The Blank Scouting Report: Enzo Fernández, Croatia 2026 and the Discipline of Reading Data

On 19 January 2026, I sat in front of a blank spreadsheet in an office in Shenzhen. At the top was a player name sent over by the scouting department: Enzo Fernández, then 21, playing for River Plate. Below it were fourteen rows of metrics waiting to be filled. The only row with a number was average distance covered: 9.8 kilometres per match. The other thirteen were empty.

I had two options. One was to write a smooth, polished report using the phrases every sporting director wants to hear: strong development potential, needs time to adapt, fits the club's philosophy. The other was to type into the empty cell the sentence nobody wants to read: insufficient information to conclude.

I chose the second, and then added the data I actually had. Enzo Fernández produced an xG chain of 0.45 per match in the Argentine top flight, inside the top five percent of the league according to the sample I had collected at the time. His average distance covered was 9.8 kilometres, below the club's internal benchmark of 11.2 kilometres for a central midfielder.

Those two numbers do not contradict each other. They measure different things. But the decision-maker kept only the number that was easiest to measure. He rejected the deal, signed a domestic midfielder instead, and filled the empty cell with belief rather than data.

Eleven months later, Enzo Fernández played a World Cup final for Argentina, and Chelsea completed a move that was widely reported at more than 100 million pounds. My spreadsheet from that day still had only one row filled in.

xG is not the truth - it is a compass, and a compass never offers a shortcut.

An empty cell is part of the data

In this profession, people are trained intensively in how to read a number, and almost never in how to handle a blank space. That is the biggest gap in professional scouting. A file with seven complete metrics and seven empty ones is not an incomplete file. It is a file with a fundamentally different level of reliability, and every conclusion drawn from it must carry a matching degree of uncertainty.

When I received the request to assess Enzo, I had three sources: event data from the Argentine league, video of the last three matches, and a report from a scout who had watched him live. The first was enough to build xG chain and xG buildup. The second was enough to assess his ability to receive the ball in tight spaces. The third was nearly worthless, because the scout had written down only general impressions.

What I did was keep the three sources separate rather than blending them into a single score. A single score creates a false sense of certainty, and false certainty is the most expensive commodity in any transfer deal.

During a transfer window, every file gets compressed into one question: buy or not buy. The correct question is different. Is the available data enough to answer at all?

Three layers of data and how they complement each other

Based on my experience watching matches in South America and Asia, a modern central midfielder has to be measured across three distinct layers.

The first is the physical layer: distance covered, number of accelerations, number of sprints above 25 kilometres per hour, and recovery capacity between matches. This is the easiest layer to measure, the easiest to compare, and therefore the most abused.

The second is the spatial layer: PPDA when the team is out of possession, passes into the final fourteen metres, receptions under pressure, and xG chain. This is the hardest layer to measure and the one that determines transfer value.

The third is the decision layer: pass quality under high pressure, ball-processing speed, and the ability to pick a position before the ball arrives. This layer can only be assessed through video and live viewing. No metric replaces it.

Enzo Fernández won at the second and third layers. He lost at the first. The man who rejected the deal read only the first layer and treated it as the whole file. His conclusion was not arithmetically wrong. It was structurally wrong.

Since that deal, every analytical report I write contains a dedicated section stating which metrics are missing, why they are missing, and how their absence affects the reliability of the final conclusion. That section is the longest in the report, and the one least often read.

Empty stadiums: isolating a variable

In 2026, when the pandemic shut down every European league, I found myself with no new data to analyse. The way a data professional handles that is not to wait, but to go back and re-examine what already exists.

I rebuilt PPDA data across five European seasons, separating matches with crowds from matches without. The result: average home-team PPDA before the pandemic was 9.6, and 8.9 when stadiums were empty.

A gap of 0.7 sounds small. It is not small. Lower PPDA means the home team presses harder. Empty stadiums made home teams press less, and the difference is equivalent to the gap between a mid-tier pressing side and a low-pressing side in the same league.

Empty stadiums were the largest laboratory modern football has ever had.

They isolated a variable nobody had previously separated: crowd noise. In every earlier model, home advantage was a mixed block of familiar pitch, travel schedule, refereeing, and crowd. In 2026 everything else stayed the same and only the crowd was removed.

My study was called Is the Crowd a Player, and it led to an official collaboration with a club in Shenzhen. But its real value lay elsewhere. It taught me that every number only means something next to the conditions that produced it. A PPDA of 8.9 in an empty stadium and a PPDA of 8.9 in front of 60,000 people are different numbers in substance, even when they are identical in value.

The 2026 season was not an exception - it was a test for every earlier hypothesis.

The same thing is happening in the current transfer window. Numbers still flow steadily, but the conditions producing them have changed: denser calendars, more international matches, and players moving between leagues mid-season. A distance-covered figure for a South American player cannot be compared directly with the same figure in Europe without accounting for fixture density and measurement method.

Croatia 2026: twelve percent does not win on its own

In 2026, while interning at a sports data company, I built a logistic model for the World Cup quarter-finals. Three input variables: PPDA, xG differential, and accumulated distance covered.

The model gave Croatia a 43 percent chance of reaching the final, against 29 percent for England. The entire data room laughed, because Croatia was seen as an underdog with a slow style and little pressing.

Croatia beat England 2-1 on 11 July 2026, after 120 minutes. Luka Modrić covered more ground than anyone on the pitch at the age of 32.

Croatia 2026 taught me this: a 12 percent probability is still a number worth betting on.

But that story is usually retold badly. Many people read the outcome and conclude that underdogs are always worth trusting. That is a poor conclusion. What the model actually said was that Croatia won because three conditions converged at once: a defensive structure that stayed organised across many matches, fitness sufficient to extend a game into extra time, and an opponent whose structure could be exploited on both flanks.

With any one of those three missing, 43 percent collapses quickly. Probability is only credible when it comes with convergence conditions. Strip the conditions away and probability becomes decoration.

From then on I wrote within a fixed frame: state the hypothesis, state the probability, state the model's confidence level, then compare it with actual results once the match is over. Publicly acknowledging the times the model was wrong matters as much as acknowledging when it was right.

Transfer-window noise and a credibility filter

The transfer window is the period when the signal-to-noise ratio drops lower than at any other point in the year. Dozens of transfer reports appear every day, and most of them come with no verifiable evidence at all.

I grade sources on four tiers. Tier one is confirmed by a club or the player himself. Tier two comes from a journalist with a strong accuracy record who names a specific agent. Tier three is mainstream media without a named source. Tier four is speculation from social media.

Tiers three and four account for most of the volume, and almost all of the engagement.

What stands out is that major deals are rarely decided by noise. Contract structure, release clauses, wage structures and payment schedules are where a transfer actually happens. A published fee of 80 million euros may consist of 60 million up front, 15 million in performance add-ons, and 5 million triggered only under conditions that are nearly impossible to meet.

In the transfer market, an 80 million euro figure can be a joke.

In another corner of the market, large sums are flowing towards leagues that sign players past their peak. Looking at the age curve, minutes played and output metrics, it is clear those deals are not aimed at raising the competitive level of the league. They are aimed at selling shirts, selling broadcast rights and selling images in new markets. For an analyst this matters, because it changes what every statistic means. A 34-year-old scoring 15 goals in a league with low defensive quality is not evidence of form. It is evidence of a quality gap.

We measure what is easy to measure, then call it the truth

This is the counterintuitive angle I believe in most, and it runs against my own job.

Football analytics today is dominated by a structural bias: we give higher weight to what can be measured, and unconsciously discount what cannot. Distance covered appears in every dataset, so it gets used. The ability to choose a position that removes the need to run has no direct metric, so it is undervalued.

The result is that players who run a lot are praised as hard-working, while players who run less but read the game well are treated with suspicion. Enzo Fernández belonged to the second group early in his career.

Every number is a witness statement; only the patient hear the full trial.

But this bias works in both directions, and that is the uncomfortable part. People who believe in data fall into the mirror-image trap: they treat xG as truth, treat the model as a verdict, and turn every match into a spreadsheet. Both groups are filling in the empty cell with what they want to believe.

Counter-evidence

Before locking in any conclusion, I force myself to write out the evidence that runs against it.

The Enzo Fernández case had a genuine weakness, and it was not distance covered. My sample at the time covered only 14 matches in Argentina plus three on video. That is a small sample. With a small sample, an xG chain of 0.45 can be inflated by two or three outstanding games, and the true value may be substantially lower.

On top of that, a midfielder covering 9.8 kilometres in Argentina might cover 11 in Europe, but he might also suffer a muscle injury from the sudden jump in intensity. Without medical data, I cannot rule that out. The sporting director was wrong in how he decided, but the risk he worried about was not imaginary.

This is the point I want to make clear: data discipline does not mean the data is always right. It means stating plainly where the data is weak, and carrying that uncertainty into the conclusion instead of deleting it to make the story tidier.

Signals for the next cycle

If I have to pick one thing to track going forward, it is the number of empty cells in a scouting file, not the value of the cells that are filled.

A file with many blanks that clearly labels which ones are blank is a trustworthy file. A file so complete that it leaves no room for uncertainty is a suspicious file. When a report on a 19-year-old in some second division already contains defensive metrics, attacking metrics, psychological metrics and a five-year development forecast, the reader should ask where the data came from, rather than why the player has not been signed yet.

For me personally, the thing worth tracking is whether the scouting industry accepts bringing uncertainty back into the process as an official part of it. When a sporting director reads a report and sees the line saying there is insufficient information to conclude, will he reject the report, or will he ask for more data?

The answer to that question determines the value of an entire generation of analysts coming after us.

I do not believe in luck - I believe in a large enough sample. But I also know that most of the time, our sample is not large enough. The job of a professional is to say so, every time, even when it makes the report look less convincing.