The Perfect Match Analysis, Written From a Blank Page

Three in the morning in Chicago. Wind off Lake Michigan pushing through the gap in the window, a sound eleven years in this city have not made familiar. I opened the file that had arrived from our newsroom's internal analysis system. Nine sections. Tables lined up neatly. A column for “Assessment,” a column for “Risk,” a column for “Conclusion.” Subheadings in bold, numbered one through nine, each with an “Evidence” block and a “Hidden Insights” block. The sort of document that, printed out and set on a meeting table, nobody would dare question.

Then I read it properly.

No team. No player. Not one possession, one scoreline, one season, one contract, one line of head-to-head history. In every position where data should have been, the same sentence repeated like a prayer: “N/A — insufficient information.” And still the frame stood. Nine sections intact. A risk matrix with six neatly classified risk categories. A glossary at the end carefully defining technical terms the document itself never used.

The machine had finished building a nine-storey house on an empty lot, and not one floor had anyone in it.

On the pixel screen, I listen for the heartbeat of the pitch. This time the heartbeat was not real.

The machine learns to tell stories

Over the past seven years, the way a sports newsroom produces content has changed so much that people inside the industry struggle to recognise it. In many places, an analysis piece is no longer written the way a writer sits down, rewatches the tape, takes notes, and types. It is split into two stages.

The first stage is deconstruction. A source document — an article, a dispatch, a match report — enters the system and is dissected into structured fields: title, source, list of information points, dominant viewpoint, entities mentioned, time sensitivity, source quality. The second stage is deep analysis, where a model trained on the language of the trade takes those fields and builds a complete document: tactics, player data, team operations and salary cap, league landscape, rules, locker-room relations, risk, media narrative, and the ripple effects across an entire industry.

It sounds reasonable. Until the first stage returns a blank page.

That night, the source document did not exist. Empty title. Empty source. Empty list of information points. And instead of raising an error, the system still emitted a document — except that where the data should have been, it printed the very instructions someone had written for it. One field said to “identify entities from the list of information points above.” Another said to “assess source quality from the source fields of the information points.” Those lines were commands, not results.

The industry calls this a scaffold leak. The template is printed, but the content poured into it has vanished. And because the template still looks good, still has a title, still has a table of contents, it slips past every automated check without anyone noticing.

What is notable is that the second stage did not fabricate anything. It did exactly what a good data analyst would do: recognised there was nothing to analyse and said so on every line. That nine-part document, in the end, was an honest report on emptiness.

Except that it looks far too much like a finished piece of work.

The trap of a document that looks done

This is where I want to linger, because it bears directly on my daily work. Based on my years of watching matches and working with statistical systems, I have come to see that the biggest risk in sports has never come from wrong numbers. It comes from right numbers placed inside a wrong frame.

An example. The heat map has been a favourite tool of broadcasters and data sites for a decade. Viewers look at it, see a bright red patch around the penalty area, and conclude that the player was everywhere. But a heat map tells you where he was, not what he did. A midfielder who stands exactly where the defensive system asks him to stand, who runs the precise channel to drag an opposing full-back out of position and open space for a teammate to shoot, may leave only a faint smudge on the heat map — while the man who gets the credit is the one who scored. The heat map builds a story with a shape but no reason.

That is precisely what happened to the nine-part document. It had a shape: nine sections, complete subheadings, three conclusions per section. It had no reason. And what makes it dangerous is this: if you only read the title, the structure, the number of sections, it is perfect.

In sports data analytics there is a harsh phrase for this kind of output: a perfectly confident hallucination. Fluent prose, immaculate formatting, smooth sentences, generated from an empty input. To an ordinary reader it is indistinguishable from a real analysis. To an automated aggregation system, even less so.

In the end, it is a new kind of fortune-telling. People once read tea leaves, cards, turtle shells. Now they read charts. The difference is that the old fortune-teller knew he was telling fortunes, while the nine-part template knows nothing at all.

There is one small detail I find more telling than the rest: the document carried a long glossary at the end — offensive rating, defensive rating, true shooting percentage, the various salary-cap thresholds, the exceptions in the collective bargaining agreement. All carefully explained. And right beside them, the document noted plainly that these terms had been deliberately excluded from the analysis, because none of them had any grounding in the data received.

Whoever wrote that note was being scrupulously honest.

The cost of an industry that always wants an answer

It would be easy, and comfortable, to turn this into a lesson about technology. About immature algorithms, dirty data, pipelines that need patching. I do not think the root lies there.

The root lies in the fact that we have built newsrooms where a wrong answer, beautifully presented, is always at an advantage over an honest silence. Editors are squeezed on volume. Aggregation systems chase output. Analysts are graded on formatting. In a machine like that, a nine-part document full of tables always gets through more easily than a single short line: “I do not have enough information to answer.”

The paradox is that what readers actually want is the opposite. Watch a match and you will see it: the moment that brings a stadium to its feet is never the moment the numbers line up. It is the moment a player — possibly one nobody remembers the name of — touches the ball in a way nobody anticipated. Where the ball rolls, we begin to tell stories. Not where the spreadsheet rolls.

I will never forget an evening in September 2026 at Toyota Park. Chicago Fire against Toronto FC. In the 93rd minute the young forward David Accam, number 11, curled a shot in to make it 2-2 in front of twenty-one thousand spectators. I was twenty-four, freelancing for a local football blog called Windy City Football, and I did exactly what a young freelancer should do: I wrote a tidy match report, correct in every particular, with the score, the timeline, the minutes.

Then I deleted all of it.

Instead I wrote eight hundred words about the heartbeat of a city inside a single touch of the ball. The club's own homepage shared it. The editors that night told me something I have carried through my whole career: they did not remember the score, they remembered the feeling.

A year later, I got lost in Moscow and found a heart. World Cup 2026. Belgium against Tunisia. After the final whistle I was stranded in the crowd because I had lingered to interview a seventy-two-year-old Senegalese supporter named Ousmane. He told me he had followed his national team through five World Cups and had never once seen them win an opening match. He clutched a shirt worn thin at the shoulders and stood in a river of singing Russians, a monument to irrational loyalty. The piece I wrote about him was shared more than twelve thousand times.

There was not a single table of data in it.

In the summer of 2026 every league stopped. Stadiums stood empty, the stands silent as an abandoned church. I was twenty-seven, chief editor at an independent sports site in Chicago, and the fear of losing my job was larger than I cared to admit. The summer was empty, and the pitch kept whispering. I built a project called “Memories of the Pitch”: I called forty-seven supporters in three countries — England, Brazil, Vietnam — and asked them about the Euro 2026 final they had watched as children. The series ran four weeks, averaged eight thousand five hundred words an instalment, and readership rose three hundred and forty per cent.

In 2026, in Doha, while colleagues jostled for Lionel Messi and Kylian Mbappé, I spent two days with Lucas Torreira, the Uruguayan midfielder who did not play a single minute across all three group-stage matches. The piece ran the day Uruguay were eliminated, drew only two thousand three hundred reads, and was shared by six international journalists and cited by a sports scholar in an academic journal.

I tell these stories not to argue that machines cannot do anything. We use data systems every day, and I owe them. I tell them to make one simple point: the value of a piece of sports writing comes from being anchored to a moment that actually happened. When the moment disappears, what remains is decoration.

And our industry is producing a great deal of decoration.

What a decent analytics room would do

If that nine-part document had been produced inside a professional data analytics department — the kind big clubs staff with a dozen people — it would never have been published. The correct reflex of a specialist handed an empty input is simple: stop, raise the alarm, hand the problem back to whoever owns the data pipeline.

That stopping is discipline, not timidity. In sports analytics, a model drawing conclusions without supporting data is treated as the most serious kind of error — because the conclusion cannot be verified, rather than because it is certainly wrong. A number you cannot trace has no value in a meeting room, even if it happens to match the coach's intuition.

In the case of that night's document, the problem sat in three places, and I list them here because they recur across a great many content systems.

The source title and the source did not exist. Without a title there are no entities. Without entities there is no analysis. Every layer behind collapsed in turn, like a row of dominoes set too close together.

The system contained a logical flaw inside its own specification. It instructed the analyst to “identify the entities mentioned from the list of information points above” and to “assess source quality from the source fields of the information points.” But the list of information points was empty, and the source fields were empty too. The specification demanded something it could not supply the data to do.

And the point I want to stress most: because the template rendered successfully, an automated consumer could easily mistake this document for a completed analysis and route it into aggregation or editorial workflows. Nobody did anything wrong on purpose. Nobody had been designed to catch it.

The blind spot nobody wants to look at

There is another reading, more counter-intuitive, and I think it comes closer to the truth.

We usually ask whether machines can replace sports journalists. That nine-part document did not ask that. It asked something far more uncomfortable: do we still have the capacity to tell the difference between an analysis and a simulation of the shape of an analysis?

The weakness of human beings lies not in reading words. It lies in being fooled by form. A document with nine sections feels more trustworthy than one with three. A piece with data tables feels more solid than one that only has storytelling. A coach who says “we need to look at the analysis further” sounds more professional than a coach who says “I don't know.”

The transfer market runs on that faith in form. Big clubs race each other like a brand arms race: buy a star to sell shirts, to fill the stands, to keep the sponsors. But the real value of those deals usually sits somewhere nobody looks: a full-back at a small club bought cheaply, slotted into the right tactical system, becoming the piece that makes the whole team work. Every contract is a sentence left unsaid. Most of those unsaid sentences are not attached to the loudest names on the bulletin.

It is the same in sports content. Big newsrooms pour money into automation, into volume, into speed. The pieces with real value — the one that made a seventy-two-year-old Senegalese supporter feel seen, the one that made a substitute midfielder in Qatar worth naming after the tournament ended — tend to be made in smaller, slower places, where a human being is still sitting there after the final whistle.

That is why I do not worry about machines writing journalism. I worry that we will forget how to check.

The disaster is not a nine-part document full of “insufficient information.” The disaster is if tomorrow the same system, with the same empty input, emits a nine-part document packed with numbers that look entirely real. Then there will be no “N/A” column to warn anyone. And we will print it, put it on the meeting table, and nod, because it looks far too professional to be wrong.

The old man in Moscow told me his story, and all I could do was write it down. That night in Chicago was the same: all I could do was write down that the system had told the truth. The question is not whether it told the truth. The question is that it told the truth in a tone of voice that made it impossible to believe it was telling the truth.

The most perfect analysis I have ever read was a blank one. And if I had to choose between a system that always answers and a system that knows how to say “I do not have enough information,” I would choose the second, every time, without hesitating.

What remains for the rest of us: does your newsroom have a place to record the sentence “I don't know” where nobody gets docked pay for it?