The Blank Spreadsheet in Pudong and Ninety Minutes That Cannot Be Invented

Twelve Empty Rows

Tuesday morning, 7:40, the training base in Pudong. Mist still clung in streaks along the touchline; the grass was wet enough that every stride from a full-back sounded like tearing cloth. I stood in the corner of the pitch, about forty metres from the goal, holding a folded A4 sheet. Twelve ruled rows, one handwritten heading: “Left foot — 35 metres.” Twelve empty rows.

Through the whole first session I wrote nothing. I just watched a number 7 stay behind after his team-mates had gone to the dressing room, place the ball on the same spot, and shoot. Those twelve empty rows were the entire record I carried home that morning.

Years later, I sat in front of a data file pulled from a match-tracking system. The header row was complete: competition, both teams, every player, columns for minutes, passes, duels, metres run. Beneath the header row, the body of the file was entirely empty. Formally valid, substantively meaningless.

The feeling was identical to that Pudong morning. And so was the temptation.

A major-tournament cycle and the pressure to have something to say

We are in the middle of a major-tournament cycle. Every morning my phone buzzes without pause: projected line-ups, unverified injury stories, fifteen-second clips cut from a closed session, hypothetical group tables assembled by someone in a spreadsheet. Reader demand in this phase is obvious: people want to know what will happen, who starts, who scores, who advances. Nobody is waiting for a reporter to say he does not yet know.

My career began in 2026, on the sports desk of a television station in Belgrade. That is where I learned something that sounds trivial: if you do not yet have the material, the bulletin still goes out, but whatever you do not have must be labelled as missing. My editor called it a “named gap.” A gap with a name lets the audience know where to wait for more. A gap filled with conjecture cheats the audience without them noticing.

From 2026 I was living in Shanghai and started going to the local club's training ground every Tuesday morning as a freelance contributor. My job then was simple enough that it was hard to explain to outsiders: stand in one corner of the pitch and count. Count how many times a player shot, how many times he switched feet, how many times the ball went in, how many minutes he lingered after the session ended. Count, and record it in pencil, because ink smears in the rain.

People often ask why I do not simply use the data that already exists. The answer sits in that Pudong morning: no existing dataset has a column for “player stays behind after his team-mates have gone to the dressing room.” It only has a column for outcome. And outcome, divorced from context, is one of the most misleading forms of information in football.

People and numbers

Wu Lei in 2026 was a right winger by trade, but he spent most of his extra time on his left foot. Over ten consecutive sessions I counted forty-two shots from a zone roughly thirty-five metres from goal, and seven of them went in. Seven out of forty-two. A conversion rate under twenty per cent — a figure that, read on a summary sheet alone, would lead anyone to conclude this was a skill not worth investing in.

But my sheet had another column, one no system records: where the substitute goalkeeper was standing inside the goal. On the first seventeen shots, he stood high, and the ball usually passed over him. From the eighteenth shot onwards, he dropped back about two metres. All seven goals came after that adjustment, when Wu Lei began switching the drop point to the far corner rather than the near one.

What I wrote in the piece “The 35-Metre Dead Angle” was not “Wu Lei is good at shooting from distance.” What I wrote was this: the player had spent four sessions doing nothing but recalibrating the reflexes of one man, and that man was the backup goalkeeper. The piece drew around fifteen hundred comments and moved me from freelance contributor to a formal internship. But the thing I kept was not the article. It was the field notebook, and one rule: data only has value when I know how it was measured, where, and by whom.

At twenty-two I counted every metre of a ritual; at thirty-one I understood it was a promise.

Table first, tape second

My professional habit is to open the data table before opening the video. Not because I trust the table more than my own eyes, but because the table tells me what to look for on the tape. A team with sixty per cent possession but only five shots in the first half has a problem converting territory into threat, and I need to re-watch its build-up on the left flank. Conversely, a team with thirty-five per cent possession but eight shots inside the box is an organised counter-attacking side, and I need to re-watch its transitions in the first seven seconds after winning the ball.

In both cases, the table does not conclude for me. It simply narrows the area I need to inspect.

A few seasons ago I spent nearly three weeks re-watching the closing twenty minutes of an upper-mid-table side chasing European qualification. What the table showed was straightforward: from the seventieth minute onward the team's high-press actions dropped by roughly a third, while the number of times opponents played through its midfield almost doubled. On tape the reason was visible. The two central midfielders could no longer cover the ground they covered in the first hour, and the left channel became an open door.

That observation required no special tooling. It required three weeks and one notebook.

Five substitutions and the twenty minutes that became an attrition war

Five-substitution rules arrived with a promise of durability for deep squads. In theory that is right: a club with twenty-five players of broadly comparable quality can rotate an entire midfield without much drop-off. But having watched enough matches across enough leagues, I see a different and less discussed consequence. Five substitutions does not make the final twenty minutes easier to breathe through; it makes them tighter. When both sides still have three changes available and can each throw a fresh striker on at minute seventy-five, the closing phase becomes a calculated war of attrition, in which the side with the better bench resources wins — not the side that played better football in the preceding sixty minutes.

The knock-on effect is that the way we read a match has to change too. When I watch a game and see Team A take the lead in the thirtieth minute, I no longer judge them by whether they hold on. I judge them by how many difference-making cards their coach still has at minute sixty-five, and how many the opponent has. A leading team with only one attacking substitution left must choose between protecting the score and keeping the ball. Quite often they choose both, and lose both.

This also raises a question about how expected-metrics tables are built. Most models aggregate chances across the whole match with equal weight. But a chance in the fifteenth minute and a chance in the eighty-fifth are not structurally equivalent, because the shape that produced them is not equivalent. Late chances typically arise from physical imbalance, and that imbalance is predictable if you know who can still run at minute sixty.

The training ground is a stage with no audience, where understudies rehearse for the lead.

The Suzhou bubble, forty days and one balcony

In 2026 the Chinese top flight was played in a controlled bubble in Suzhou. I was one of the few reporters living with the Shanghai squad for the full forty days. No travel, no crowds, no life beyond the hotel and the training pitch. Time moved by the training schedule rather than the fixture list.

In the evenings, Oscar — the Brazilian number 8 — often stood alone on the balcony. He could not get back to São Paulo, could not see his two young children. One day he asked me to film a short video message for them. I shot it on a phone, unscripted, unarranged, just to let him speak. When it was posted it drew roughly three point eight million views. That number says nothing about the quality of the video. It only says that a great many people, in a great many places, were in the same position.

What stayed with me more was what happened afterwards. Trust in the dressing room increased. Chen Wei, the backup goalkeeper, a man who appeared in almost no bulletin all season, began talking to me about his greatest fear: being forgotten. Not being sold, not being injured. Just being forgotten.

From that I built the “Bubble Diary” series, told in the voices of the players who never got on the pitch. No goals, no scorelines, no tables. Only the people sitting on the bench and waiting.

Inside the Suzhou bubble, we scored with lonely shots.

Kazan, the night of 27 June

In 2026, on the back of that run of training-ground pieces, I was sent to Russia for the World Cup as a content assistant. I had no official accreditation. On the night of 27 June, at Kazan, Germany lost to South Korea without reply and went out in the group stage. I stood in the supporters' section, among people with no seat.

In the front row, about five metres from me, a man named Park Jae-won, fifty-four years old, knelt down and kissed the grass. He said one sentence to me, which I recorded verbatim in my notebook: “Sixteen years it has taken to get this night.” He did not talk about winning. He talked about the waiting.

The piece “Kazan, Where the Tears Belonged to the Strong” went on to draw around two point three million reads. I did not retell the scoreline in it, because hundreds of other reports had already done that.

Kazan taught me this: tears do not belong to the losers, they belong to those who stay.

A blank cell is not an all-clear

Back to the empty data file on my desk. There is something about blank cells that our trade routinely mishandles.

In a football dataset, a zero and a blank are completely different things. A zero means it was measured and the value was nil: this player attempted no shots in the second half, this team took no corners. A blank means it was not measured: we do not know whether that player shot, we do not know whether that team took corners. In professional reporting these two cases must be presented differently. In practice they are usually treated the same, and that is the most serious blind spot in football analysis today.

A report that found no risk and a report with insufficient data to look for risk are opposites in value. The first is a conclusion. The second is a gap. If a reader skims both and sees no red flags in either, they will process them identically. That is the moment a silent system becomes a lying system.

In football we meet a version of this every week. A team wins three-nil while taking six shots, and the bulletins call it a convincing victory. Another loses one-nil while controlling the entire game, and the bulletins call it a crisis. We are using the outcome cell — the only cell always filled in — to infer every other cell. But the outcome cell carries the largest variance of anything in the table, which makes it the least reliable basis for inference.

The second blind spot belongs directly to the major-tournament cycle. When a short tournament takes place, the sample is tiny: three group games, seven at most for a finalist. On a sample that small, randomness weighs so heavily that any tactical conclusion is fragile. A team winning three group games does not prove its system is better than the system of a team that won one and drew two. Yet we write as though it has been proven, because we need a story to file that day.

The third blind spot sits with the writer. Amid the roar, the beat keeper very easily forgets his own beat. When a favoured team wins, nobody re-checks the source. When a disliked team loses, nobody re-checks the spokesperson. A reporter who stands at a training ground every week has a large advantage over someone reading feeds in an office: he knows who trained today, who did not, who worked alone, who left early. That advantage is worth something only if he is willing to write it down, including when it contradicts the story currently going around.

The beat keeper is not permitted to fall asleep inside the roar.

What an empty dataset can teach

I still keep the A4 sheet from that Pudong morning in 2026, twelve empty rows, in my desk drawer. It has no further use. I keep it to remind myself that every piece begins with a gap, and that the gap is not a problem to be concealed. It is the starting point.

Handling an empty data file is like handling an empty ruled sheet. The first step is to check whether the file is corrupted, whether the download was correct, whether the logging system is failing. The second step, if the file really is empty, is to state plainly that the file is empty. The third step is to find another source — and in my trade, that other source is usually the grass.

There is a line I wrote in my field notebook long ago: the pitch never lies, only the watcher tells the truth. The grass does not care whether you have a dataset. It only records footprints. An injured player can still run, but his footprints differ: shorter stride, uneven push-off, and in the sprints he chooses to arc wide instead of driving straight. None of that lives in any cell of a standard dataset. It lives on the pitch, and only there.

The next signal

Next week I will go to the training ground again on Tuesday morning. I will carry a new sheet, ruled in advance, and I will begin by writing nothing at all. If after the first session the sheet is still blank, I will not fill it with conjecture. I will come back the following Tuesday.

Football is a sport of small samples and large gaps. Everyone wants to read the conclusion. But a conclusion is trustworthy only when behind it stands someone who stayed long enough in one corner of a pitch, counted enough times, and was willing to admit what he had not yet counted.

If an empty data file lands in your hands this week, will you fill it with a story, or will you stand up and go and look?

I do not write to be read. I write so that the pitch has a witness.