When the Data File Is Empty: Verification Discipline and the Cost of Sourceless Conclusions

Opening: 2.1 xG and a squared notebook

At dawn on 6 July 2026, in a small flat in Tianhe District, Guangzhou, I sat in front of a screen with a squared notebook and a blue ballpoint pen. The quarter-final between France and Uruguay in Nizhny Novgorod ended 2-0 to France. The post-match graphics displayed a line I still remember character by character: France, 39 per cent possession. In a group chat with fellow sociology students, a message appeared almost instantly: "France won by luck."

I wrote four numbers in the notebook. France: 39 per cent possession, 2.1 xG, 10 shots, 5 on target. Uruguay: 61 per cent possession, 0.4 xG, 6 shots, 1 on target. The possession column and the expected-goals column pointed in opposite directions. That was the first time I understood that what television calls "control of the game" and what a model calls "quality of chances" can be two different stories, even contradictory ones. Over the following three weeks I rewatched every match of the tournament and hand-built an xG table for each team by logging shot locations and situation types.

Before 2026, I watched football. After 2026, I read it.

Six years later, on a Tuesday afternoon in Guangzhou, a data file arrived. It was empty. No title, no source, no information points, no identified entities. What I want to describe here is not the story of an empty file. It is the story of the professional reflex required when facing one: what to write, and more importantly, when to refuse to write.

Context: how the data infrastructure changed the way football is narrated

There is a historical gap that few Vietnamese football writers bother to reconstruct. On 15 December 2026, the Bosman ruling opened free movement for out-of-contract players within the European Union and turned the transfer market from a barter system into a pricing system. In 2026, Transfermarkt launched and, for the first time, published a public number for the value of almost every professional footballer on the planet. By the 2010s, StatsBomb, Opta, Understat and FBref had turned event data into a mass commodity.

Within twenty-five years, supporters moved from having only a scoreline and a commentator's voice to being able to look up xG, PPDA, progressive passes, field tilt and projected transfer value within three clicks.

In China, where I work, and in Vietnam, where I grew up, this data was absorbed faster than the discipline to use it was formed. An analysis piece can cite xG from Understat, minutes from FBref and market value from Transfermarkt, then conclude something about a player's future in a single sentence. The reader sees three sources and believes three verifications occurred. In reality, those three sources may all be copying a single data provider.

The deeper problem is structural. Once the data infrastructure looks complete, reader expectation shifts from "tell me about the match" to "give me a conclusion". That expectation creates a very specific pressure: fill every gap, including gaps the input data does not authorise you to fill.

That is why I am writing from a situation that seems meaningless: an analysis file returned with all nine analytical dimensions marked "insufficient information".

Core 1: Chiesa, and the trap of reading data through impressions

In June and July 2026, at twenty-one, I watched Euro 2026 and was captivated by Federico Chiesa like millions of others. The press called him the breakout star of the tournament. The basis for that claim was easy to verify: two goals for Italy, one against Austria in the round of 16, one against Spain in the semi-final.

I pulled the data and cross-checked three sources: FBref, Understat and StatsBomb event data. Across five matches: 1.8 xG, two actual goals, 41 per cent shot accuracy. Leading European wingers at the time sat at 45-50 per cent. He beat the model at finishing and sat below standard at shot selection.

I wrote a 2,000-word analysis for my personal blog arguing that Chiesa's Euro 2026 performance was structurally unsustainable because it rested on chance conversion far exceeding chance quality. On 9 January 2026, Chiesa tore his anterior cruciate ligament in Juventus's 4-3 win over Roma at the Stadio Olimpico. He returned after nearly ten months and needed much longer to recover his rhythm.

Chiesa does not break the data. He breaks the way we read the data.

Two layers must be separated. The statistical layer: 1.8 xG across two goals signals a high-conversion run, and high-conversion runs regress. The physiological layer: a winger whose style relies on maximum acceleration and repeated sharp changes of direction in tight spaces carries above-average ligament load. The two layers resonate. The correct conclusion is not "Chiesa will decline" but "Chiesa's risk structure concentrates in two places: conversion efficiency and ligament load".

On injuries, my professional position has been fixed since then: rushing back from ACL reconstruction is destroying the second phase of many careers. Psychological fear is harder to fix than the body.

Core 2: the empty Anfield and the fact that noise is data

In 2026, as the pandemic emptied stadiums, I was twenty and writing my undergraduate thesis. Liverpool lost six consecutive Premier League home games between 21 January and 7 March 2026, the longest such run in the club's top-flight history.

The first media reflex was to label it a crisis. My reflex was to isolate variables. I took Liverpool's PPDA and compared two periods. In 2026-20 the figure sat around 8.2. During the behind-closed-doors period it rose to roughly 12.5. A rising number means less aggressive, slower pressing, and easier build-up for opponents.

But the cause was not tactical. It was a variable the dashboard does not display: the crowd.

The empty stadium taught me that noise is data.

High pressing is a socially driven collective behaviour. It requires a player to leave position before knowing whether a teammate will follow. That decision is made in roughly two-tenths of a second and is underwritten by two things: trust in teammates, and social pressure from the stands that punishes retreat. With 53,000 people gone from Anfield, the social pressure vanished. The decision to step out shifted from default to optional. PPDA rose.

When 53,000 spectators fall silent, the numbers start talking.

A second variable cannot be ignored: on 17 October 2026, Virgil van Dijk tore his ACL in the 2-2 draw with Everton at Goodison Park after a challenge by Jordan Pickford. Losing Van Dijk means losing the player who covers the space behind a high defensive line.

Two logically independent variables overlapped in time. A writer choosing only the crowd variable produces an elegant piece that is half wrong. A writer choosing only Van Dijk produces a correct piece that is half empty. I built a three-column table: home, away, rest interval. Liverpool collapsed at Anfield while remaining acceptable away in the same period. If the sole cause were Van Dijk, the collapse would have been distributed more evenly. It was not. The correct answer is a function of both variables.

Core 3: the transfer market is where impatience gets priced

On 3 August 2026, Paris Saint-Germain activated Neymar's release clause at Barcelona for 222 million euros, still the highest transfer fee in football history. Its analytical value lies in the psychological anchor it set for the entire market.

The transfer market is where impatience gets priced.

Most bad deals are not bad because the player is poor. They are bad because of timing. A club signing a striker on 29 January is not buying quality; it is buying relief from pressure. Late-window fees reflect buyer panic more accurately than seller quality.

One structural loophole is underrated, and I will state my position plainly: signing-on fees for free agents are more toxic than transfer fees, because they sit outside the core surveillance of financial rules.

When Club A buys a player from Club B for 80 million euros, that number appears in the press, in filings, on the balance sheet and in every compliance model. It is amortised across the contract: 16 million euros a year over five years. Everyone can check it.

When Club A signs a free agent, there is no transfer fee. But there are signing-on fees paid to the player and the agent. These are disclosed with far less transparency and may be booked as a prepaid cost or a one-off charge. Either way, they are hard to reconcile against a monitoring system designed to track transfer transactions.

UEFA approved its Financial Sustainability Regulations on 7 April 2026, with a squad-cost ratio starting at 90 per cent of revenue and falling to 70 per cent from 2026-26. The Premier League's Profit and Sustainability Rules permit maximum losses of 105 million pounds across three seasons. Both systems assume wages and transfer fees are the main costs, and both can track them. Signing-on fees do not fit that assumption neatly.

In parallel sits a mechanism few supporters know. FIFA's solidarity mechanism distributes 5 per cent of transfer compensation to clubs that trained a player between the ages of 12 and 23. The money follows the player, not the selling club. In theory, a small Vietnamese academy that trained a player for four years retains a claim on future moves. In practice, these payments often go unclaimed because small clubs lack the legal department to track thousands of transfers a year.

FIFA banned third-party ownership effective 1 May 2026. The ban did not remove the economic incentive behind it. When a flow is blocked, it tends to find a detour: loans with purchase obligations, indirect co-ownership, intermediary structures.

The writer's task is to distinguish between a structure that is legal but ethically questionable and one that breaches rules. If no specific rule was violated, I do not call it a violation. If I only have a feeling, I label it a probability estimate with a confidence level.

Core 4: source tiering and the two-source rule

When I read a transfer story, I do not ask whether it seems true. I ask who reported it first, what access that person has, and who benefits from the information appearing.

Agents have an incentive to inflate prices. Selling clubs have an incentive to manufacture competition. Buying clubs have an incentive to cool supporter expectations. Writers have an incentive for engagement. Any source-grading system that ignores these incentives is easy to manipulate.

My internal rule: every number must come from at least two sources that are independent by supply chain, not by website count. Three sites copying one provider are one source. I also separate nominal value, probability-weighted expected value and annual book cost. These three figures can differ by forty per cent.

Core 5: risk first, and the value of saying "insufficient data"

When an analysis file arrives with no title, no source and no information points, the correct technical response is not to guess but to invoke null handling. All nine dimensions must be marked insufficient.

That looks like failure. It is actually a process success.

Data does not make revolutions. It only strips the paint off legends.

But data cannot protect itself from being fabricated. Only discipline can. A system without a null-handling mechanism will generate conclusions automatically, because every template has blank fields and human nature fills blank fields. A completed nine-dimension form looks identical to a genuine nine-dimension analysis.

Consider risk ratings. If a system finds no risks, the output looks like "low risk". But "no risk found" and "low risk" are different states. The first is a statement about your observational capacity. The second is a statement about the world. Merging them is a serious logical error and more dangerous than missing a risk, because it manufactures false safety.

In football this appears in softer forms. A club unpunished for two years is described as well governed. A player uninjured for three seasons is described as durable. Both confuse observed data with inference about mechanism.

Contrarian angle: correlation is not causation, and the filled-in report trap

The more data you have, the easier it is to mistake correlation for causation. With eighteen variables per match, you will always find one that correlates with results. With eighteen variables and a small sample, finding a spurious correlation is close to certain.

My most frequent check is to reverse the question. If this metric caused the outcome, it should move before the outcome. If a winning team presses little all season and still wins, low PPDA is a team attribute, not a cause.

A second, more expensive point: the biggest risk in modern football analysis is not bad data. It is a report that is filled in and therefore looks like analysis. A wrong number gets caught. A meticulously formatted report containing no information point passes through every layer of review unnoticed.

Vietnamese football writing imports models faster than it imports discipline. A piece can use xG without naming the model, PPDA without stating the sample, transfer value without naming the source. Readers gain the form of data without its value. That is information inflation: more words, less knowledge.

Takeaway: signals to track next round

Data does not erase emotion. It explains why emotion exists.

Three signals. First, the structure of signing-on fees for free agents: if free-agent deals rise while disclosed transfer spending falls, money is moving out of the observable zone of financial regulation. Second, ligament-load and sprint-volume metrics among players returning from long-term injury. Third, the number of Vietnamese analysis pieces citing data sources with publication dates attached.

Every number tells a story. The story is not inside the number.

A good analyst is not the person with the most data. It is the person who knows exactly when the data is insufficient, and who has the nerve to say so instead of filling the blank with a conclusion that sounds convincing.