The Empty Spreadsheet: Notes on What Sports Data Never Says

The Empty Spreadsheet and the A4 Sheet in Hai Phong

In June 2026, in a press room in Hai Phong, I placed on the table a single A4 sheet printed with the statistics of 14 matches played by Rimario Gordon. The Jamaican striker had just been signed by the home club for a transfer fee of 250,000 US dollars, no small figure for a V.League side at the time. Across those 14 matches, his expected goals figure stood at 0.32 per game, the lowest among ten foreign forwards then playing in the league. I circled the bottom row in red and wrote my forecast in the margin: five goals this season.

A senior editor in the room said something I still remember verbatim: "What would a woman know about strikers." I did not argue. I simply pushed the sheet toward him, my finger resting on the source column: 1,105 minutes played, 68 shots, 22 touches inside the opposition penalty area, 4 shots blocked from inside twelve metres. The empty cells on that sheet mattered just as much: no column for top speed, no column for successful escapes from marking, simply because the league did not collect them.

At the end of the season, Rimario scored exactly five goals. His contract was terminated. Nobody said anything further in that press room.

But what I want to write about today is not a won argument. It is the empty cells.

That night in Hai Phong taught me one thing: people look at the price board, I look at the movement board. And a movement board, if you want to read it, requires knowing where it falls silent.

A Profession Built on Empty Cells

I work as a transfer market administrator for a sports news outlet. The job sounds glamorous: tracking players, valuing them, forecasting. In practice, most of my hours go into cross-checking spreadsheets nobody bothers to read: minutes played, shot locations, duel win rates, turnovers in one's own third.

For a league like the V.League, a full European-standard dataset does not exist. There is no tracking data, no complete defensive pressure model for every round, and in some rounds not even accurate touch counts. Newcomers to the trade do something naive: they fill empty cells with feeling. They watch three matches, see a player running hard, and write "good stamina." An empty cell gets filled with an adjective.

I learned that an empty cell is a category of information, not a defect to be hidden. There are three kinds of gaps, and each tells a different story.

Collection gaps occur when a metric is perfectly measurable but the league does not measure it. A league with expected goals but no expected goals on target is a textbook case. The gap tells you the league is only halfway toward quantifying finishing quality.

Structural gaps occur when a metric cannot exist. You cannot compute the average number of passes an opponent completes before losing the ball for a league where nobody records passes. This is an infrastructure limit, not analyst laziness.

Deliberate gaps occur when data exists but is not published. A player is injured; the club says nothing. Salaries are never disclosed. The medical department holds training-load figures, and those figures sit in a drawer. This gap is the most dangerous because it is not neutral.

Those three categories determine how I write. With a collection gap, I state plainly: the league does not measure this, so I draw no conclusion. With a structural gap, I find a proxy metric and label it as a proxy. With a deliberate gap, I flag it in red and return later, once the information is dense enough to speak.

The chart does not lie, but it does not tell the whole story. I go looking for the part left blank.

Fourteen Matches, Sixty-Eight Shots, and One Column With No Number

Back to Rimario. The dataset I built for him had four layers. The first was volume: 14 matches, 1,105 minutes, an average of 79 minutes per appearance, enough to rule out the theory that he simply did not play enough to score. The second was chance quality: 68 shots, 31 of them from outside the box. The third was position: 22 touches inside the opposition area across 14 matches, or 1.57 per game, a very low figure for a centre-forward. The fourth was output: 3 goals in those 14 matches.

An expected goals figure of 0.32 per game can be misread if taken in isolation. People often say: "The number is low because his teammates never pass to him." That is a reasonable hypothesis, and I tested it. If teammates never pass, touches inside the box should be very low, and indeed they were 1.57 per game. But when I separated the 31 shots from outside the box, I found that 24 of them came from beyond twenty metres with no defender near him. An isolated striker has to shoot from distance, but not that kind of distance. That is the shooting of a man who no longer believes he can escape his marker.

This is where data begins speaking through silence. My sheet had no column for "number of times he won position ahead of a defender." The league does not collect it. But I had a proxy: the number of fouls opposition defenders had to commit on him. Across 14 matches, that figure was nine. An average of 0.64 per game. Among the other foreign forwards, the next lowest was 1.8 per game. He was not generating pressure on opposing defences.

The end-of-season total, five goals, matched the forecast, not because I performed magic. It was the result of reading one empty cell correctly and choosing a sensible proxy. Looking only at 3 goals in 14 matches, I could have concluded "unlucky." Looking only at 68 shots, I could have concluded "positive." Both would have been conclusions read from half a map.

That episode shaped how I write every piece since. I do not open with a judgement. I open with the empty cell and state plainly what I am using to fill it.

The Tank and the Empty Cells of a Model

In June 2026, the newsroom assigned me a World Cup preview special for the tournament in Russia. I built a four-variable model for Germany: 67 percent average possession, 2.1 expected goals per game in qualifying, 91 percent passing accuracy, and chances created from set pieces. The model produced a semi-final finish. I ran the headline "The tank cannot stop in the group stage."

My model had four variables. It was missing at least five others, and I knew it, yet I wrote as though I did not.

The first missing variable was pitch temperature. In June in Russia, pitches dry faster, the ball rolls quicker, and a possession side needs more touches to control tempo. I had no per-stadium pitch temperature data, so I ignored it.

The second was Mexico's high press. In regional qualifying, Mexico had shown willingness to push high, but because they were not among Germany's major opponents, I underweighted them.

The third was the psychology of a reigning champion. This is the classic deliberate gap. Nobody can measure it, and because it cannot be measured, it gets defaulted to zero. I defaulted it to zero too.

The fourth was history. And this is the one I regret most, because the data was fully available. Before Germany in 2026, three reigning World Cup champions had been eliminated in the group stage of the following tournament: France in 2026, Italy in 2026, Spain in 2026. Three times in the last four editions. That ratio was sitting on my desk, and I excluded it from the model because it was not a technical metric.

On 17 June 2026, Germany lost 0-1 to Mexico. On 27 June 2026, Germany lost 0-2 to South Korea and were eliminated. My article was mocked by readers for a week.

Germany left the 2026 World Cup — every model has a day it goes bankrupt; only historical data remains. From that shock I learned a line I repeat before every piece: respect the model, never trust it absolutely. I abandoned declarative writing entirely. For every match, I now publish two scenarios, each with an explicitly stated uncertainty factor.

Twenty-Six Rounds With Crowds, Nine Without

In May 2026, with the world paralysed by COVID-19, the Bundesliga became the first major league to return with empty stadiums. I recognised a rare opportunity: a natural experiment in which the only removed variable was the crowd, while everything else — clubs, rules, calendar — stayed roughly constant.

I split the season into two parts: 26 rounds with crowds and 9 rounds without. I compared four metric groups.

First, home advantage. The home win rate fell from 55 percent to 43 percent. Measured as the gap between home win rate and away win rate, home advantage dropped by roughly 15.3 percent. This is the clearest quantitative evidence that crowd noise has real effect, not merely a feeling.

Second, discipline. Yellow cards rose 22 percent. The most plausible explanation: without crowds, referees lose the social pressure of a packed stand, and players lose the sense of being watched.

Third, pressing intensity. The average number of passes completed by the away team's opponent before losing the ball fell from 11.4 to 9.8. Away teams were pressing earlier and harder because noise no longer made them hesitate.

Fourth, goals per match, which barely moved.

I wrote a three-part series. A German tactical analyst shared it, and I gained 2,000 followers.

But what I learned was not in the 15.3 percent figure. It was in the variable I realised I had omitted for years. In an empty stadium, I realised I had failed to count one variable: emotion does not appear in a spreadsheet. Every model I had built for previous seasons assumed the crowd was a constant. It turned out to be a variable, and it moved from 55 to 43.

My writing changed from that point. I began using before-and-after comparison as the skeleton of every analysis. Not because it looks elegant, but because it forces me to identify which variable changed, rather than merely describing the present state.

Belgium Had the Best Attack, Italy Had the Best Defence

In July 2026, I predicted Belgium would win the European Championship, for a single reason: they had the highest total expected goals in the tournament. I wrote the piece, offered no second scenario, and felt confident.

Italy under Roberto Mancini won. Their average passes-allowed-before-recovery figure was 8.7, the lowest of the 24 competing teams. Opponents completed an average of only 8.7 passes before losing the ball. I had missed that metric because I was fixated on expected goals.

After the final, I spent three weeks building my own pressing dataset across 14 major leagues, reaching back to 2026. The result forced me to publish a public admission of error under the headline "I was wrong: data has nothing but the truth." The central finding: European champions from 2026 through 2026 all recorded a passes-allowed-before-recovery figure below 10.

What stands out is why I had not discovered this sooner, for a very specific reason: in the dataset I used, the pressing column was blank in seven of fourteen leagues. I had looked at that empty cell hundreds of times and defaulted to treating it as unimportant, rather than asking why it was empty.

Since then, every match analysis I write combines at least two data dimensions: attack and defence. And my headlines became humbler. I ask "Could…?" instead of "Certainly…".

The Three-Layer Verification Rule

After four major errors — the Rimario call was right, but the four that followed were wrong — I built a three-layer process for every piece.

Layer one is provenance. Every number must answer three questions: who recorded it, when, and how. A number that cannot answer all three does not enter the article.

Layer two is cross-verification. Every conclusion must be tested against at least two independent metrics. If the attacking metric and the defensive metric tell different stories, I do not pick one. I keep both and write out the contradiction.

Layer three is context. This is the hardest layer, because it requires me to state what I do not know. Every piece now includes a short passage specifying which factors the data excludes.

Layer three produced a concept I use constantly: the silence threshold. That is the point at which the number of empty cells exceeds the number of populated cells, making every conclusion an act of speculation. When I hit that threshold, I write one sentence: not enough information to conclude. That sentence once drew complaints from editors. It has also saved me from retracting pieces many times.

Three in the morning, the market is asleep. That is when the numbers are most awake. I usually recheck datasets around that hour, when nobody is calling to ask when the piece will be ready.

When Esports Taught This Lesson Before Football

I entered the industry in 2026, starting as an esports player and tournament organiser before moving into media. Esports taught me to read empty cells before I ever encountered expected goals.

In esports, every patch note is a shift across the entire ecosystem. Pick rates and ban rates change overnight. The most interesting part is this: empty data cells in esports tend to appear exactly where they matter most. A rising team posts a high win rate, but with a sample of only six matches the sheet is nearly meaningless. A player shows beautiful numbers, but if those numbers were accumulated in a lower division, they have no comparative value.

Football arrived at the same lesson later, and more expensively. A player with 14 matches of data is a small sample. A league with nine crowdless rounds is a small sample. A single European Championship is a small sample. My profession consists largely of working with small samples and stating clearly that they are small.

The only real difference between the two worlds: esports publishes its data, football does not. But publishing does not mean completeness. An esports sheet may have 200 columns, and 40 of them are blank across every tournament. A good reader of spreadsheets is someone who knows what those 40 columns are.

Goalkeepers' Feet and Fixture Density

There are two places where I believe the sports analytics industry misreads because it reads too few empty cells.

The first is goalkeeping distribution. Over the past seven years, goalkeeper passing accuracy has become one of the most cited metrics. Manuel Neuer is held up as the archetype. In Vietnam, Dang Van Lam has repeatedly been praised for his distribution.

My concern is not the praise. My concern is that this metric is often placed beside an empty cell: the basic reflex metric. In many datasets I have reviewed, a goalkeeper's expected goals conceded is merged into a single column with actual goals conceded, making it impossible to separate what belongs to the defensive line and what belongs to the keeper. When that empty cell goes unmentioned, a goalkeeper whose reflexes are declining can retain a high transfer valuation thanks to a handsome passing number. The market reads the published metric and ignores the missing one.

The second is injury. Blame usually falls on pitch conditions, luck, or physique. But when I total a squad's actual minutes across a peak month, the figure routinely exceeds the physiological recovery threshold. No medical department can rescue a team playing two matches a week. Clubs can buy more doctors, more recovery machines, more nutritionists, but they cannot buy more time. This is an empty cell that coaching staffs see and boards rarely budget for.

I raise these two points not to assert they are absolutely correct. I raise them because they are places where an empty cell gets read as a zero.

Correlation Is Not Causation, and an Empty Cell Is Not Evidence

Here I have to argue against myself.

The 15.3 percent drop in home advantage sounds very solid. But the Bundesliga returned mid-pandemic, with a compressed calendar, unusual accumulated fatigue, and unusual player psychology. I may have isolated a single variable — the crowd — when at least five variables moved together. My before-and-after comparison is tidier than reality. Reality is messy.

Likewise with the pressing finding about European champions. A figure below 10 appeared for every champion from 2026 to 2026. But that is five instances. Five instances is not a law of nature. It is a small sample, and I presented it in a piece that travelled further than it deserved.

And here is the subtlest error, the one I commit most: turning an empty cell into evidence. When a club does not disclose injury status, I tend to infer a serious injury. When a player has no match data, I tend to infer decline. Silence carries no positive information. It is only silence.

The human reflection I force into every piece is this: data is a map, not the territory. Maps contain blank stretches drawn in white. A good cartographer writes "unsurveyed region" rather than drawing an imaginary river into it.

People remember Hai Phong for the noise. I remember it for the subsequent success rate, and for one remark in a press room that I answered with data instead of anger. But I must also admit: there were times I answered with data when I should have said I did not yet know.

Signals for the Next Round

Three signals I am tracking in the period ahead, and how I will track them.

First, the quality of data collection in domestic leagues. When a league begins recording per-player passing metrics, empty cells shrink, and that is when transfer valuation models become more trustworthy. I will track it by counting exactly how many columns are published each round, not by reading commentary.

Second, fixture density. I will total each club's minimum rest days between matches during compressed periods and cross-reference against publicly disclosed injury cases. If those two curves move together for three consecutive months, I will have enough basis to write something serious.

Third, my own silence threshold. I will record how many times in a month I choose not to conclude. If that number is zero, it means I am writing carelessly.

My numbers do not need applause. They need to be right — time is the referee.

What I keep after twenty-two years of watching this industry is one small habit. Before writing anything, I look at the dataset and ask myself: which cell is empty, and why is it empty. Most answers are uninteresting. But those empty cells, rather than the filled ones, are where I find what to write next.