When the Data Goes Blank: Lessons From an Analysis With No Content

At 3:12 on the morning of 13 August, a file landed in my inbox. The name: "Deep Analysis — Stage 2". Inside were nine complete sections, comparison tables, a one-to-five-star rating scale, and a glossary of terms at the end. I scrolled to the first section. The conclusion column read: "Insufficient information". Section two: "Insufficient information". And so on to section nine, the same sentence repeating like a chorus.

Only one cell was filled. The field for domain: football.

The sender was a young colleague, with one line attached: "Could you take a look — should I add the numbers in?" I sat in front of the screen for a long while. I had asked myself that same question on a June night in Kazan, when the spreadsheet was full and the article was empty.

My first data notebook was a symphony, but back then all I could hear was the drum.

It is worth setting out what actually happened, because without the mechanism this is just a technical glitch.

In our workflow, every article passes through two stages. Stage one reads the source text and breaks it into information points: who, what, when, which figure, which source, what level of reliability. Stage two takes those points and analyses them in depth: tactics, finance, results, league position, risk, media. The symphony is only audible once stage one has written the notes.

This time, stage one returned nothing. The list of information points was empty. No club named, no player identified, no headline, no source, no publication date. One trace remained: the domain field was filled in correctly. Football.

That says a good deal. Classification usually runs on headline, URL or section tag — things that sit outside the body text. Extraction needs the body text itself. When the shell is filled correctly and the core is hollow, the failure is in retrieval, not classification. The source article may have been behind a paywall, rendered by script the crawler could not execute, reduced to a headline with no body, or lost to an encoding error and discarded silently. Nothing in the record identifies the actual cause.

More to the point: this happened mid-transfer-window. That is the densest period of the year in a football reporter's inbox, when forty links can arrive in a day, most of them unverified rumours, most of them from accounts that live on engagement. In that environment, an empty file is the most dangerous object there is, because it looks exactly like a complete one.

A complete template with empty cells is more dangerous than an error message. When a system reports an error, the writer knows to stop. When it returns a file with a title, nine sections, tables and a rating scale, the writer receives a different signal: the work is done, only a few figures are missing. I once stood in a stand while the scoreboard showed 0-0 in a match that had never kicked off. Nobody in the ground believed the board. But print that board on paper and hand it to someone who was not there, and they will write in their notebook: goalless draw.

In football we have another name for this phenomenon: a team sheet with eleven blank names but the formation still drawn correctly.

Circular dependency turns one small failure into a total collapse. The list of players is derived from the list of information points. So is the list of clubs. If the first stage extracts nothing, the second has nothing to identify, and with it goes every section that depends on names: tactics, transfers, league position, discipline, dressing room. What remains are the sections about method.

On the pitch, that is a side built around a deep-lying playmaker who is suspended. You lose one player. You lose the whole carrying system.

The real danger is the pressure to fill the gap. The more detailed the template, the stronger the pull to complete it. My young colleague asked whether he should add figures, and by the logic of the trade the question is reasonable: articles need numbers. But any figures added here would not come from the source article. They would come from memory, from habit, from what is usually true in similar pieces. That is the moment analysis becomes invention — and worst of all, invention that keeps the voice of expertise.

I have seen it happen to me. In April 2026, at an AFC Champions League match between Guangzhou Evergrande and Shanghai SIPG, I sat in the stand and counted seven losses of possession by right-back Zhang Linpeng. My editor asked for 1,500 words, then cut it to 500 and told me readers wanted the story, not the spreadsheet.

Then came the 2026 World Cup, when I followed Germany. After the 0-2 defeat to South Korea in Kazan on 27 June, I calculated that Germany's passing accuracy in the final third was 68 percent, 14 percentage points below their own figure in 2026. Those statistics were correct. My article was not read. A colleague's piece about chaos in the dressing room travelled everywhere. The ball is round, the story is not — the 2026 World Cup taught me to read between the numbers.

The lesson was not that data is useless. The lesson is that correct data can still be insufficient data, and where it is insufficient, people fill the space with something else.

In data analysis, a minimum sample is the only shield. An expected-goals divergence after three matches is noise, not a trend. A side that concedes fewer than its expected goals against over the first five rounds may simply be enjoying a goalkeeper in form. The rule I set myself: no conclusion about a gap between process and results without an adequate sample. The same rule applies to a file. Missing data is a different category from low-confidence data. It is a gap, and a gap does not get scored.

For that analysis, the methodologically correct outcome was to write "insufficient information" in every section and stop. Stopping there is a complete conclusion.

The transfer window makes everything harder, because speed is rewarded. A fee quoted today can be wrong within three days. A release clause can be misread because one sentence about its active period is missing. Agents have an obvious incentive to leak; clubs have an obvious incentive to deny; and neither needs to lie — only to under-say. In that environment, a file with no source, no date and no name cannot be given a reliability rating. It sits outside the scale, not at the bottom of it.

There is something else data will never capture. In 2026, when the leagues stopped, I still went to the training centre every day. The security guard there, Chen Rong, fifteen years in the job, told me about Yang Liyu — a player who came to the ground alone at 6:30 in the morning for 27 consecutive days during quarantine. I had intended to write a piece praising his discipline. Looking further, I understood he was in crisis because he could not see his family. Every empty stadium night is a slow drumbeat; I write it down so nobody forgets.

That detail appears in no spreadsheet. It is also the reason a writer cannot be replaced by a template.

There are two misreadings I encounter constantly.

The first: reading "insufficient information" as "no risk". That is the most serious inferential error in this trade. A club with no financial-breach news is not necessarily a clean club. A player with no injury news is not necessarily fit. A team absent from the rumour pages is not necessarily out of the market. The absence of information has never been evidence of anything, including something good.

The second misreading belongs to readers and writers alike: this industry rewards the product that looks complete. A piece with enough numbers, enough tables and enough conclusions will always be shared more than a piece saying there is nothing to say yet. The same holds in Vietnam: during every transfer window, a rumour about a foreign striker appears in the morning and by the afternoon sits on dozens of sites, with a fee rounded off for neatness and a phrase like "reportedly". The moment a rumour is written, it acquires a life of its own, and the correction that follows never runs as fast.

The counterintuitive point: the right decision on that file was to remove it from the workflow, not to grade it unreliable. A source with no headline, no author and no date cannot be rated, and what cannot be rated does not proceed.

The internal signal I will be watching over the coming weeks is a gate: whether a newsroom dares to publish an empty result. An organisation can build tables, charts and rating scales, but if it will not say "no data yet", then all of it is just a tidy presentation of guesswork.

The best sports writer is one who knows his notebook can lie. I do not write to predict; I write so that someone reading later will know how we lived.