When the Data Sheet Returns Zero

The clock on the wall of a small apartment in Shenzhen read 2:14 a.m. In front of me sat a spreadsheet with nine columns and a single row of results that was completely empty. I had spent four hours collecting data, run the model three times, and each time the system returned the same sentence: insufficient information to reach a conclusion. I first felt that sensation in the summer of 2026, just turned eighteen, computing expected goals for the France–Belgium World Cup semi-final. My model gave France 1.6 and Belgium 0.8. France won 1-0 with a header from a corner. The spreadsheet was not wrong. It simply went silent at exactly the point where I needed it to speak. Since then I have built a data-reading process with nine layers of checks: tournament context, format, squad, region, finance, rules, risk, media narrative, and industry transmission. Each layer has to answer a minimum question. If it cannot, that layer is marked empty rather than filled with guesswork. It sounds dry, but this is the hardest skill in the trade. Beginners fear gaps. They see an empty column and immediately fill it with a story: this team presses better, that player is out of form, that coach has lost the dressing room. Those stories sound plausible, and it is precisely their plausibility that makes them dangerous. In football, gaps are everywhere. Broadcast footage captures only part of the pitch. Positional data only records players while the ball is moving. The feeling of a defender being dragged out of position in the 88th minute lives in no data file. The good analyst is not the one who fills gaps fastest, but the one who points at the right place to say he does not know. Conclusions drawn from the office are often detached from the rhythm on the pitch, and that distance only shows itself when you are willing to stand still and look at it. The France–Belgium semi-final on 10 July 2026 in Saint Petersburg was the first lesson. I collected the coordinates of every shot from public statistics pages, weighting by distance, angle and situation type. The result: France 1.6, Belgium 0.8. The actual score: France 1-0, Samuel Umtiti heading in from a corner in the 51st minute. The problem was that my model weighted set pieces lower than reality warrants. A corner is not a random sequence of events; it is a routine rehearsed hundreds of times on the training ground, with blocks, decoy runners and a calculated drop point. For a month I rewatched the entire tournament, logged every set piece, and adjusted the weights. The model became more accurate. But I learned something more important: expected goals measures the quality of chances, not the quality of organisation. In 2026, when the pandemic turned stadiums into empty concrete shells, I was a data analysis intern at a sports company in Shenzhen. I collected figures from 240 Chinese Super League matches. Two facts stood out. The home win rate fell from 47 percent to 39 percent. And PPDA, the number of passes a team allows the opponent per defensive action, dropped from 11.2 to 10.5. In other words, teams pressed harder and contested more, yet scored less efficiently. The easiest explanation is psychological. No crowd, no pressure, and the home side loses its anchor. That explanation may be partly right, but it ignores what I call the background hum. A stadium with a crowd generates a rhythm: noise rises when the home team pushes forward, silence falls when the ball goes out to the flank, a collective exhale follows a clearance. That rhythm shapes player decisions below the level of conscious thought. When it disappears, players must generate their own rhythm, and most have never been trained to do so. Crowd or no crowd, the match still needs someone to tell its story — but the storyteller must know what he is leaving out. My internal report was published on the company news site, and a few local analysts cited it. What I remember most, though, is a line from an assistant coach: he did not need a data table to know this, he needed a data table to convince the board. That is another truth about my profession. In November 2026, I worked as a data assistant for an online sports outlet covering the World Cup in Qatar. On 22 November, Saudi Arabia beat Argentina 2-1. I calculated the winner's expected goals: 0.35. Argentina: 1.9. The article drew heavy criticism. Many readers felt that publishing such a low figure insulted the underdog's victory. I did not take it down. I wrote a second piece using tracking and positional data to show that Argentina controlled the ball but left gaps in two decisive phases: the first when the back line pushed too high right after taking the lead, the second when the midfield failed to drop in time after a long ball. Saleh Al-Shehri and Salem Al-Dawsari needed only two moments. What I defended was not the 0.35 figure itself. What I defended was the right to say that a match can be won by two moments and still be dominated for the other 88 minutes. Both things are true. A European football magazine noticed the second article and invited me to contribute as an independent data specialist. In June 2026, at the Euros, I followed the Georgia national team for two weeks. It was their first appearance at a major finals. Qualifying data showed their average expected goals against was 0.9 per match, among the lowest, even though they did not dominate possession. I wrote that Georgia could surprise Portugal. On 26 June 2026 they won 2-0, with goals from Khvicha Kvaratskhelia and Georges Mikautadze. This time the model was right. But I did not celebrate. A model that is right once proves nothing. What mattered was the structure: Georgia won by accepting they would concede the ball, holding their defensive distances, and waiting for two transition moments. That structure is predictable. The emotion in the dressing room afterwards is not. Correlation is not causation, and this is where my trade slips. The fact that the home win rate fell from 47 to 39 percent in a season without crowds does not prove that crowds were the cause. 2026 also brought a compressed schedule, centralised venues, quarantine, and changes to substitution rules. Any of those could have contributed, and no model separates them cleanly. I once told a colleague that expected goals does not lie. It simply never tells the whole truth. My colleague replied: then it lies politely. I think he is half right. The problem is not the tool, but users turning the tool into a courtroom. I do not build tables for matches; I build tables for doubt. There is a reverse risk few mention: when an analytical sheet returns an empty result, our reflex is to fill it with narrative. A coach gets sacked, a player gets blamed, a tactic is declared obsolete. Those conclusions may be right. They simply were not drawn from a data file that never contained anything. Football does not live inside the cells, it lives between them. That is why I still rewatch footage after the model has finished, even though it takes three times as long. My nine-column spreadsheet was still empty at dawn. I saved the file as null-result, then wrote a short note: source unreadable, re-run the extraction, no inference permitted. That was the entire output of a night's work, and it was still an output. Next season will bring thousands more columns and thousands more brilliantly named metrics. What I want to keep is not the ability to compute faster, but the ability to know when to stop my hands on the keyboard. If a metric cannot measure the moment you are trying to explain, does it belong in your article at all?