The Empty Risk Box: The Quiet Trap of the Transfer Window

THE EMPTY RISK BOX: THE QUIET TRAP OF THE TRANSFER WINDOW

In January 2026, in a meeting room in Boston, I was handed a twenty-six-page scouting file on a twenty-three-year-old forward playing in a European second tier. The file was beautifully built: a physical section, a technical section, a psychological section, an integration section, a comparison against three players in the same position. The last page was titled Risk Matrix. That page was blank. Not a single box was marked, not a single red flag was raised.

The sporting director read the blank page and drew his conclusion: this player is risk-free. I asked one question: who checked? Nobody could answer, because the honest answer was that nobody had checked. That player's league does not publish detailed injury data, has no public GPS tracking, and the scout had only watched three matches on video. The risk page was blank because there was no data to fill it with, not because risk was zero. An empty risk box is a question nobody has asked yet, and it looks exactly like an answer. In eighteen years of working with data, I have never met a faster-killing trap: the silence of data read aloud as a clean certificate.

The night of June 24, 2026 at Foxborough taught me the first lesson about that kind of silence. It took me a few more years to understand all of it.

The transfer window and the blanks nobody counts

The transfer window is the densest information market of the year, and also the noisiest. Every day brings hundreds of lines about negotiations, release clauses, buy-back options, signing fees, agent commissions, medical markers. Most of it is noise. But the danger of the transfer window does not live in false information. False information still makes people suspicious. The danger lives in missing information: a file with no injury data, a club that does not disclose its wage structure, a league with no motion-tracking numbers, a player with no senior international history.

European football has travelled a long way with data, but the travel has been uneven. The Premier League publishes tracking metrics down to the individual phase of play. Some Nordic leagues disclose medical data and running volume. In many other leagues, all you have is a scoreboard, a team sheet and a three-minute clip. When a data-rich club buys a player from a data-poor league, it carries a dangerous habit with it: it reads a thin file to the standard of a thick file. It fills the blanks with feeling, then calls that feeling experience.

I call this the silent-failure error. A report full of empty boxes looks exactly like a clean report. Neither has a red flag. The reader cannot tell them apart unless the writer says so. And in my trade, the only thing that separates them is provenance: who measured, with what, when, and across how many matches. Without that provenance section, every number in a file is uncitable, and every conclusion drawn from it is a suspended conclusion.

The Foxborough night and the first con

In June 2026, New England Revolution hosted Toronto FC at Gillette Stadium. Toronto held 72 percent of the ball, fired twenty-one shots, and finished the match with 2.3 expected goals. The final score: New England won 1-0, the only goal scored by Diego Fagundez.

I was an intern writing match reports then. My editor asked me to celebrate the heroics of the home defence and goalkeeper. I refused, pulled the data from StatsBomb, and wrote a piece headlined Toronto deserved to win 3-0. It reached fifty thousand reads in twenty-four hours, and the editor had to publish a correction. The lesson was not that the data was right. It was that the scoreboard is not wrong; it simply answers a different question from the one people think it answers.

A match holds two kinds of truth. The first is the result, recorded instantly, legally binding, repeated thousands of times over the following twenty years. The second is the process, erased from collective memory within forty-eight hours. The result is the con that time has memorised; xG is the testimony. But I have to complete the thought: xG is not truth either. xG is a testimony, and every testimony has a witness, a moment of telling, and a motive for telling. An xG model built on a league without ball-location data produces far cruder numbers than a model built on a league with goal-mouth tracking. Reading two xG figures from two different systems and comparing them directly is a basic error, and in the transfer window that error repeats every single day.

After that 2026 piece I dropped emotional match reporting entirely and set myself one rule: if the numbers contradict the story, trust the numbers. Then I added a second clause, which later turned out to matter more: if the numbers do not exist, do not write a story in their place.

Croatia 2026, PPDA and the table that measured pride

Thanks to the piece that spread in 2026, I was invited to build data for a new sports platform during the 2026 World Cup. Before the quarter-finals I constructed a PPDA table for all thirty-two teams. Croatia's figure was 8.9, meaning that in each defensive phase they allowed the opponent an average of 8.9 passes before the ball was blocked or recovered. It was the lowest figure among the remaining eight teams.

I wrote about Marcelo Brozović: 13.8 kilometres covered in a single match, nine ball recoveries against Argentina. I put a question in the piece: Croatia does not have luck, Croatia has a system. When they reached the final, I became a name mentioned on data programmes, and a Championship club hired me as a part-time consultant.

But the thing I want to say here sits outside the table. The Croatia PPDA table of 2026 did not measure pressure; it measured pride. The figure 8.9 does not say Croatia ran a lot. It says Croatia refused to be stretched. A team that accepts sitting deep, accepts letting the opponent keep the ball, but sets an absolute ceiling on the number of passes it will pardon. That ceiling is a collective psychological decision, encoded into a number. Reading the number without reading the decision behind it is reading the label without reading the goods.

I realised something about my own trade. A model is only a language. One person can say PPDA like a parrot, and another can say PPDA like someone who understands exactly where that team's pride sits. In the transfer window, the difference between those two people is measured in money.

The empty stadiums of 2026: a natural experiment that asked nobody's permission

In early 2026 the pandemic froze the world and the stands went empty. The Boston consultancy where I worked as a mid-level employee cut forty percent of its staff. I did not ask for an exemption. I wrote a report: Stadium Effect, evidence from 372 Bundesliga matches before and during the behind-closed-doors period. The dataset showed the home win rate falling from 45 percent to 31 percent, and penalties awarded falling by 28 percent.

The empty stadiums of 2026 were a natural experiment: football does not need a crowd to expose its nature. When the roar of twelve thousand people disappears, what remains is eleven individuals plus a coaching staff. The crowd then appeared as a number: how much is home advantage worth, measured in points. The answer the Bundesliga gave was fourteen percentage points in win rate, and twenty-eight percent in penalties.

Huddersfield Town hired me to consult for the final eight rounds of the Championship. I proposed a rotation model based on sprint distance above six metres per second: any player who ran below 80 percent of his personal threshold in two consecutive matches would be benched. They took fourteen points from twenty-four and stayed up by exactly one point.

That report left me with something uncomfortable. The 372-match dataset had enough matches, enough teams, enough control group. But there was one blank I could not fill: I could not measure what a player feels when he scores in front of an empty stand. Which means a good model can still contain empty boxes, and what decides its quality is whether it declares those boxes instead of filling them with assumptions.

Morocco 2026: two testimonies that never reached the scoreboard

The 2026 World Cup was the year I learned the most from a team nobody ranked among the favourites. Before the tournament I published a series arguing that Morocco defend, and that they operate on data. But the word data here needs separating from the word data everybody uses.

Goalkeeper Yassine Bounou finished the tournament with goals prevented 4.3 above expectation. Achraf Hakimi made 6.8 progressive passes per match. Those two numbers say very different things. Bounou's number is a differential number, meaning it only means something against a baseline built on shot quality and starting position. Hakimi's number is a volume number, meaning it only means something once you know where Hakimi plays on the pitch and what he has been asked to do.

I predicted Morocco would reach the semi-finals. When they eliminated Portugal 1-0, international platforms called me. And the detail I remember most is this: most of those calls asked the same question. Can the next team repeat it. I could not answer. Because what made Morocco 2026 sits inside a structure that is not publicly disclosed: how they divided defensive responsibility between midfielders and full-backs, how they accepted surrendering the ball in certain zones. On the open data table, that part is a blank box.

Ronaldo, real xG and inflated xG

In the summer of 2026, a Saudi investment fund asked me to appraise Cristiano Ronaldo for a contract extension, and paid for a forty-page report. I did my part. Ronaldo's actual expected goals created in that period was 0.55 per match; once set-piece situations were included, the figure was pushed up to 0.82. The gap between those two numbers is an expensive piece of testimony: a significant share of the output was generated in situations the team fully controlled, not in live phases that demand reaction speed and the ability to create separation.

I recommended against paying more. The fund objected. Three months later, Ronaldo's market valuation dropped 15 percent. I do not tell this story to praise myself. I tell it to offer an analytical structure the transfer window lacks: separate the gloss of media effect from real capability, then state the financial warning threshold explicitly. The market does not price players by quality. The market prices by supply and demand, by name, by league and by nationality. Transfer data is like a tide: looking at the surface tells you nothing, you have to measure the seabed.

And on the seabed there is always a blank every report should declare: workload data at thirty-eight years old differs from workload data at twenty-three. Same metric, two different levels of confidence. That is why I now add a column called confidence to every report, where I grade the data I am using.

Back to the last page of the file

Back to January 2026. That twenty-six-page file contained not one numerical error. It contained one structural error: it treated an absence of data as a signal of safety. Within three weeks of that meeting I built a seven-point screening process for every incoming file, and at each point all I need is a binary answer.

One: where does the injury data come from, and is there a clear time matrix. Two: is physical output measured by GPS or by observer impression. Three: how many full ninety-minute matches has the scout actually watched. Four: how many positions has the player appeared in, and for how many minutes in each. Five: the current contract structure, the release clause, and the share of economic rights held by third parties. Six: the wage structure of the buying club, to know how a proposed salary reshuffles the internal hierarchy. Seven: the blank box. To me the seventh is the most important: an explicit list of what nobody could verify. A file that cannot list its own blanks is an unfinished file, however beautiful it looks.

Release-clause structure and wage bill are the real story of the transfer window. A cheap release clause plus a high salary creates a category of risk the scoreboard never displays. A player performing well in a data-poor league, then paid like a star of a data-rich league, produces a gap called expectation. That gap is written down nowhere. It sits in the blank box.

The silent-failure trap

This is where I have to argue against myself. Traditional scouting thinking holds that missing data is a minor problem that the human eye can compensate for. I hold that it is the largest problem, and that its largest feature is invisibility.

A club that receives a report with three red flags will hold a meeting. A club that receives a report with no flags will sign. Between those two situations, the second is far more dangerous, because the decision-maker believes he has just been given complete information, when in fact he has received a blank form. In esports compliance, we work on a principle: silence is not exoneration. A category that cannot be screened must be recorded as unresolved, and must never be recorded as compliant. Football is missing this principle at the scouting stage.

I also have to guard against the opposite reaction: using blanks as an excuse not to decide. Some clubs turn caution into a habit of delay, and lose a player every window waiting for more data. The transfer window has a clock. A perfect file on January 20 is worth less than a good-enough file on January 3, because the player has signed elsewhere. My job is not to demand perfection. My job is to state clearly how thin the data is, so the decision-maker knows what he is betting on rather than betting blind.

One more error I warn myself about. I carried a habit from esports into football: the belief that everything can be measured. Esports logs every millisecond. Football is not like that, and never will be, because football has eleven bodies in open space, grass, wind, a referee, and a ball that touches a player's foot for less than two minutes on average across ninety. Pressure in football is not measured by the number of runs. It is measured by the number of passes a collective refuses to give away. That is why PPDA taught me in 2026: pressing is not running a lot, it is running at the right moment. I never quit data, I only changed suppliers.

And my current supply is missing something. In transfer files, the emotional section is treated as filler. But when I ask a player why he has underperformed for three straight matches, the answer sits in no metric. It sits in a wife who has not obtained a visa, or a child who has no school. A model that treats that as noise will forecast wrongly, and will forecast wrongly with confidence. xG judges nobody; it merely exposes the truth the result conceals. What both the result and xG conceal together is the human being.

The call for the next cycle

The next transfer window will offer one signal worth tracking, and that signal is not in transfer fees. It is in how many clubs begin publishing blank boxes inside their scouting reports. The first club to circulate a file stating plainly that we have no reliable medical data for this player over the last three years will be laughed at in the meeting room. Then they will win, and nine months later the whole league will copy the template.

I am not waiting for a better model. I am waiting for a new habit: reading silence as a statement. The result is the con that time has memorised; xG is the testimony. But the most credible testimony in the transfer window belongs to the person who dares to say he does not know.