The Empty Data Sheet: The Discipline of a Football Analyst

2:47 in the morning, Beijing. Outside the window, the traffic on the Third Ring Road has thinned to the point where I can hear the ventilation fans on the building opposite. I open the file the desk has just sent, scroll down, and stop at the first field.

Team name: N/A.

Competition name: N/A.

Squad data: N/A.

Timestamp: N/A.

Twelve pages. Nine analytical dimensions. Not one line of factual data.

If I were the man I was fifteen years ago, I would close the file, pour another coffee, and write from instinct. I know exactly what that feeling is: fingers already on the keyboard, a conclusion already formed at the back of my head, and the data table reduced to decoration for that conclusion. People call it professional instinct. I used to think so.

On the wall opposite my desk hangs a sheet of A2 paper on which I hand-drew 27 geometric symbols representing pressing patterns. I built that system in 2026. Looking at it now, I see a neat paradox: 27 symbols, and nothing for them to mark.

My niece, a second-year media student, poked her head in while I was staring at the screen. She asked: "What will you write now?"

I answered her. That answer is this article.

In March 2026 I joined the sports department of Belgrade Television. My equipment was a notebook, a pencil, and a VHS camera on the gantry. No Expected Goals. No passing maps. Only one rule my first editor gave me on my first afternoon, which I have kept intact for 35 years: if you did not see it, you do not write it.

In 2026 I began building the geometric symbol system. My first article on the Chinese Super League received 312 reads and five comments. Nobody was angry, nobody was impressed. A perfect silence. I spent three months rewatching 80 Shanghai SIPG matches and found one concrete detail: the zone between their midfield and defensive lines was the origin point of seven goals conceded in the 2026 season. From that anchor I drew 27 separate pressing patterns. Each one is a spatial configuration, not a name.

In 2026, when the leagues were suspended and the stadiums empty, I did not watch a single live match for eight months. I built a database of 1,200 attacking patterns covering World Cup 2026 through the 2026-20 season, and processed it in Python. The most notable result: teams that press aggressively within 30 seconds of losing the ball recover it at a rate 23% higher than teams that press more slowly. I wrote a 15-page report, the longest piece of my then-20-year career.

My workflow reduces to five steps: gather sources, deconstruct the events into information points, map them onto nine analytical dimensions, cross-check, then publish. Tonight's file died at step two. Not from laziness. Because there was nothing to deconstruct.

Three situations must be distinguished, because each demands different handling.

First: the data package is withheld for commercial or rights reasons. This is the easy case. The data exists; it simply has not reached me. You wait, you buy, or you find an independent substitute source.

Second: the data exists but is unverified. This is the dangerous case, because it looks exactly like real data. A transfer rumour from an anonymous account. A figure that has passed through seven layers of resharing. A statistic sliced out of its original season.

Third: the package is genuinely empty, as tonight. No data because the event has not happened, has not been recorded, or does not exist.

What many people in this trade refuse to accept is that these three cannot be handled the same way. And the third is the only case in which the correct answer is an empty answer.

A usable information point needs four components: an entity, a timestamp, a unit of measurement, and context. Remove any one and it is merely a sentence.

"Germany have declined" is not information. No specific entity, no timestamp, no unit, no opponent context.

"At the 2026 World Cup, Germany's defensive line sat on average 62 metres from their own goal, and the centre-back pairing of Mats Hummels and Jerome Boateng won only 48% of their duels" is information. Entity, timestamp, unit, context.

The distance between those two sentences is my entire profession. Data does not lie, but it chooses whose voice it answers to.

I wrote about Germany before their match against South Korea in 2026. They lost 0-2 and were eliminated in the group stage for the first time in 80 years, since 2026. That article reached 870,000 reads. What I remember is not the read count.

What I remember is that before publishing, I deleted two paragraphs. One was about psychological saturation after the 2026 title. One was about domestic media pressure. Both read smoothly. Both were plausibly true. Both had zero supporting data. I deleted both.

That was one of the best decisions I have made as a writer, and nobody knows about it. That is the nature of this kind of decision.

The nine-dimension mapping step is the most time-consuming part of my workflow, and it is also the step that most clearly exposes where an empty package fails. Each dimension requires its own type of entity.

Tactical analysis needs lineups, starting shapes, positional data and block height across match phases. Finance needs contract values, wage structures, revenue streams, net debt. Results need specific matches and form sequences with a defined sample length. League landscape needs tables, squad values, and resource comparisons within direct competitive tiers. Governance needs rule texts and disciplinary precedent. Management and dressing-room analysis needs personnel, contracts and internal relationships. Risk profiling needs an event to assess. Media-narrative analysis needs a story that actually exists. Industry transmission needs an upstream originating event.

An empty package means none of the nine dimensions has raw material. And a nine-dimension system without raw material is not a poor system. It is an empty table. The table is still good. It simply has nothing on it.

Now the part I want to dwell on, because this is where people fall.

Suppose someone asks me to predict a match for which I have no data. In that moment, my head already contains a probability. If I write "Team A have a 55 percent chance of winning", where did 55 come from?

Nowhere. It is a number I invented. And it is far more dangerous than saying "I do not know", because it wears the shape of knowledge. A reader cannot distinguish a rate computed from 1,200 patterns from a rate conjured in three seconds. Both print in the same typeface.

To produce the 23% figure I published in 2026, I needed 1,200 patterns. Those patterns came from roughly fifty or more fully coded matches. A single high-level football match yields on average twenty to thirty codeable attacking patterns. Which means that for a conclusion with directional value, I need at least several dozen matches.

A missing match cannot be replaced by an afternoon of thinking.

When the sample size is zero, every rate is meaningless. Not imprecise. Meaningless. No denominator, no fraction. This is the simplest arithmetic in the trade, and the most frequently violated.

There are three ways people fill empty cells. I have seen all three, and I have done all three myself.

The first filler is reputation. "This team has always defended well." Reputation is a prior, not evidence. A prior is useful for deciding which question to ask, not for writing the conclusion. Morocco at the 2026 World Cup is the cleanest example. Relying on reputation alone, one writes "Morocco defend deep and counter" and stops there. Based on my own match-watching across that tournament, I coded 14 of their matches, and what I found was a different mechanism entirely.

The second filler is media narrative momentum. When three major outlets publish that a team is rising, what is rising is usually the story, not the team. Narratives carry inertia, and inertia is easily mistaken for on-pitch trend.

The third filler, and the hardest to detect, is a model run on bad inputs. You feed a good model a corrupted dataset, and the output is a number that looks valid. The model does not notice that the input is nonsense. It simply computes.

These are not three moral failures. They are three structural failures, and they occur most readily under deadline pressure. The mechanism closely mirrors what I analyse in football itself. A system never collapses starting from the final defeat. It starts at a break point much earlier, usually at a decision that looked small and entirely reasonable when it was made. A rate invented at 2:47 in the morning is one such break point.

One more distinction must be stated clearly, because it is constantly muddled: negative data versus empty data.

Negative data is information. "Team A created no clear chance across the entire second half" is a fact with an entity, a timestamp, a unit and context. It says something about Team A.

Empty data says nothing. "No match report exists for Team A" says nothing about Team A. It says something about the person holding the report.

Confusing these two is the origin of most unsupported conclusions I read each week. A match that was not recorded is not a match without events. A transfer with no leaked rumour is not a transfer that is not happening.

The cross-check step in my workflow follows one simple rule: a fact is publishable only with at least two independent sources, or one source plus match video I have watched myself. No exception for friendly sources. No exception for fast sources. No exception for sources that are usually right.

That rule costs me a lot of exclusives. It has also meant I have not had to issue a correction in ten years.

Back to Morocco and Achraf Hakimi, because this is the fullest example of what verified data actually looks like.

Morocco at Qatar 2026 became the first African national team to reach a World Cup semi-final. Across the 14 matches I tracked, one detail went almost unnoticed early on: Hakimi repeatedly vacated the right-back position and drifted inside, forming a five-man midfield line during defensive phases. That removed the opponent's central reference point and with it the ability to assign man-marking duties. My analysis video reached 1.2 million views on a platform in Beijing, and the national broadcaster kept me in a guest seat throughout the tournament.

But the number I want to tell is not 1.2 million views. The number I want to tell is the ninth match.

I found that structure in the ninth match. For eight matches I looked at it without seeing it. My eye needed eight matches to code a repeated movement into a named symbol. The pattern was there from the first match. The observer was the late arrival.

This is the deepest reason I cannot fill an empty cell with instinct. My instinct, even after 35 years and 1,200 patterns, still needs eight matches to see something sitting directly in front of it. An analyst confident he sees the truth in three seconds is an analyst who has never counted the times he looked straight past it.

I know this sounds like humility. It is not. It is arithmetic.

There is another temptation, subtler than the three above: turning emptiness into an argument. People write "the silence from the coaching staff suggests..." or "the absence of information on this deal is a signal...". Those lines sound sharp. They convert the absence of data into a category of data. Sometimes they are right. Most of the time they are simply a polite form of invention.

There is a truth about this trade I learned late: a club dies before the match kicks off, at the negotiating table and on the transfer paperwork. But that holds only when there is paperwork to read. With no paperwork, it is not an insight. It is a good-sounding sentence.

So what should an analyst write when the package is empty? My answer fits in one line: write the state, not the conclusion.

Writing the state means describing precisely what exists and what is missing, with reasons. "No positional data is available for Team A across their last three matches; therefore it is not possible to assess whether their block is rising or dropping." That is an honest sentence, and it is useful to the reader.

Writing a conclusion without data means inventing the ending and then hunting for evidence. That is a reversed process, and it is the default process of most sports content today.

Tonight I chose the first path. I returned an empty sheet to the desk, with a short note explaining why it was empty. That was the entire output I could honestly produce at 2:47 in the morning.

Over 35 years I have built a system, and one of the greatest risks of any system is that it persuades its owner that everything has a systemic cause. That belief is wrong. There are moments my model cannot explain, and I have to say so rather than force them into an existing symbol.

A missed penalty in the 88th minute is largely not the output of a pressing system. It is a missed shot. A red card in the third minute is not the output of a tactical structure. It is a late tackle. If I assign a systemic cause to those events, I am doing exactly what I condemned above: filling an empty cell with a model that sounds plausible.

The counterintuitive conclusion I want to leave here is this. An empty data sheet is the most honest document in the room. Every other document in that room contains somebody's decision, a decision about which figures to include, which season to cut, which phenomenon to name. The empty sheet contains a single truth: we do not yet know.

And the danger is not deliberate fabrication. It is fluent fabrication. An unsupported conclusion written well spreads faster than a supported silence. That is why honesty in analysis is not a virtue. It is a skill that must be trained, and it is harder than writing.

I hear the line "football is just luck" more than any other. I understand why people say it. It is a form of self-defence. But when I hear it in a discussion about a match where both sides have data, I always ask myself: if everything is luck, what exactly were my 1,200 patterns for?

The answer I found after eight months of building the database is not that football contains no luck. The answer is that luck has structure, and that structure is measurable in a small but stable proportion. I do not believe in luck. I believe in the 23% that shows up a second time.

If tonight's data sheet is empty, then what I must write is an empty sheet. There is nothing to analyse. There is one thing to record.

Read a data table the way you read a battlefield map: the smallest detail is an arrow. But a blank map is not a bad map. It is a blank map, and the reader has the right to know that before drawing their own arrows onto it.

In this major-tournament season, while everyone is swept along by flags and stories, the question I leave behind is simple. If you strip all the data out of the analysis you just read, what percentage remains is what you actually saw, and what percentage is what you filled in yourself?

A sporting culture does not live in the stands; it lives in how people defend the shirt. And in this trade, defending your own colours means refusing to invent a single percentage when nobody has yet stepped up to defend it.