A Spreadsheet of Empty Cells: The Line Between Sports Analysis and Fiction

Two forty-seven in the morning, Nagoya. I open a spreadsheet with forty-seven columns and three hundred and twelve rows, the standard framework for a post-match read. Column for first-half PPDA. Column for transition xG. Column for touches inside the penalty area. Column for midfield duel win rate. All empty. Not a single number. Not a single player name. No competition, no timestamp. The sheet returns exactly one value, repeated in every cell: insufficient information.

Nine years into this job, I am used to missing data. I have never met missing data this clean, missing so completely that no gap remains to slip a finger into. And the first feeling was not confusion. It was itchy fingers. Professional instinct pushed me to fill it in. That was the most writable moment of the week.

My analysis runs in two stages. Stage one is decomposition: identify the subject, the competition, the time marker, the raw information points. Stage two is deep analysis: build out the technical frame, player form, tournament system, world landscape, rules and institutions, coaching staff, risk surface, public narrative, and the industry transmission chain. Eleven blocks, hundreds of cells.

When stage one returns zero, stage two has nothing to stand on. Technically, the process must stop. But the profession does not stop. Deadlines do not stop. Editors do not stop. And more importantly, readers do not stop. Every morning thousands of people open a sports page waiting for a read. The emptiness of the data does not reduce demand for conclusions. It only pushes the pressure to manufacture conclusions out of nothing one level higher.

I have watched that pressure turn into product, and I have also watched it get blocked. In 2026, when the Bundesliga returned after lockdown, I sat typing match after match into a separate sheet: twenty-eight matches without crowds, home teams won seven. Twenty-five percent. The previous season, the matching figure was forty-one percent. I wrote it up, posted it to an international analytics forum, and a German editor shared it. There was nothing miraculous in it. I simply chose to count instead of choosing to narrate. Home advantage was never an advantage, only noise encoded into goals. When the noise disappears, the number testifies on its own.

But tonight, I have nothing to count.

This is the moment to state plainly a mechanism the analytics world rarely admits: when a data cell is empty, the writer's instinct is not silence, it is substitution.

The first mechanism is substitution by proxy. No PPDA, so use successful duels. No xG, so use shot count. No shot count, so use possession share. Every substitution sounds reasonable, because these variables correlate in most matches. But correlation is not causation, and a wrong proxy delivers a wrong conclusion with very high confidence. I once read an analysis concluding a team pressed badly simply because that team chose to concede the ball. Their PPDA was low, but low because they opted for a controlled low block, not because the midfield was lazy. The data was right. The conclusion was inverted.

The second mechanism is sample inflation. Three matches are three matches. But once a story is already in place, three matches quickly become a trend, then an identity, then a system. I wrote my thesis on decoding the low block, and the biggest lesson came not from the model but from checking sample size. When Morocco shocked the world at the 2026 World Cup, I re-ran every match in the dataset. In the group stage, their opponents generated an average of 0.6 xG per match. In the knockout round against Spain, they allowed eleven touches inside their own penalty area and not a single shot on target across one hundred and twenty minutes. Morocco's approach was an extreme form of passive pressing, holding roughly twenty-five metres between lines. Those numbers need no embellishment. But if I had only one match, I would not have dared write a single word.

The third mechanism, and the most dangerous, is narrative backfill. The writer starts from the conclusion and works backwards to find evidence. The conclusion comes first, the data comes after. When stage one is empty, narrative backfill meets no resistance at all, because no data can contradict it. An empty cell is a blank page, and a blank page always agrees with whoever is writing.

I have been on the other side of this mechanism, in the role of the suspect. 2026 World Cup, I was seventeen, sitting in front of the television in Nagoya watching Japan face Belgium in the round of sixteen. Japan led two nil through Haraguchi and Inui, then lost two three in the last fifteen minutes. All of Japan talked about spirit. I scraped every passage of play and took notes: in the first half Japan had a PPDA of 6.7, pressing very well. In the second half, PPDA dropped to 14.8, the midfield stopped closing down, and Belgium's xG rose from 0.7 to 1.9. I wrote it up on a blog with those numbers as evidence. A group of male supporters commented: what would a girl know about tactics.

I did not answer with emotion. I answered with a chart.

Precisely because I lived through that, I understand why narrative backfill is so tempting. The conclusion is the part that gets rewarded. The process is the part that gets skipped. Nobody shares a spreadsheet. People share a sentence.

Euro 2026 is the counter-example, and the lesson I still keep. Before the tournament I analysed Mancini's Italy: average PPDA of 9.1, ahead of every major side, sixty-two percent possession, very few goals conceded. I wrote that Italy were not defending in the traditional sense, that they pressed through controlling the ball, and I predicted they would win. A male journalist on Twitter mocked me: what would a woman calculating xG know about the history of Italian football. I stayed quiet. Italy lifted the trophy, conceding exactly three goals across the tournament. My old piece was dug up and shared more than two thousand times.

What stands out is that I was right because I had data, not because I had belief. If my stage one had been empty that year, I could still have written a very persuasive piece about Italy, and it would have been garbage beautifully presented.

When Japan push high, I do not see miracles, I see the formula of collapse. That sentence holds for a team. It also holds for an article. An analysis that pushes high without data to press with is self-destruction: it says a great deal, asserts forcefully, and has nothing underneath.

There is another data family caught in the same trap, and I want to say it plainly. Distance covered and sprint counts get packaged as effort indices, but ineffective running also produces pretty numbers. A player covering twelve kilometres in a match proves nothing beyond the fact that he covered twelve kilometres. I once built a heat map for a midfielder with the highest distance figure of the round and found that most of that distance sat in dead zones, lateral running, chasing a tempo the opponent had set. So when an analysis praises effort through distance, I always ask more: where was that distance, when, and for what. For the same reason, I do not read the five-substitution rule as a purely positive reform. It deepens squads, true, but it also turns the final twenty minutes into a war of attrition, where the team with the better bench rather than the better football takes points. I have checked this against season data: the share of goals from the seventieth minute onwards rises markedly in leagues using the full five substitutions, while chance quality falls. More goals, but not better goals.

I live in Japan, I write about badminton for the Japanese market, and I read Vietnamese football from a distance. I do not compare the two scenes through pride or through an inferiority complex. I compare with numbers: official matches per year for a young player, flight hours per season, registered squad slots in the national league, the average number of years a young talent is retained before being sold. Those numbers know neither pride nor shame.

This is where I want to go against my own profession.

In sports analytics, people fear the lying analyst. That fear is aimed at the wrong address. The lying analyst is easy to catch: data can be verified, sources traced, numbers cross-checked. He will be caught, sooner or later.

Harder to catch is the honest analyst who is trying. That person does not invent numbers. That person simply fills the empty cell with the most plausible thing. That person writes may, appears to, the trend suggests, words soft enough never to be fully wrong and firm enough always to look right. That person produces something unfalsifiable, and something unfalsifiable belongs to literature, not to analysis. The problem is not the writer's ethics, it is a professional structure that rewards conclusions and does not reward silences.

Every pressing sequence is a testimony, every number is a confession. But when no pressing sequence was recorded, there is no testimony at all. And an analysis written under those conditions, however beautifully presented, is a testimony the investigator wrote himself.

People need belief to place a bet, I need data to be certain. Tonight I have no data. So I am not certain. The only professionally correct act is to say so, and leave the cell empty.

Meanwhile, in another city, a twenty-two-year-old woman is typing scores for a provincial badminton tournament in central Vietnam. She has no software, no multi-angle cameras, only a notebook and a phone filming each rally. She will log two hundred matches before anyone notices her. I would rather write for her than for glossy analytics decks. The only decent way to do that is never to tell her that sitting still can also produce an article.

My spreadsheet is still empty when morning comes. I do not fill it in. I rename the file no-data-available and leave it exactly as it is.

I do not know whether the data will arrive next week. But I know the first thing I will check when it does: the gap between a row of data and a sentence of conclusion. If that gap is wider than the sample size allows, I will delete both and start again. In this trade, keeping one cell empty is sometimes harder than filling an entire sheet.