The Empty Frame: When Football Returns a Report With No Data

One night in Chengdu

The clock in my study read 2:14 in the morning. July rain fell evenly on the roof of the old apartment block, a sound I have heard for more than a decade of summers and never grown used to. On my second monitor, a nine-part analytical report lay open. I scrolled down, up, then down again. Forty-two rows of data. Not one row contained a real number.

The first column, where a competition name should have been, read “N/A — insufficient information.” The second column, where a club name should have been, read the same. The information-points list was bare. The entities-involved field had never been filled. Nine analytical dimensions — tactics and technique, club finance and the transfer market, results and public-opinion cycles, league landscape, rules and governance compliance, management and dressing-room, risk profile, media narrative, industry transmission — all ended with the same sentence: insufficient information to assess.

It is the most formally complete sports report I have read in twenty-nine years in this trade. And it contains not a single event.

An outsider would call it a failure. I am not so sure.

I found the error not in the centre of the pitch, but at the edge of the frame.

I first wrote that line in 2026, after spending nearly four months reviewing 147 controversial refereeing incidents across 28 rounds of the Chinese top flight for a sports television channel in Chengdu. Of those 147 incidents, 12 incorrect offside decisions traced directly to camera placement rather than to a referee’s eyes. The most memorable came in Round 25, Guangzhou Evergrande against Shanghai SIPG. Wu Lei’s goal was disallowed for a 15-centimetre margin, yet not a single camera in that match’s system sat on the horizontal plane required to draw the offside line.

I began building a private database of viewing-angle error. One line per incident: camera angle, distance to the touchline, frame rate, estimated deviation in centimetres. From then on, every piece of analysis I wrote carried a camera-angle diagram with it. That is how I learned that a wrong answer can always be corrected, while an answer that contains nothing must be sought elsewhere.

But there was a third kind of failure I could not yet distinguish back then: a process returns a result, the result has the right shape, the right headline, all the fields present — and nothing inside. Not wrong. Not right. Just empty.

Two tiers of a process

Modern sports analysis runs on a two-tier pipeline. Tier one reads raw text — match reports, club statements, transfer stories, referee reports — and extracts information points: who, did what, when, how much, sourced from where. Tier two takes that list of information points and applies a professional analytical framework on top: tactics, finance, rules, risk, media.

The entire power of tier two depends on one condition: tier one must return something. When tier one returns an empty list, tier two still runs. It still produces nine sections, tables, headings, footnotes. The form is intact. The content is zero.

I have met the manual version of this error many times; I simply never called it by that name. In 2026, at the World Cup final between France and Croatia, I sat in front of six different video feeds analysing the handball incident involving Ivan Perisic. Referee Néstor Pitana reviewed it on VAR and awarded France a penalty in the 35th minute. Of the six main broadcast angles I had, only one showed Perisic’s arm opening in the way people describe as “unnatural.” That was the only angle the referee was shown in the VAR room, across 1 minute and 47 seconds.

It took 37 rewatches before I understood that the human eye is not a measuring instrument.

What matters is not whether Pitana was right or wrong. What matters is that the system had pre-selected a very narrow set of inputs, then produced from that narrow set a conclusion with the shape of certainty. The other four camera angles did not disappear. They simply were not fed into the process. The output looked complete. The input had been filtered.

In 2026, when European stadiums played in silence, I had the chance to observe what I call VAR without a crowd. The empty stadiums of 2026 showed me this: VAR does not save football, it exposes football. Crowd noise had long masked most exchanges between referees and the VAR room. When the noise vanished, viewers heard the entire decision chain — and realised it is not a straight line from data to verdict, but a negotiation among three or four people holding different degrees of certainty. I logged dozens of those audible exchanges and sorted them into three groups: confirming a fact, checking consistency with earlier incidents, and agreeing on who carries responsibility. Only the first group concerns images. The other two concern people.

Across all three episodes — 2026, 2026, 2026 — the problem was never an absolute shortage of data. The problem was that data existed but never entered the process, or entered and was trimmed at some stage without anyone marking it.

Now return to tonight’s report.

Dissecting a blank page

An empty report is a thing of complete structure and zero content. On the face of it, useless. In my trade, however, it is useful in a very specific way, because it is the only trace showing where the process broke.

There are four hypotheses for an empty result, and they differ completely in consequence.

The first is an ingestion failure. The raw text never reached the extractor — a broken path, an unusual format, a validation step blocking the wrong thing. In this case the original article still exists somewhere intact; only the pipe is blocked.

The second is a parsing failure. The text arrived but the extractor did not recognise the structure. An unusual headline, a table layout, or mixed languages can cause the parser to skip the entire body and return an empty list silently. This is the most dangerous of the four, because it raises no alarm. It simply produces nothing.

The third is a genuinely content-free source. An empty file, a headline-only statement, an error page saved as an article. Here the empty result is the correct result.

The fourth is a deliberate empty result. The operator marked “insufficient information” on purpose rather than guessing. This is the only one of the four in which the empty result is a conscious act rather than an accident.

These four hypotheses lead to four different remedies. Check the path. Rewrite the parser. Close the file and find another source. Print the report and file it.

What caught my attention tonight is that the report is not silent in the same way everywhere. It is empty only where real facts are required — competition name, club name, figures, dates. Where only form is required, it is complete to the point of politeness. Tables have column headers. The risk framework has six category rows. The glossary defines its own terms. The document even flags three levels of risk, the highest being “empty input data.”

In other words, the report diagnosed its own condition.

In medicine this would be a patient who tests negative yet still gets admitted. In data analysis it means the pipeline did not collapse; it simply ran through a void. And the break sits at the boundary between tier one and tier two, not inside tier two.

This is where I want to pause a little longer, because it goes straight to my own work.

When I analyse a VAR incident, I split the process into three stages: capturing images, choosing which angle to present, and interpreting what is seen. Of those three, the second is the least discussed and the most powerful. I once recorded that across the 147 incidents of the 2026 season, several would see the certainty level of a conclusion drop from “clear” to “possible” simply by changing the presented camera angle. Same incident. Same referee. Same law.

The two-tier analytical pipeline works identically. Tier one captures. Tier two presents and interprets. When tier one returns an empty list, tier two can still present beautifully. It just has nothing left to interpret.

We thought we were searching for justice; it turns out we were only searching for a prettier camera angle.

I wrote that line for a 2026 piece on VAR and still find it accurate when applied to data reports. A report with nine full sections and no verifiable facts resembles a VAR room with the screens on, the referee seated correctly, the procedure followed to the form — and no image ever displayed.

There is one more detail worth noting. The empty report never tries to fill its gaps. It does not write “the club is understood to be negotiating,” does not write “sources close to the deal reveal,” does not write “according to unverified reports.” In precisely the places where an ordinary sports story would place a soothing sentence, it places a named blank.

Naming a blank is a technical act, not a literary one. Writing “insufficient information” pinpoints exactly where the deficiency is. Writing “reportedly” conceals it.

During the transfer window, that second manoeuvre occurs more often than at any other time of year. I have tracked the transfer market long enough to separate three kinds of story: stories with a contract, stories with an agent, and stories with nothing but temperature. The third kind accounts for most of the traffic and nearly all of the error. A club that has never sent an offer can still appear in ten stories in a single day, if enough people mention it at once.

Tonight’s report is the opposite type. It carries no temperature, no source, no agent, no contract. And for that reason it is the most honest piece of reporting of the day.

At 45, I have softened noticeably in how I judge such documents. At 35, I would have thrown it in the bin and called it a defective product. Now I keep it, bookmark it, and use it as a reference specimen for a concept the sports-analysis industry badly needs and barely measures: the data-completeness rate.

The data-completeness rate is the ratio of mandatory fields correctly filled to the total number of mandatory fields. For an ordinary match report it is usually recorded above 90 percent, which is why nobody rechecks it. Once I did recheck, a troubling pattern emerged: the most important fields — data provenance, capture timestamp, measurement conditions — are the ones with the lowest completion rates. The least important fields, such as competition name, round, and match date, are almost always complete.

Tonight’s report is the extreme version of that pattern. It has the full framework for nine analytical sections, but a factual completion rate of zero. Measured on form, it scores perfectly. Measured on content, it scores nothing. The gap between those two measurements is where most analytical error in football hides.

I used that measurement to audit several transfer datasets available to me during the current window. The result was unsurprising. Fields covering transfer fees, contract duration, and release clauses are typically filled with far lower accuracy than fields covering player and club names. Yet those first three fields are exactly what determines the real value of a deal. A story saying club A is interested in player B can be entirely wrong without anyone checking. A story saying B has 11 months left on his contract is verifiable, and far more important.

This is why I always tell younger editors that the structure of clauses and the wage bill is the real transfer-window story, while the names are merely the presentation layer.

Back to the central issue. If an empty report can appear without anyone noticing, then the next question is: how much football analysis published every week is built on an empty tier one that nobody checked?

I have no figures to answer that quantitatively, and I will not invent any. But I have one qualitative observation solid enough after twenty-nine years: wrong conclusions in football analysis almost never originate in the interpretation stage. They originate in the input stage. An expert can interpret brilliantly from a bad dataset, and the result will be a brilliantly presented but wrong conclusion.

In refereeing, this is called an error at the fact-establishment stage. A referee is not wrong when applying the law; a referee is wrong when establishing the wrong fact to which the law applies. The same logic holds for data analysis.

There is one marker of an empty tier one that I have tested often enough to trust: when an analytical report cites no verifiable fact at all, it is almost certainly not because the author was lazy, but because the author had none. Writers who hold data always leave traces of it — a timestamp, a figure with units, a full proper name. Tonight’s report leaves no such trace on any line.

And this is the part I consider most important in the whole story. When tier one is empty, the process does not stop. It keeps running. It produces a document that can be read, cited, shared, and used as the basis for a decision. Nothing in the pipeline automatically blocks that, unless the operator has enough composure to block himself.

In football we have such a mechanism, and we call it VAR. But as noted above, VAR can only check what is fed into it. It cannot check what was removed before it was switched on.

When a gap becomes a temptation

The natural human reflex before a gap is to fill it. This is not a defect specific to football; it is how the brain operates in order to survive. In football, however, that reflex carries concrete professional consequences.

An editor needs a headline. A fan needs a conclusion. A sponsor needs a forecast. The sentence “we have no data” satisfies none of those three needs. The sentence “the club is likely to act in the coming days” satisfies all three, and cannot be disproved within a week.

I have seen this mechanism from both sides of the screen. In my reporter years I wrote sentences like that — not because I believed them, but because the newsroom needed a story and I did not yet have enough data to write the right one. Later, working in data analysis, I understood that each such sentence was an occasion on which I trimmed my own tier one.

From a referee’s standpoint the problem is even starker. A referee has no right to say “I don’t know.” The match keeps running. The clock keeps running. A decision must be produced within seconds, even when every camera angle is inconclusive. My trade calls this decision-making under imperfect information, and it is why the phrase “clear and obvious error” exists in the VAR protocol.

That phrase is a compromise. It says VAR intervenes only when the mistake is large enough, to avoid re-adjudicating every phase of play under a microscope. But it never says how large “large enough” is. There is no definition in centimetres, seconds, or degrees. No threshold is written into the law. Which means the threshold is set by human perception, match by match, league by league, moment by moment.

In 2026 I measured that the accepted offside-deviation threshold differed between matches within the same round. Not by much, but it differed. And when the threshold shifts, consistency vanishes — not because referees are poor, but because the standard does not exist in measurable form.

Tonight’s report chooses the opposite path. It sets a threshold of zero. No data, no conclusion. In an industry where everyone is racing to say more, faster, more confidently, a document that is entirely silent.

I do not hold that silence is automatically a virtue. An empty report produced out of laziness is still empty. But there is a difference between a gap left unfilled and a gap filled with unverifiable material. The first invites scrutiny. The second invites belief.

During the transfer window, the second always wins on volume. That is something I already knew and no longer find surprising.

What I keep

I have kept tonight’s report in its own folder, beside the viewing-angle error database I built from 2026 onward. The two belong together: one is a set of incidents where the process had data but not enough of it, the other a document where the process had no data at all. Both point me to the same joint in the pipeline — the boundary between what is captured and what is presented.

If you read a piece of football analysis this week, try one simple exercise. Count the verifiable facts in it: an absolute date, a figure with units, a full proper name, a contract clause you can look up. If the count is zero, you are reading the body of a report whose input tier never finished running.

And once you have counted it one time, you will not be able to stop counting.