The Empty Cell: An Audit of Stage-1's Zero Rows, and Where Football Data Pipelines Actually Break

1. The Third of July, and an Empty Cell

Moscow, 3 July 2026. The fifth round of the shootout between England and Colombia at Spartak Stadium. On my laptop was a sheet listing every World Cup shootout kick since 2026 — who took it, which foot, which way the keeper dived, which round of the shootout. Hand-coded three weeks earlier, the sheet had already flagged Carlos Bacca's tendency. Jordan Pickford dived to his left and stopped the ball. The broadcast producer sitting beside me did not even look across. Earlier that same day he had told me on the record that xG was a fad and that tactics were not my lane. Nobody in the press box agreed with the log. The log was right.

Today I have a sheet of the same species in front of me. It has eight fields. Seven are empty or N/A. The eighth contains a single word: football. That is the whole of Stage-1. No title, no source, no article type, no information points, no entities, no time sensitivity. You cannot build a shootout database from this sheet. You cannot build anything from it. And that is the most interesting fact in the entire file.

When I started a kinesiology degree in Liverpool as a sixteen-year-old, I assumed the hard part of this work would be gathering data. By the time I took a junior event-coding job at a football data collector in Bootle in 2026, I had learned that the hard part is confirming the data arrived at all. This piece is about that confirmation. It is not a match report. It is an audit — of one empty sheet, and of the habits that fill that emptiness with things that were never there.

2. What Stage-1 Is, and Why Nothing Downstream Moves Without It

Football data journalism runs in three layers. The first is extraction: pulling raw material out of a match report, a federation circular, a press conference transcript, a transfer window document, a league table. The second is audit: isolating every sentence in that material which can be verified independently. The third is analysis: joining those verified sentences into a conclusion nobody has published before.

Stage-1 is the name of the first layer. It turns raw material into structure — title, source, type, claims, entities, time limits. If that sheet is empty, the second and third layers have no work left to do. You cannot draw a conclusion from evidence that does not exist.

I have done this by hand. In October 2026 I was sent to the FIFA U-17 World Cup in India. Fifty-two matches. The final was England 5-2 Spain at Salt Lake Stadium, Kolkata, before 66,684 people. I coded that final alone — ninety minutes, more than 1,100 events, no second analyst beside me. An English freelancer on the press row addressed me in Hindi, assuming I was the translator. I slid my timestamped log across the desk and said nothing at all.

After that I started a private error ledger. Every misread pass, every misattributed shot, dated and explained. Within a year, a short method note travelled with everything I filed — the data source, the sample size, the limits of the claim. Editors began trusting my figures over a rival's because of that note.

I raise all of this because the sheet in front of me looks familiar. It is the sheet I would have handed over without an audit, and the person who received it would have written something brilliant on top of it. Roughly forty per cent of that brilliance would have been my imagination.

3. Question One: Why Title and Source Are Non-Negotiable

The first two fields on this sheet — title and source — are both empty. Technically that is an absence of information. In practice it is an absence of context, and the difference matters.

In football, the same claim becomes a different object depending on where it came from. Take the claim that a club bought a defender for two million euros. If it comes from the club's official statement, it is a ledger entry — date, fee, contract length, all verifiable. If it comes from a source close to an agent, it is a possibility: the fee may be inflated, deflated, or split into instalments. If it comes from a fan account, it is entertainment.

All three require different verification standards. I can print the first as fact. The second I write in conditional language and name the source, so the reader can judge. The third I do not print at all.

An empty title-and-source field means I cannot determine which category the material belonged to. And if I cannot determine the category, I cannot select a standard.

One example from my own archive. 13 July 2026 — East Bengal 4-1 Mohun Bagan, Federation Cup semi-final, official attendance 131,781. I always write that figure as official, always name the competition, and never call it a final, because it was a semi-final. That is a discipline of two words — 'semi-final' and 'official' — and it is the discipline that separates a verifiable figure from a decorative one. Without the source field, you do not have that discipline available to you.

4. Question Two: Why 'Unclassified' Is the Most Expensive Error

The second field — article type — reads Unclassified. This is where I get most uneasy, because in football journalism type is not a label. It is the entire verification frame.

For several years I have sorted football material into five categories, and each has its own rule set.

| Type | Core claim | Verification level | Biggest risk | |---|---|---|---| | match-report | score, goal times, lineups | high — official match sheet | mis-timed events | | transfer-rumor | fee, contract, signing date | low — network of sources | inflated fees | | tactical-analysis | formation, xG, pressing triggers | medium — model-dependent | treating model output as fact | | opinion-column | judgement, evaluation | not applicable | opinion presented as fact | | press-conference | manager/player statements | high — transcript | quote cut from context |

That table is my current working method, and it only functions because I know which type of material I am holding. When the type is Unclassified, the table is worthless.

The real difficulty is that all five types are written in the same language. A transfer story and a tactical analysis both contain active verbs, numbers and names. The difference is that in the first, every sentence is a claim that can be verified; in the second, every sentence is an interpretation that can be challenged but not called false.

July 2026. My employer cut twelve data roles, mine among them. I kept my forty-page methodology audit and launched a paid newsletter. Two hundred and sixty subscribers in the first month, nine hundred by December. That growth came because I stated at the outset what kind of material I claimed and what I did not. A reader does not come to me for transfer news. They come for the audit of how reliable transfer news is. If the type is unclear, that bargain collapses.

5. Question Three: The One-Line Summary and the Author's Position

The third and fourth fields — one-line summary and author stance — are both empty. They look separate but do the same job: they state what the material wants to say, and in whose interest.

In global sports business this distinction becomes very concrete. Suppose an ISL club moves its shirt sponsorship from a local business to a global brand. One report can say: 'the club expanded its partnership to grow brand value.' Another can say: 'the club broke a contract with the community that bought its tickets for twenty years.'

Both sentences describe the same event. The difference is the direction in which the writer leans. The author-stance field records that lean. If it is empty, I cannot tell whose interest the material serves. And if I cannot tell that, I cannot use the material as fact.

I have a personal example. Since I was seventeen I have watched shirt sponsorship money pull clubs away from their cities. When I look at the Kolkata derby attendance — that 131,781 of 13 July 2026 — and compare it with the brand-exposure figure a club reports in its annual review, I can see two different things being measured. One is a community. The other is an audience. Every club uses one word for both, and that word is the fraud.

I do not write this as a declaration. I show it through case selection — which figure goes at the bottom and which at the top. The author-stance field is what governs that selection.

6. Question Four: Information Points, the Column Without Which Nothing Works

Here the audit settles on one point. Of all the Stage-1 fields, exactly one cannot be trimmed, and that is Information Points. Every other field is a convenience. This one is the only necessity.

An Information Point is a sentence that can be detached from its source and checked independently. The most useful ones carry four parts: the claim, the claimant, the location of the claim (which pitch, which document, which minute), and the date of the claim.

That list is now empty in front of me. This does not mean the article contained no facts. It means the extraction step failed to record where the facts were. Those are different propositions, and the difference changes the whole diagnosis.

Here is how I explain it. In June 2026 the Bundesliga returned behind closed doors on 16 May. I pulled the first hundred matches after that restart and looked at the home-win rate. It had fallen. My home-advantage coefficient, built on a decade of full stadiums, had stopped working. I rebuilt it rather than reporting around it.

Later that year I logged the 2026-21 ISL from Liverpool. The entire season in Goa, three venues, no fans, no promotion, no relegation. I woke at 4:30 a.m. for 7 p.m. IST kick-offs. That season taught me that a figure which is not data is only a promise.

If a machine handed me an empty list for that season, I would not write analysis on it. I would first ask whether the machine had read the raw text at all.

7. Question Five: Entities and Pronoun Resolution

The fifth field — entities — reads 'not extracted.' This looks like the dullest field on the sheet and is in practice the most destructive.

Football's language is built on pronouns. 'He said,' 'the club announced,' 'they confirmed,' 'the team made a change.' All of these mean one thing: the writer is treating a specific person or institution as already established in context.

If entity extraction has not run, I cannot know which club 'the club' is. In a league where names have changed repeatedly, that is dangerous. Indian football has seen clubs change owners, names and cities over the past decade. 'The club asked for the entry fee to be waived' — which club? Which year's club? Two separate institutions have existed under the same name in two separate years.

I learned this years ago. At the Bootle desk, the rule was: do not log a pass until you are certain of its recipient. If the recipient was in doubt, we used a separate code. A handful of events went into that category every week, and every week somebody cleared them out ruefully.

By the same logic, if pronoun resolution has not happened at entity level, no claim can be attached to any person. And a claim attached to nobody is not a claim.

8. Question Six: Time Sensitivity and Its Expiry Date

The sixth field — time sensitivity — reads 'not assessed.' This is the field I insist on most, because I have paid for getting it wrong.

Every fact in football has an expiry date. Injury status goes stale in hours. A fee is meaningless after the window shuts. The league table changes weekly. A manager's remark is stale after the next match.

So the time-sensitivity field demands two things: the publication date, and the validity period of the claim. An Information Point that does not say how long it will remain true is a trap.

One concrete example from Indian football. The FIFA ranking. India's best-ever position is 96, reached in February 2026 and equalled in July 2026. For most of the intervening period India has sat in the 100 to 130 band. Inside that band, small movements arrive every month. A three-place rise in one month is not progress; it is noise. A real trend only emerges from a multi-year series.

Now consider what happens when the time-sensitivity field is empty. Somebody lifts a 'recently' out of a federation press release and attaches it to a statement that is two years old. The reader assumes it is today's news. That error is not careless; it is a direct consequence of the empty field.

In my newsletter every figure carries its date. An undated figure does not get space. That rule cost me two paying clients who found it harsh. It is also why my readers know that when I print a number, it was true on the day it was measured.

9. Question Seven: A Label Is Not Data

The eighth and final field is the one that contains something: Domain Label: football.

This is a thing I take seriously, because it is the easiest lie on the sheet. It looks like information. It is one word, it is clean, and it will show green on any dashboard. But nothing can be decided from it.

============================

| Field | Status | What it decides | |---|---|---| | Title / Source | N/A | nothing | | Article type | Unclassified | nothing | | One-line summary | empty | nothing | | Author stance | N/A | nothing | | Purpose | N/A | nothing | | Information Points | empty list | nothing | | Entities | not extracted | nothing | | Time sensitivity | not assessed | nothing | | Domain label | football | nothing |

============================

The most striking part of this table is that eight of its nine rows reach the same verdict. It is a table that is not a lament but a conclusion. When you count a football question rather than assert it, every row often gives the same answer, and at that point the argument is over.

The label 'football' tells me which dictionary to use. It does not tell me which sentence to read. And analysis happens on sentences, not dictionaries.

10. Where the Pipeline Broke: Four Hypotheses

Now the real question. If the sheet is empty, where did the emptiness come from? From experience I have four hypotheses, and an order in which to test them.

Hypothesis one — the upstream fetch failed. The raw text never arrived. The link was broken, the page was behind a login, or the content simply was not available when the fetch ran. In that case the extraction step made no error: it returned what it found, and it found nothing. This is the most common and least discussed failure. The way to catch it is simple: check the fetch log for the status code and the byte count. A zero-byte response produces an empty sheet, and the next step quietly passes it along.

Hypothesis two — the parse failed. The raw text arrived, but its markup had changed. Football websites change templates every few months, and each change blinds an old parser. The text exists; the sheet says nothing was found. This is the second most common failure. Catch it by measuring the text length against the empty result. If the text ran to thirty thousand characters and extraction returned zero, the fault is in the parser, not the text.

Hypothesis three — the schema mismatched. The text arrived, the parse ran, but the extracted fields did not match the expected schema. The extractor labelled a goal as a 'score-event' while downstream looked for 'goal.' Both were correct, and the result was zero. This failure is the cunning one, because no single component is wrong on its own.

Hypothesis four — the silently swallowed exception. The most dangerous of the four. Somewhere a try-catch block intercepts every error and returns an empty list. The pipeline never crashes, no alarm fires, and a zero arrives every time. I have seen this in employment. When twelve data roles were cut in July 2026, one reason was that the people above believed the system was running, because their dashboard was green. The dashboard was green because every error had been suppressed.

Taken together, the four produce one general rule: an empty result is not a fact; it is a symptom. Treating it as a fact is the error that hollows out the whole chain.

11. What a Populated Sheet Looks Like — A Template

I want to be explicit that what follows is not the extraction of any real article. It is a blank template, with the field names set out and a description of the kind of value each should hold. It is a template, not a fact.

| Field | What to fill | Minimum condition | |---|---|---| | Title | the piece's title, as written | no editing | | Source | publisher + publication date (day-month-year) | date mandatory | | Type | match-report / transfer-rumor / tactical-analysis / opinion-column / press-conference | pick one | | Core claim | one sentence, twenty words maximum | no adjectives | | Information Point | claim + claimant + location + minute/page + date | each a separate field | | Entities | person / club / competition / venue, pronouns resolved | resolution mandatory | | Time sensitivity | publication date + validity period | expiry mandatory | | Verification status | verified / unconfirmed / disputed | name the source |

I built this template three months ago. It took a day. Using it costs me twenty minutes per article. My correction rate has fallen to almost nothing. That is the arithmetic people are reluctant to count, because the investment shows at the top and the saving hides at the bottom.

12. The Real Price of an Empty Cell

Now the part that angers me most, and I will put it in numbers.

Across the entire chain of trust, Stage-1 is the cheapest step. Running it takes seconds. Building it takes months. But an empty Stage-1 does not reduce everything downstream to zero — it converts it into invention.

Consider what happens when a newspaper or newsletter writer sits down with an empty sheet and a fixed word count to fill. The sheet gives them nothing, but the deadline demands something. In that moment the writer does three things. They lift material from other sources and fold it into this piece. They fill gaps with general knowledge. And they do the most dangerous thing of all: they draw a conclusion and write it as a fact.

This is where most of the untruth in football journalism is manufactured. Nobody intends to lie. They are trying to fill an empty sheet.

I understood this while logging the 2026-21 ISL — three venues, Goa, no crowds, no promotion, no relegation. That season produced a great deal of writing claiming the league was becoming more competitive. The truth was that the season had no stake at all. When no team can go down and none can come up, the points table is a count, not a contest. But the empty sheet had no field for that.

This is the centre of my method. If it is not in the log, I will not put it in the log to reach a conclusion. The sentence looks simple. Living up to it is a full-time job.

13. The Closed-Shop Ledger — When Counting Wins the Argument

Now the example that shows most clearly how far apart a populated sheet and an empty sheet really are.

From 2026 to 2026 the Indian Super League was a closed shop. That does not mean the league was bad. It means the doors in and out were shut. I never wrote a direct opinion in that argument, because opinions were cheap. I wrote counts.

============================

| Measure | Count | |---|---| | Clubs admitted (per AIFF circulars) | countable | | Clubs arriving by promotion | zero until 2026-24 | | First promotion slot | 2026-24, Punjab FC | | Second promotion slot | 2026-25, Mohammedan Sporting | | I-League seasons completed | countable | | I-League seasons abandoned | countable |

============================

The conclusion from that table is not a lament. It is arithmetic. Across roughly a decade, not one and not two but a single promotion slot opened for Punjab FC in 2026-24 and a second for Mohammedan Sporting in 2026-25. That is a number. You may like it or not; the number does not move.

I write this because it is exactly what a populated sheet provides and an empty sheet cannot. If the Information Points had been filled — on which date which club was admitted, on which date which season was cancelled, which circular said what — you could build that table yourself. With an empty sheet you cannot build it. You can only write an opinion, and no editor can check an opinion.

The notebook that arrived from Kolkata had this answer written in it eleven months before the argument asked the question. That is the entire value of an archive.

14. The Derby Record Book, and the Conditions of Precision

I keep this book in three parts — date, competition, official recorded figure. All three are mandatory.

13 July 2026. Federation Cup semi-final. East Bengal 4-1 Mohun Bagan. Official attendance 131,781. I have attached four rules to that figure, and all four map directly onto fields in my sheet.

One: I always write it as 'official,' never 'estimated.' Two: I always name the competition, because 131,781 means nothing if you do not say which competition it came from. Three: I never round it upward — writing 131,781 as 'about 132,000' is a lie, however small. Four: I never call it a final, because it was a semi-final.

That fourth rule is the most necessary and the most frequently broken. Calling a semi-final's attendance a final's attendance enlarges the romance and shrinks the precision. And a reduction in precision means that next time you cite a figure, the reader will not bother to check it. Precision is a slowly spent capital, and rounding it upward spends a great deal at once.

The second job of this book is to show me that Kolkata club crowds run above sixty thousand while the national team draws considerably thinner outside Kerala, Goa and Bengaluru. That gap is not a political statement; it is a comparison. But writing that comparison requires two separate books, each with a date and a competition on every row.

When the sheet is empty, that book becomes useless, because it has no basis on which to add its own rows.

15. The Chhetri Ledger and the Second Meaning of Emptiness

Sunil Chhetri's international goal count is 95. It is a closed, itemised record that closed in November 2026. That is how I enter it — a list closed with a date, not a mood.

The question that follows is not how sad we are. The question is who scores next, at what level, from which academy, in which competition. That is a count of names. And counting names requires the one thing missing from this empty sheet: Information Points.

For several years I have kept U-17 and U-19 score sheets. There is no great discovery in them. There are names, dates, and which state each player came from. When somebody writes that Indian football has a striker crisis, I ask how many score sheets they hold. The answer is often zero. The crisis is elsewhere.

16. The Contrarian Angle: The Zero May Be the Only Honest Artifact in the Room

Now the part where I have to reverse my own argument, because if I do not, I will commit the very error this piece is written against.

So far I have argued that an empty sheet is a symptom. That is true, and it is half the truth.

The other half is this: if the source material genuinely did not exist, the empty sheet is the only honest artifact in the room. And the person who fills it in and sends it downstream is the one introducing the actual error.

The distinction matters. On one side is a technical failure — the material existed and the pipeline failed to read it. On the other is a logical failure — the material did not exist and somebody performed as though it had.

In the first case the fault lies with the machine. In the second it lies with the person who forces the machine never to return empty-handed. I have seen this second kind of failure often, and it always has the same causes: a deadline, a word count, and a tool that guesses rather than returns zero.

This is where I have to doubt my own models. My home-advantage coefficient was one such artifact, and when the stadiums emptied it broke. I had two options: hide it and write on old numbers, or rebuild it. I chose the second, and I recorded the error. My newsletter carries a public corrections ledger — every retracted figure, its date, and the reason. It cost me. Two paying clients left. But the trust of the analysts whose numbers I now verify before quoting was built on that ledger.

A coefficient is only a promise, and promises expire when the stadiums empty. The same applies to an empty sheet. Even a zero should carry an expiry date.

17. What to Watch Next

If this sheet came out of an automated process, the repair work is not in the analysis. It is in the extraction. I am keeping three signals, and all three can be measured in any future tournament cycle.

Signal one — the ratio of text length to extracted facts. If the text runs to thirty thousand characters and the fact count is zero, the problem is not analysis. Signal two — the count of empty results, with dates attached. If that number rises above ten per cent, you are not looking at a shortage of material; you are looking at a sick pipeline. Signal three — the pieces published with a full word count after an empty sheet. Set those aside and read them, and count how many sentences have no source at all.

There is a column for what happened and a column for what we wanted. A writer who merges the two saves a deadline and spends their capital. The third time they cite a figure, nobody will check it.

Nobody in the press box agreed with the log. The log was right. I write this now because that same log has sat me down in front of an empty sheet and told me that between silence and lying there is a third road: measure it, record it, date it, and if you find nothing, write a zero.