The Empty Analysis: Football Data Journalism and the Trap of Sourceless Conclusions

2:40 a.m., seventh floor of an old building in Gràcia, Barcelona.

I opened my inbox and found a file from a young collaborator. Three thousand words. Ten subheadings. Three carefully colour-coded tables. The first line declared that a La Liga club had "completely lost control of midfield" across its last three matches. I read to the final line, went back to the first, and did what I have done for ten years: I went looking for sources. There were none. No match matrix. No PPDA figure. No event-data link. No timestamp. That three-thousand-word building had been erected on a total void.

What kept me awake was not the carelessness. I have seen that for a long time. What made me sit up was the confidence. The writer did not hesitate, did not ask a single question, did not acknowledge a single limitation. And I recognised that I was looking at a pure specimen of the fastest-spreading thing in my profession: analysis born from a void, yet presented as though it were born from data.

In the summer of 2026, I saw the Opta ghost – and from that day, my eyes have not trusted what they see. But I kept one other belief: an article can say "I do not know" without losing credibility. In fact, that is the only way to keep it.

The industry of gaps

My trade is now measured in two units: length and latency. Everyone sees length. Only insiders feel latency. During a transfer window, the average latency of a rumour – from appearance to confirmation or denial – runs from a few hours to a few days. But most content produced inside that window does not wait for the outcome. It goes out immediately, in language soft enough that it never needs correcting.

The incentive structure here is plain. A news site needs traffic. Traffic comes from coverage, not from accuracy. A wrong analysis published fast can gather many times the readership of a right analysis published three weeks late. In the economics of attention, error costs almost nothing while hesitation costs a great deal. The result is an ecosystem where the speed of producing conclusions is rewarded, and verification is treated as an unnecessary cost.

About fifteen years ago, when I still wrote for print, every analysis passed through an editor, a fact-checker, and a slow but real cross-referencing process. Now an analysis passes through a publish button. The process did not vanish because someone sabotaged it. It vanished because it was no longer technically necessary, and because it was never counted in any success metric.

I am 68, and I have seen three generations of this trade. The first wrote with their eyes. The second wrote with their eyes plus video. The third writes with event-data tables. Each generation produced the illusion that it was more objective than the last. All three share one weakness: the writer is never required to prove that he actually read the thing he is describing.

The transfer market is a monastery where numbers chant; I merely transcribe what they pray. But when I opened this season's transfer pages, I heard something else: an echo. Articles reflecting one another, each loop adding a little more confidence and not a single new piece of evidence.

Dissecting an empty analysis

I spent three weeks doing something strange: classifying empty analyses by their shape. The results were not random. They have structure, rhythm, their own grammar.

The first pattern is the "treatise". This is the style that uses dense jargon to hide the absence of any data behind it. A typical sentence: this team "loses structure in transition". It sounds professional. But to say that, the writer must show at least one thing: the number of counterattacks conceded per turnover, or passes intercepted in their own third, or the average distance between lines when the ball is lost. Without those, the sentence is merely music.

The second pattern is "floating numbers". The writer inserts a percentage – a figure with no sample, no source, no time window. Seventy-two percent of what, across how many matches, under which definition? A metric without a definition is not data. It is an adjective written in digits.

The third pattern is "goalpost drift". The piece opens with a verifiable claim, then quietly slides into an unverifiable one. For example: "this team completed 88% of passes over the last three rounds" – verifiable. Two paragraphs later: "which shows they have lost their identity". The second sentence is no longer data. It is a judgement inheriting credibility from the first.

The fourth pattern, and the one that worries me most, is "the gap filled with silence". The writer knows the data is missing, so chooses to say nothing about the missing part. The article remains full, coherent, and conclusive. It is just that the conclusion is supported by nothing beyond its own fluency.

All four patterns share one trait: none of them can be proven wrong. And a claim that cannot be proven wrong is not analysis. It is ornament.

I learned this painfully. In 2026, when I moved from print to an online platform, my first data-driven piece covered Valencia's 3–0 win over Las Palmas on matchday two of La Liga. Valencia scored three but generated roughly 1.4 xG. Las Palmas pressed ferociously, with PPDA dropping to 7.2 – a figure meaning they lunged into almost every opposing pass. They conceded three, largely because their back line pushed high while midfield rushed forward. I wrote that, backed by three separate data sources.

The newsroom was politely silent. A colleague told me he could not understand why I "looked at tables without watching the match". He was half right. I did watch the match. But I did not trust what I saw until the table said the same thing. I then spent three weeks building a homemade xG model and validating it across the first 76 matches of the season, to answer one question only: were my eyes deceiving me?

The answer was yes. Which is why I am writing this.

On the Moscow night, I did not sleep. Not because of football, but because the numbers were whispering a prophecy.

Thanks to the credibility of those La Liga pieces, I was invited to write a data column for the 2026 World Cup. I published a prediction that France would win, at a point when they had impressed nobody in the group stage. My basis lay in two things: the French U21 side held the highest rate of passes into the opponent's final third among leading youth setups, and Antoine Griezmann's average shot at that moment carried roughly 0.21 xG per attempt – above the average for Europe's leading forwards. The piece was dismissed as "dry as roof tiles". When France lifted the trophy, a Spanish editor told me: "You were right, but nobody reads the way you write." That night I wrote in my notebook: truth needs to be told with emotion, not only with digits.

When the stadiums fell silent in 2026, I understood something: football never died, it merely stripped off its clothing and revealed its skeleton. In the behind-closed-doors period, I was granted access to real-time data from a Segunda-tier club in Catalonia. Home win rates fell from roughly 46 percent to roughly 38 percent. But passes into the final third rose by about 11 percent. Two metrics moving in opposite directions, and both of them correct.

That is the most important lesson football has taught me in twenty years: two metrics can be simultaneously correct and simultaneously meaningless if the reader does not understand what they are measuring. A falling home win rate does not prove that crowds create an advantage. It shows a correlation. Turning correlation into causation demands something else: a mechanism, a test, a chain of evidence long enough to rule out alternative explanations. Most of the analysis I read this season jumps straight from correlation to causation without pausing for a single second in between.

The real structure of a transfer window

In a transfer window, the information gap is not an accident. It is a product.

When a club negotiates, at least four parties hold fragmented information: the selling club, the buying club, the agent, and sometimes a third party holding economic rights. Each has different motives to leak or to stay silent. The agent leaks to create price pressure and widen the market for his client. The selling club leaks to push the price higher. The buying club leaks to reassure supporters or to apply pressure on a different target. And across all those currents, the one thing never leaked honestly is the structure of the deal itself: add-ons, contingent clauses, payment schedules, and sell-on mechanisms.

That is why I tell young colleagues: do not read the headline number. Read the footnote. A deal announced at eighty million euros but paid over five years, with forty million contingent on collective and individual achievements, and fifteen percent of any future sale – that is a deal with a completely different risk profile from one paying eighty million up front. Same figure. Two structures. Two stories.

And this season I keep seeing the same error repeat: deal rankings sorted by headline number, not by structure. Such a ranking is not wrong arithmetically. It is wrong structurally, and in modern football structure is what determines who survives the next three seasons.

The same holds for injuries. Medical confidentiality places supporters and media in a state of controlled blindness. Clubs publish recovery timelines in ways that favour their asset values. A knee problem described as "minor" can be ligament damage. A return scheduled "within weeks" can be three months. Nobody lies directly, and nobody tells the whole story. So any analysis built on club injury communication must carry a risk flag from the first line.

I once believed in feel. After Opta, I believed in probability. After COVID, I believed in structure. Structure is the only thing that does not change when context changes. A club can change manager, formation, and star player. But wage structure, contract structure, and academy infrastructure change far more slowly. And it is those slow things that predict what happens eighteen months out.

Two fields where gaps are best hidden

I want to pause on two fields I have followed for years, because they show that the data gap is not only a journalism problem. It is a structural problem.

Esports has a measurement paradox so elegant it is uncomfortable. In-match data is enormous – every input, every movement logged at high frequency. But data about the profession itself barely exists in public, verifiable form. Average career length, post-retirement income, labour-market re-entry rates – those figures are rarely published to statistical standards. The consequence is concrete: when a discipline invests only in measuring what happens on stage and never measures what happens to people after they leave it, every analysis of that discipline's sustainability is guesswork.

Women's football sits on the opposite side of the same problem. Here, commercial figures are published fairly regularly, but they are usually packaged for a specific communications purpose: social responsibility reports, sponsorship dossiers, equality campaign releases. When a women's league expands in parallel with a sponsor's media campaign, I always ask myself what within it is long-term investment and what is an indispensable component of a brand strategy. The difference between those two only becomes visible after three years, when the campaign ends and the broadcast contracts have to stand on their own feet.

Both fields teach the same lesson: abundant data does not mean sufficient data. The question is not how many metrics exist, but which metrics exist because they are useful, and which exist because they are easy to publish.

The contrarian angle: honesty is becoming a competitive advantage

This is where I turn against myself.

For ten years I argued that carelessness is the enemy of data journalism. I still believe it. But I have changed my view on its consequences. I used to think that when misinformation floods in, readers would grow suspicious and leave. That did not happen. Readers did not leave. They shifted state: they read for entertainment, not for understanding. And when readers read for entertainment, conclusions become part of the game rather than claims about the world.

An empty analysis functions more as a market segment than as a crisis. And that means honesty – saying I do not know, saying the data does not permit a conclusion, saying this analysis cannot be executed because the input is empty – has become a competitive advantage rather than a limitation.

I realised this rather late, on an evening when I had to return an assignment without writing a single word. I told the editor the input contained nothing to analyse, and that any conclusion I produced would be fabrication. I assumed I would lose the work. Instead, my next piece – a straightforward account of why I could not write – drew more readers than any ordinary analysis I published that year.

Readers do not reward silence. They reward the reason for silence. And in an ecosystem where everyone has an opinion about everything, the ability to explain why a question cannot be answered is becoming the scarcest skill of all.

But I must raise a warning here, because I detest laziness sheltering under the name of honesty. Saying "not enough data" is a valid conclusion only when the speaker has genuinely gone looking. Otherwise it is simply another form of carelessness, dressed up as performative humility. Between "I do not know because I have not searched" and "I do not know because I searched and there was nothing" lies an enormous gap in professional ethics. An empty analysis has value only when it is the outcome of a complete process, not when it is the starting point.

And here I must admit something about myself. At 68, my first instinct on receiving an under-sourced document is to doubt the sender's quality. But I have learned that this scepticism can be a form of laziness in reverse: blaming the input in order to avoid working with it. When an analysis reaches me with every field blank, the first move is not to judge. The first move is to check whether the gap belongs to the sender or to the world. Those two kinds of gap demand entirely different responses.

A signal for the next cycle

I am 68, but the data is younger than I have ever seen it – every season it grows another set of teeth. The tools now arriving will make verification far cheaper than fabrication. When the cost of checking falls below the cost of inventing, the entire incentive structure flips.

I do not know exactly when. But I know which signal I will be watching: editors who start paying for verification instead of paying for length. When that happens, empty analyses will disappear – not because they are wrong, but because they will have become more expensive than the truth.