The Empty Report and the Discipline of the Football Analyst

The Empty Report and the Discipline of the Football Analyst

In the winter of 2026, in a rented flat in Marseille's 8th arrondissement, I laid eleven VHS tapes in a row on the wooden floor. Three days earlier, José Mourinho's Porto had beaten Monaco 3-0 in the Champions League final at the Arena AufSchalke. I watched that match eleven times within seventy-two hours. On the eleventh viewing, I paused at minute 67 and noticed the most important thing in the whole process: my notebook had one completely blank column, labelled "intention." I had recorded every pass, every off-ball metre, every duel. I had not recorded the players' intentions.

That moment shaped the rest of my career. An honest blank column is worth more than a filled column of fabricated numbers. Twenty-two years later, sitting in front of data dashboards in Marseille, I see the football analysis industry filling that blank column with something more dangerous than silence: fluent prose with no data anchor.

The context of the contest

The football analysis industry has changed in ways nobody in the Marseille press room in 2026 could have imagined. When I entered the profession, a good analyst sat in a studio with a notebook and recorded every phase the naked eye could catch. Today, a single match in a top European league generates millions of positional data points, and every player is tracked by at least two optical camera systems plus a sensor in the captain's armband. Volume of data is no longer the problem. The problem lies elsewhere: very few of us are trained to say "I don't know."

In experimental science, where I earned my master's in movement science, a null result is not a failure. It is a finding. When an experiment produces no signal, the researcher records exactly that and states the limits of the method. If you invent a signal that does not exist, you do not merely deceive colleagues — you corrupt the entire chain of knowledge on which those who follow will rely. In football, that chain is corrupted every day, and those corrupting it often do not know what they are doing.

Last week I received a nine-dimension analytical report on a match that did not exist. Every data field was empty: no title, no source, no information points, no named entities. Only one label was filled in — "football." The machine had run the correct schema, returned a structurally valid object, and inside it there was not a single fact to analyse. The striking thing was not the technical failure. The striking thing was the first reaction of many around me: "Then write something, readers need content."

That is precisely the breaking point. A system designed to produce fluent text from an empty input will produce fluent text. It will not stop and say it has nothing to say, because stopping is not in its architecture. Neither are we. We are rewarded for confidence, not accuracy, and that reward is creating a dense layer of content of unverifiable claims.

The trade of blank columns

In 2026, I sat in the Olympique Marseille press room after a 1-3 defeat to PSG. I was twenty-eight, the only female researcher in the room. I asked about the gap between midfield and the left flank. A male journalist sneered and asked whether women watch football with emotion. I did not answer directly. I pulled out a movement map of twenty-two players I had drawn from video, and pointed to seven occasions on which Bixente Lizarazu was left open on the left channel. The room went silent. A week later, I was invited to write a tactical column for La Provence under the nickname "Madame Tactique."

The lesson I drew was not that women can analyse football. The lesson was that precision comes from the smallest detail, and the smallest detail only has value when it is counted correctly. Seven times left open is a verifiable number. "Marseille's defence played loosely" is an unverifiable sentence. Both can be said in the same confident tone. Only one survives the pressure of video.

Since then, my rule has been to begin every piece with a concrete number or phase, and never to let a claim stand without a base. But there is a paradox that took me years to accept: my trade is the trade of blank columns. Most of what happens on a pitch cannot be measured with existing data. We measure passes, not decisions. We measure positions, not fear. We measure running speed, not hesitation.

An honest analyst works with those blank columns every day. He marks them, circles them, and tells readers that this zone is unexplained. A dishonest analyst fills them with dressing-room language — "hunger," "character," "fighting spirit" — and turns unmeasurable words into persuasive weapons. Both write pieces thousands of words long. Only one holds up when the next match is played.

Porto 2026: an equation, not a miracle

When people look at Porto 2026 and see a miracle, I see an equation waiting to be solved. This is the sentence I wrote at the top of the twelve-thousand-word analysis I completed after two weeks, based on eleven viewings of the final. Colleagues found it dry. A university lecturer in Lyon used it as teaching material. Both reactions were true to its nature.

The basic data of that match: Porto had less possession, only around 43%, but created five clear goalscoring chances, while Monaco created one. That is a paradox explicable only by geometry. Mourinho did not try to control the ball. He tried to control the space into which Monaco was forced to pass. Didier Deschamps' Monaco played a game based on circulating the ball through midfield, and Porto built a structure of four synchronised moving blocks, narrowing the lateral passing lanes in midfield, then trapping opponents in precisely the channel they thought was empty.

I described this mechanism with the geometric symbols X, Y, Z in my analytical table. The X axis was the width of the pitch, the Y axis its length, and the Z axis the time a pass needs to travel from sender to receiver. When Porto forced Monaco to pass with a Z value above the safety threshold, Monaco's turnover rate spiked, and every turnover was a Porto counter into the space Monaco had just opened behind their advancing back line. Porto's three goals were not three separate moments of inspiration. They were three instances of the equation producing the same result.

Based on my experience watching matches at this level, I can say that most "miracles" in European football share a similar structure. A team weaker in personnel finds a variable the opponent cannot control, then optimises that variable to an extreme. Porto 2026 optimised pass time. Other teams optimise set pieces, or duels in the opponent's half, or match tempo. A miracle is the name we give to an equation not yet solved.

On my twelfth viewing, in 2026, I found a new detail. In the second half, Mourinho moved the starting point of the central defensive block about five metres lower than in the first half. He did not change the personnel. He changed the coordinates. Monaco took thirty minutes to notice, and by then the score was 2-0. This detail appears in none of the commentary of the time, because measuring defensive-block coordinates by half requires a process almost nobody applied in 2026.

Collapse is not the end of the tunnel. It is the largest dataset life provides. Monaco's failure in that match is a perfect dataset on how a system reacts when stripped of a familiar variable. Deschamps kept his ball-circulation structure until the seventieth minute before changing. That is thirty-one minutes of pure data on delayed tactical recognition. If I were a young coach, I would use those thirty-one minutes as the first lecture.

The mechanism of fabrication

There is a gap between "insufficient data" and "no data," and that gap is where fabrication lives. When a field is empty because the source did not provide it, that is insufficient data. When a field is empty because the event was never observed, that is no data. Both must be recorded clearly. In practice, both get filled with speculation shaped like fact.

I call this mechanism "schema-shaped fabrication." A system asked to produce a nine-dimension report will produce a nine-dimension report. If the input is empty, it will still produce all nine sections, each with a heading, tables, and a conclusions section. The external shape is perfect. The internal content is zero. And a reader looking only at the shape cannot tell that report apart from a real one.

This is more dangerous than an ordinary wrong article, because an ordinary wrong article can be caught by checking facts. A schema-shaped empty report cannot be caught by checking, because there are no facts to check. It can only be caught by one question: what is the source of this claim?

In my industry, that question is rarely asked. We read a long analysis, find it fluent, find it has numbers, find it has firm conclusions, and we believe it. We do not check where those numbers came from, by what method they were collected, on what sample, under what conditions. That is why I always put a data footnote at the end of a piece, and why I treat the absence of footnotes as a red flag more serious than a wrong conclusion.

One case I have tracked for years is the misuse of xG. Expected goals is an index of chance quality based on shot coordinates, shot type, number of defenders, and other variables. It is useful when used to compare chance quality between two teams in a large sample. It is meaningless when used to conclude about a single match, because a single match is a sample with enormous variance. I have read hundreds of pieces concluding that a team "deserved to win" based on an xG gap in one match. That is using a large-sample tool for a small-sample question.

The same mechanism repeats with PPDA — the number of passes opponents complete before each defensive action. This index measures pressing intensity. Lower values mean more aggressive pressing. But PPDA does not distinguish organised pressing from chaotic pressing. A chaotically pressing team can have lower PPDA than an organised pressing team, while the latter is in fact far more effective. If you do not place the index in its structural context, you will read a conclusion opposite to the truth.

And here is the point I want to stress: an index detached from structural context systematically produces conclusions opposite to the truth, not randomly. That is why listing numbers is not analysis. That is why I never let a data table stand alone in my writing.

The economics of transfer rumours

Transfers are the market of hope, and hope rarely follows valuation. This market is the perfect environment for schema-shaped fabrication, because every rumour can be constructed with a full structure: an anonymous source, a fee figure, a timeline, a motive. Nobody can verify until the deal closes, and when it does not close, the rumour disappears leaving no trace of accountability.

I track this market from a structural angle, not a news angle. When a club in a lower European league develops an outstanding young player, and he shines for a season, I do not look at the list of big clubs supposedly interested. I look at his contract structure. How many years remain? Is there a release clause? A sell-on clause? Does the parent club need to sell for financial balance?

The answers to those questions predict a player's future more accurately than any rumour. A player with two years left at a club that needs money will be sold within eighteen months. A player who has just signed a new five-year deal will go nowhere, no matter how many clubs are supposedly interested. Contract structure is a constant. Rumour is a noise variable.

I have observed a recurring pattern over many years. A small club achieves a surprise result through a group of well-coached young players and a manager who has built a fitting system. Within two transfer windows, big clubs come and take away each core player. That club does not collapse immediately, but the system collapses, because the system was built around specific people with specific characteristics. The small club's success becomes the opening form of another talent raid. This is a law of the system, and it does not depend on fans' feelings.

At the same time, another part of the market is operating in the opposite direction. Leagues in the Gulf bring in European stars past their peak with large contracts. Tactical-level analysis reveals a structural problem: a league with only a few late-career stars cannot create a competitive system. A system must be built from youth development, from a domestic league with depth, from a playing culture formed over generations. Aging European stars are a catalyst for media, not for football. They become image ambassadors for a tourism market rather than pillars of a football nation.

That is a signal transfer data can read. When a league recruits many players over thirty on high fees, that is a signal of a commercial strategy, not a competitive one. When a league recruits many young players on low fees and long contracts, that is a signal of a building strategy. The two signals mean entirely different things, and they are often merged under the same headline "league X is rising."

The problem of selling predictions

There is a dark side effect I rarely write about directly, but which is always present in my analytical choices. When football became fully digitised, live data on every phase could be sold to betting companies in the same window it reaches analysts. The same data stream serves two purposes: decoding the match and pricing probability. In many cases, the party paying more is the second.

This creates a distortion in the industry's development. Data companies optimise their products for betting-market needs, not analytical needs. Indices are designed to predict outcomes, not to understand mechanisms. When analysts use those indices for the second purpose, they are using a tool for the wrong job, and they often do not know.

I do not write about prediction. I write about mechanism. The difference matters: a correct prediction can be produced from a misunderstood mechanism, and a correct mechanism can produce a wrong prediction in a single match. Football is a high-variance system, and anyone who claims to know the outcome of a single match is selling you a product they do not own.

The contrarian angle: the most dangerous part of an empty report

When people receive an empty report, they usually look only at what is missing. They see blank fields and ask how to fill them. But the most dangerous part of an empty report is often the one part that is filled in.

Imagine a report with ten fields, nine blank and one containing a classification label. Most readers will focus on the nine blanks and ignore the label. But that label is a claim. It says this content belongs to a specific field. It was produced by a classification algorithm, and classification algorithms have error rates. If the original content is not football, that label is a false claim, and it sits in the only position a reader might trust.

This is the structure of a subtle failure in an analytical system: when everything is blank, people conclude the system failed. When one field is filled wrongly, people conclude the system succeeded. The detectability of an error is inversely proportional to the amount of information around it. An error sitting alone in a sea of blanks will survive longer than an error sitting among many verifiable facts.

I apply this principle to what I write. If a piece of mine contains one unverifiable claim among twenty verifiable ones, that claim is more dangerous than if it stood alone. Readers trust probability. The more correct details around it, the more readily a wrong one is accepted.

The contrarian view I want to propose is this: the best protection against misinformation is not to demand more information. Demanding more information produces more information, including more misinformation. The best protection is to demand traceability. Every claim must answer the question: where did it come from, when was it observed, by what method. A claim that cannot answer that question, true or false, has no value in a chain of knowledge.

The traceability standard and the writer's discipline

In my work I apply a four-layer traceability standard. The first layer is origin: whether information comes from direct observation, published data, or an indirect source. The second layer is timing: every fact must be tied to an absolute date; relative expressions such as yesterday or this week are not used. The third layer is units: a number without a unit is a meaningless number, and a number with a unit but no sample is also meaningless. The fourth layer is confidence: I clearly distinguish what I know, what I infer, and what I assume.

These four layers are not administrative ritual. They are tools of the trade. When I analyse a match, I need to know which layer I am standing on so I do not accidentally turn an assumption into a conclusion. Most serious analytical errors I have witnessed in thirty-seven years come from mixing these layers.

Invisible is the largest data layer we have not yet measured. I write this word alone on a line in my notebook, and I keep it there as a reminder. Everything we measure about a match is a tiny part of what happened. Players move in the space we record, but they decide in another space for which we have no camera. Our precision is precision over a portion of the picture. Admitting that does not weaken analysis. It makes analysis honest.

Based on my experience watching matches across eight World Cups and eight Olympic Games, I have noticed a law about how the public receives analysis. When a piece offers a firm, simple conclusion, it spreads fast. When a piece offers a conditional conclusion with limits, it spreads slowly. Spread rate is inversely proportional to accuracy. This is a property of the information ecosystem, not of football, and it cannot be fixed by individual goodwill.

What to verify in the next match

A major-tournament cycle compresses emotion and expands temptation. When everyone is swept up in flags and stories, the pressure to produce content rises and the pressure to verify falls. Pieces written in this period have the shortest lifespan and the greatest potential for harm, because they are read the most.

In the next match, when you read an analysis with a firm conclusion, ask three questions. Where did the numbers come from and over what sample. Does the conclusion distinguish between what was observed and what was inferred. And if the author's hypothesis is wrong, what sign on the pitch would show it.

Fate is not decided in the press room — but it begins to be written there. And in an age when every press room has a machine writing in its place, the analyst's greatest discipline is to keep the blank columns and to say clearly that they are blank.