The Beautifully Formatted Blank Page: The Silent Failure Inside Football Analytics

Opening

At 2:14 a.m. on 9 March 2026, in a rented office on Rue de Bonnel in Lyon's 7th arrondissement, I opened a file a scouting company had sent with a trial contract: fourteen PDF pages on a twenty-one-year-old defensive midfielder in the Belgian second division. Page three carried a green status bar reading "Form stability: 92%." Page seven read "Injury risk: low." Page eleven read "No red flags on attitude."

Three weeks later I obtained the raw JSON export of that same report. The field stability_index returned null. The field injury_risk returned null. The field attitude_flags returned an empty array. The platform's rendering engine interpreted null as "no negative signal," and "no negative signal" was painted green. Not a single data row about that player had ever passed through the system. The upstream provider for the Belgian second division had stopped pushing data eleven days earlier over a licensing dispute.

I was holding a flawless report about a human being the system had never seen.

That was the moment I understood that the biggest risk in modern football analytics is not a model computing the wrong answer. It is a model staying silent, and that silence being printed in green ink.

Context: an industry built on empty cells

Over fifteen years in this trade I have watched a top-flight French club's data department grow from two people to nineteen. I have watched a sporting director open a tablet in a press room to prove that an incoming striker had better burst metrics than the one just sold. I have watched a lawyer use a PPDA chart to argue that a coach should not be sacked. And I have watched an investor use a projected wage bill to persuade a bank to extend a loan.

Football analytics is no longer a late-night blogging hobby. It is a four-tier supply chain: raw data collectors in the stands, event-data processors, visualisation platforms, and end consumers — clubs, agents, bookmakers, newsrooms, investment funds.

Every tier has its own reason not to say one thing: the underlying data may be missing.

Collectors are paid by matches covered. Processors are paid by metrics emitted. Platforms are paid by seats. Newsrooms are paid by pageviews. In that chain, an empty cell generates revenue for nobody. An empty cell painted green does.

I have read hundreds of pages of scouting-platform marketing material over three years. Not one of them devoted a section to the question: "When we do not have data, how will we show it to you?" Coverage clauses sit in appendices, in eight-point type, and usually describe how many leagues are supported rather than what share of matches in each league actually has complete data.

Process metrics and the trap of consensus

Most public argument about football data orbits xG — a model estimating the probability that a given shot becomes a goal. It is a good tool. It measures chance quality rather than conversion luck. xGA measures the same thing at the other end. PPDA measures pressing intensity as passes allowed per defensive action.

These three have become a common language, and precisely because of that they have become a shelter for lazy conclusions.

The test I consider most valuable in the entire analytical toolkit — the divergence test between process data and results — is the least used in journalism. It requires a long enough series, a large enough sample, and an acceptance that the two can tell different stories. It also requires the writer to tolerate having no conclusion for several weeks.

The lesson I took from the 2026 World Cup: referees read spreadsheets too. When I sat through forty-eight group-stage matches and hand-logged 1,247 decisions, what I learned was not that referees favour anyone. What I learned was that a decision can be correct under the laws, plausible to the naked eye, and still sit two standard deviations off the general distribution. Watch one match and you see a referee. Watch forty-eight and you see a pattern. And if you watch only the metrics published by a single vendor, you see exactly what that vendor wants you to see.

Consensus around metrics manufactures false safety. When three newspapers cite the same xG figure, the reader feels that three independent sources have confirmed each other. In reality all three are citing one vendor, one model, one run of the numbers. Three harmless data points stitched together into a money map that leads to a village with no football pitch — the same mechanism I have seen repeat across club financial filings.

Finance: where empty cells cost the most

If there is one place where missing data causes damage in real money, it is a club's accounting office.

Start with transfer amortisation. A fee is not booked once. It is spread across the contract's length. A player signed for four years at €100m carries a €25m annual book cost. Sell him after two years for €40m and the club books a €10m accounting loss, even though cash-flow-wise it received €40m. This is why some deals that look catastrophic in the press are rational arithmetic on the ledger, and vice versa.

The sell-on clause is the second layer of the same mechanism. Manchester City inserted one when they let Jadon Sancho join Borussia Dortmund in 2026; when Sancho moved to Manchester United in July 2026 for a reported fee around €85m, a share of the value flowed back to Manchester. Paths like that never appear in transfer bulletins. They appear in the notes to the accounts, usually on the thirtieth line of a four-page note.

COVID-19 closed every stadium on earth, but the holes in financial reporting never socially distanced. In April 2026, with Ligue 1 cancelled mid-season, Olympique Lyonnais published a 112-page emergency report on Euronext. I spent three weeks reconciling line items against DNCG filings and found a €7.8m "intermediary fee" paid to a Luxembourg company incorporated exactly two months before signature, whose legal representative shared a name with the agent of a substitute player. The structure had every marker of a shell company. Whether it broke a law was a question for investigators, and I did not answer it myself.

Five years later that same club fought a legal battle with the DNCG over financial self-sufficiency, was administratively removed from Ligue 1 in June 2026 and restored on appeal in July. The story was told in the press as a drama about a president. Most of the pieces I read never opened the accounts to see where the revenue structure and the debt structure actually sat.

In England, the Profit and Sustainability Rules produced precedents nobody can ignore. Everton were docked ten points in November 2026, reduced to six on appeal in February 2026. Nottingham Forest were docked four points in March 2026. Manchester City faced 115 charges from February 2026. Juventus were docked fifteen points in January 2026, adjusted to ten in May.

Each of those sanctions is a lesson in methodology. To judge a club's exposure you need the wages-to-revenue ratio, the operating margin of football activities, the squad's amortisation structure, and the permitted loss for each assessment window. Those four variables rarely appear together in one article. When they are absent, the writer has two options: say the evidence is insufficient, or build a story on a single number. The second option always gets more readers.

Sporting culture is at its most beautiful viewed from the stands and at its most nauseating viewed from the accounting office. Yet it is from that same accounting office that people learn whether the club they love will still exist after June.

Transfer money flow: read the agent-fee column

The transfer market never lies if you read the agent-fee column instead of the player-price column.

The player-price column is written for the audience. It is a number agreed between two clubs and published in the way most flattering to both. The agent-fee column is written for regulators. It sits in contract registration files, in federation annual reports, and in documents very few journalists bother to read.

Moisés Caicedo joined Chelsea in August 2026 for a reported £115m, the highest fee ever between two English clubs. Enzo Fernández arrived at the same club in January 2026 for £106.8m after winning the young player award at the 2026 World Cup. Those figures have been repeated thousands of times. What is repeated far less is the instalment structure, the performance add-ons, and the percentage flowing to intermediaries.

The technique I have used most over six years is three-tier tracing: source of funds, intermediate stopping point, ultimate beneficiary. In September 2026, still a master's student, I received a scanned copy of a €45m contract between a FIFA subsidiary and an energy group under the Qatar World Cup framework. A €3.2m "access fee" clause routed money to a Bahamas account where the recipient company held a registered address but no physical office. I confirmed the company was incorporated two months before signature. The 4,800-word investigation ran on 15 November 2026. FIFA later opened an internal audit and acknowledged that a large share of the costs lacked full supporting documentation.

Qatar built stadiums on hot sand; I exposed sponsorship contracts signed on quicksand.

But I have to state my limits clearly: I proved a structure, not a motive. There is a gap between "this structure cannot be explained by ordinary business logic" and "this structure is bribery." Plenty of investigations collapse because the writer closes that gap with an adjective.

Governance: empty cells that are not allowed to exist

At the governance layer, missing data is not merely damaging. It is an act.

Third-party ownership was banned by FIFA because it lets an investor hold part of a player's economic rights and therefore an incentive to interfere in transfer decisions. Tapping-up is prohibited yet extremely hard to prove, because the evidence lives in meetings with no minutes. Article 19 tightly regulates cross-border moves of players under eighteen, and it is the area where academies in Africa and South America face the most pressure.

Then there is multi-club ownership. City Football Group, the Red Bull network, Eagle Football — each creates a validity question when two affiliated clubs qualify for the same European competition. That question has a clear answer in the text. The question without a clear answer is: when one club in a network sells a player to another club in the same network, who determines the market price?

I tried to trace one such internal transfer chain for two months. I stopped because I could not get three independent sources. I wrote a four-page note on why I stopped, and I did not publish it. Professionally correct, commercially wrong.

Dressing rooms, cycles and what models leave out

There is a paradox in football analytics: the more data there is, the easier it becomes to ignore what cannot be measured.

Fixture density is the clearest example. No model predicts injury better than counting rest days between matches. When a team plays twice a week for six straight weeks, no medical department saves them. I tracked one Ligue 1 club through 2026-24 and logged every muscle injury. The pattern was not in pitch type or age. It was in the gaps between matches. The dataset that club bought had a "match load" column but no "cumulative monthly match load" column. The missing column was the most important one.

At the tactical layer, homogenisation is hard to see in metrics. The inverted winger has become the default at most big clubs because it creates central overloads and frees the full-back. But when everyone does it, the touchline empties, and pure wingers — players who can hold width, stretch a back line and create horizontal space — are undervalued by models that measure effectiveness inside the shape the system prefers. This is a subtler bias than a technical bug: the system measures what it was designed to measure and ignores what it was not.

At the management layer, two effects are routinely misread. The "new manager bounce" is usually attributed to the newcomer's talent; in most cases I have checked it is regression to the mean plus an easier fixture list. The contract year is usually read as a surge in motivation; in reality it is three incentives running at once — the player proving himself, the club wanting to sell before losing him for nothing, and the agent building negotiating leverage.

Media, rumours and the expectation cycle

Of all football information, transfer news has the highest noise ratio and the lowest verification ratio.

The only way to grade a transfer rumour is to tier the source. Tier one: a journalist with a verification record and direct relationships at the club or agency. Tier two: a major outlet citing tier one. Tier three: a tabloid citing tier two and adding an emotional detail. The problem is that on social media all three appear in identical format, identical type size, identical emoji.

Agent motive is the variable most articles skip. A rumour can be floated to pressure a parent club, to raise the price of another client on the same roster, or simply to test fan reaction before a real deal. I counted seven cases across the last two windows where a rumour surfaced exactly one week before another player represented by the same agent signed a new contract.

In esports, players' win rates are public but investors' win rates are not. Football is walking the same road more slowly: player performance data floods the market while capital performance data stays near zero.

Industry transmission: from academies to broadcast contracts

Every football story has three transmission layers, and the third is usually skipped.

Upstream is the talent supply chain: academies, football schools, local scouting networks. Midstream is clubs and competitions. Downstream is broadcast rights, commercial revenue and derivative markets.

A change upstream takes five to eight years to surface downstream. When a country tightens rules on minor transfers, the first consequence is not a big contract. It is academies relocating to neighbouring countries and youth tournaments appearing with opaque ownership structures. Nearly a decade passes before that consequence reaches a European wage bill.

Downstream, the broadcast contract is where the largest money flows and where the fewest people read. A three-year rights deal routinely carries dozens of pages of annexes on merit-based distribution, rebate clauses if the season is cancelled, and resale rights for third markets. Those three clauses determine a mid-table club's finances more than any transfer.

The reasonable case on the other side

I have to say something my peers often avoid: the analytics industry has genuinely made football more transparent, and I am a beneficiary.

Before xG became common language, judging a team rested almost entirely on the authority of the speaker. A famous pundit said team A defended well, and it became true because nobody had a tool to argue. Today a first-year student can check it against open data. That is an advance in accountability I would not trade away.

It is also true that refusing to conclude when evidence is thin is a professional virtue rather than a weakness. In many cases "insufficient information to assess" is the most accurate and most useful answer an analyst can give. The problem is that this answer does not sell advertising, so it rarely gets printed.

And here is what I remind myself every week: scepticism is a method, not a personality. People call me a cynic; I call myself someone who reads the books behind the pitch. The difference is that someone who reads the books accepts that most pages are ordinary and only a few lines are anomalous. An investigative reporter who finds anomalies on every page has stopped investigating and started campaigning.

There is one technical distinction I hold tightly: risk of wrongdoing versus evidence of wrongdoing. A company incorporated two months before a contract is a high risk indicator. It becomes evidence when documents about the ultimate beneficiary and the actual services supplied are added. The distance between those two levels is the whole border between journalism and indictment.

A doctrine for empty cells

Back to the fourteen-page report in Lyon.

What bothered me most was not that the company was wrong. It was that their system behaved exactly as designed. null is not a bug. It is a valid value. The failure sits at the presentation layer, where a valid value is translated into a positive message that nobody re-checks.

In medicine, when a test cannot run, the result is recorded as "indeterminate" and the doctor knows information is missing. In aviation, when a sensor fails, a warning light illuminates rather than going dark. Football analytics has no equivalent convention. We built a system in which losing a signal looks identical to having no problem.

That is why I propose a minimum convention for any published analytical report, inside a club or inside a newsroom: every table carries a coverage line stating what percentage of fields actually hold values; every conclusion carries a limits line stating what would make it wrong; and every empty cell is displayed as an empty cell, not as a green tick.

Those three lines cost far less than a €15m transfer executed on the strength of a blank report.

Closing

That fourteen-page report still sits in the second drawer of my desk, printed, stapled, with three handwritten notes in the margin of page three.

I keep it because it reminds me that the best tool in this trade is not a forecasting model but the habit of asking one simple question before any table of numbers: where did this data come from, who pays for it, and what is missing here.

The football analytics industry will keep growing, and most of what it produces is useful. But an industry only matures when it learns to publish its blank pages too — and when readers learn to recognise that a beautifully formatted blank page is still a blank page."