of every figure.

Second, there must be a clear distinction between three kinds of sentence: fact, inference and conjecture. A good report must state which one it is using. When a headline carries a transfer figure, the reader needs to know whether it came from a contract, from an agent's account, or from a reporter's inference based on an alleged wage.

Third, methods must be public. When I publish an investigation I attach a methodology appendix describing how I collected and cross-checked the data. That makes the piece duller for the skimmer, but it makes it refutable. A refutable article is a healthy article. An irrefutable article is usually one that says nothing.

The path of money and the reader's responsibility

In ten years watching this industry I have moved from hand-recording 1,247 refereeing decisions in a ruled notebook to cross-checking a 112-page financial report against regulator files, and then to sitting with an empty data table trying to understand why it was empty.

Those three stages share one logic. The path of money in football always leaves a trace. The trace can be buried under layers of companies, layers of contracts, layers of media intermediaries. It does not disappear. People put money into a small club in a provincial town because something at the other end is worth more than the money spent.

Three harmless data points combine into a money map leading to a village with no football pitch. A village with no football pitch means football is not the reason. Something else is.

For Vietnamese fans, I do not suggest reading every financial report. I suggest three smaller things, each doable in a minute.

When you read a transfer story, find where the original figure came from. If the article only names another outlet, you are reading the second or third layer of the amplification chain.

When you read a report citing xG or PPDA, check whether the author names the data provider and the time range. A full-match average with no half-by-half split usually conceals more than it reveals.

And when a club is deducted points for breaching financial rules, remember that the real story lies in accounting decisions made years earlier, by people who never appear in the report.

What I do not know

I do not know whether European football's financial regulations genuinely narrow the gap between clubs or merely push money into more complex forms. The available evidence leans towards the second, but it is not strong enough for a conclusion.

I do not know whether expectation metrics will keep improving to the point of approximating tactical intent, or will hit a ceiling and become a new layer of decoration. The last twenty years lean towards the first, which is why I still use them, always with caveats.

And I do not know whether the growth of language models will leave sports information cleaner or dirtier. It could cross-check thousands of records in seconds. It could also generate thousands of empty records that look as polished as the table I described.

The only thing I know for certain is this. An information system without a validation gate does not collapse loudly. It collapses quietly, by producing complete, correctly formatted documents with emptiness in every cell.

In football people like to say the result does not lie. The result lies in its own way, and its way is to let the reader feel confident that he has understood.

Sports culture is at its most beautiful seen from the stands; at its most repulsive seen from the accounts department.

Both are true of the same match. The writer's job is not to pick a side but to stand in the middle and show the reader both at once.

Three harmless data points combine into a money map. An empty data table combines into a reminder that the most dangerous thing is sometimes not false information, but information containing nothing at all, presented as though it contains something.

And the last question belongs to anyone in this trade, in Vietnam or in Europe, at a major outlet or on a personal page: if your data table were empty tomorrow, would you know?