The Empty Sheet Audit: When the Data Pipeline Returns Zero, Who Writes the Story?

Eleven fields on one sheet. Ten of them read N/A. The single field still alive held two words — cricket_world. No headline, no source, no publication date, no one-line summary, no author stance. The list of information points was empty. And in the field marked 'Entities Involved', where a value should have sat, there was an instruction instead: identify from the information points above. That field was never filled. It simply retained a note about how it was supposed to be filled.

My habit is to open the ledger before I read the story. For seventeen years I have sat over ball-by-ball logs, tagged every delivery into a phase, and asked every number the same three questions: where did you come from, who counted you, and what was your definition. So when a sheet comes back entirely blank, it is not a slow afternoon for me; it becomes a case. It is connected to cricket — that much is legible. What it is about — that is not. The whole of my work sits inside that gap.

The chain that runs before a story exists

Any match report or transfer report passes through a chain: headline → source → article type → summary → information points → entities → time sensitivity. Each link feeds the next. If the information points are empty, the entities cannot be derived. If time sensitivity is undetermined, there is no way to tell whether the document in front of me is a match report, a transfer story, an ICC governance item, or a piece about cricket's money — and those four demand four different audit methods.

On cricket websites, scorecards and live blogs are frequently rendered in JavaScript. A conventional extractor selector never reaches them; it returns an empty string, and every downstream link fills itself with a default value. N/A here is not a finding. It is the written form of an absence.

An article type of 'Unclassified' is a different class of event altogether. That is not a judgement about an article; it is a failure of classification. The summary and stance modules hang off that branch, and when the branch itself is missing, even the generic fallback never fires. In newsroom language: the compositor left the page empty and the press ran anyway.

This is why I pre-register my three adjustments and freeze them — venue, phase, era. Death overs means 16 to 20, the powerplay 1 to 6, the middle 7 to 15. Until a definition survives three seasons, the piece does not get filed. The remaining variables — the Impact Player rule, pitch, innings parity — go into a limitations paragraph, and I deliver the verdict at exactly the confidence the sample supports.

Three ledgers, built from the source up

In 2026 my first IPL assignment on a Bangalore digital cricket desk was RCB's fourteen-match home season at M. Chinnaswamy. I logged every ball and tagged overs 16 to 20 as death. At home, RCB's seamers conceded 12.6 runs per over in that phase; Yuzvendra Chahal's 18 death overs went at 8.1. The figure circulating on the desk had been built on a different definition, so I went back to the death overs and counted them myself.

The sample is small — fourteen home matches, eighteen overs from one bowler — and the claim cannot be made larger than it. So the piece ran with a methods table: which over counts as death, which delivery was assigned to which phase, and what was excluded from the count. Since then, 'clutch' is a word I do not use without the denominator printed beside it. That page also taught me that the labour of an audit goes into fixing definitions, not into adding numbers.

The following year I was put on the IPL mega auction and asked to build a phase-fit model: expected runs by batting position, death-over economy, powerplay wicket probability. The board flagged death bowling as the squad's biggest gap and rated Chris Woakes a 6.9/10 fit, with the warning that his powerplay overs would be overpriced at the money in play. Royal Challengers Bangalore bought Woakes for INR 7.4 crore at the 2026 mega auction in Bengaluru and finished the season sixth. The auction model had one price wrong, and it was Woakes.

One label has to stay attached here: a transfer fee is a number someone chose, not a law of nature. The model prices contract value, not merit. That is why I keep price and performance in separate columns, and state which question the model can answer and which it cannot.

The empty stadium was the cleanest experiment the sport has ever handed us, and letting it go still feels expensive. In 2026, sixty matches were played in the United Arab Emirates with zero spectators and a home-away distinction that existed only on paper. I matched the 2026 home splits against the 2026 venue-neutral data: chasing sides won 33 of 60, toss winners won 32, and death-over scoring moved from 9.9 to 10.6 runs per over.

Before drawing anything from those numbers, I matched ground dimensions and squad strength, because it is easy to say that home advantage disappeared when the crowd did. After matching, the answer was different: home advantage had not vanished, it had been removed from the sample. Crowds change rhythm, not talent. Such experiments do not repeat, so I keep that season's data apart.

In all three cases the raw file existed; the labour went into definitions and coding. The sheet that opened my morning did not even offer that convenience, because the log itself was empty. A small sample is a known limit — I write it in the limitations paragraph and move on. An empty sample is not a limit; it is a missing instrument. The two are worlds apart.

Where the mistake was mine

The reverse is also true, and it must be written down, or the whole accounting drifts in one direction only. Not every empty field is a failure of the game; sometimes it is my own selector failing. Hand me the page for that same article and our extraction layer might pull the entire ball-by-ball log and a story might exist. Then the problem is the instrument, not cricket — and changing an instrument does not change the world, only the page.

Second: disagreement is not my profession, and it should not become a habit either. Where the numbers agree with the crowd, that agreement should be written plainly. If a franchise has had a genuinely good auction and the model says so too, dismissing it as headline noise is not auditing, it is posturing. A correction that only ever runs one direction is not a correction; it is a bias.

Third, and most important: an empty field is not a discovery. This sheet says 'Unclassified', which tells us nothing true about cricket. It tells us the classification layer is broken. Conflating the two is the exact error cricket coverage repeats every season — mistaking the absence of news for a signal.

What to measure next time

Next time a cricket report comes back as an empty sheet, do one small thing before rewriting the prompt: run the extractor on that single article alone and measure the length of the body it returns. If it is zero, the fault is in the layer, not the story. Until then I will publish my raw table with the coding rules attached and invite the counter-count — because a claim that survives the next data release is always heavier than today's noise.