The Lesson of the Empty Scorebook: Why Questions Must Precede Data in Cricket Analysis

There is a page in an old scorebook at M. Chinnaswamy Stadium in Bangalore that carries no figures. Across Karnataka's 2026-84 Ranji home season I hand-coded 4,180 deliveries — line, length, field position, bowler spell, batter handedness. But the page I am talking about has no ink; it has a question written on it. The question was: do I have the right thing to measure?

A document reached my desk a few days ago that revived that old question. It was the output of a cricket-analysis pipeline — the result of stage one, meant to be the foundation for everything downstream. The domain label was printed clearly: cricket_world. The classifier had done its job. But the table beneath it was empty. No title, no source, no article type, no one-line summary, no author stance, no writing purpose. Most striking of all, the list of information points was entirely blank. And beneath that emptiness sat a note that stopped me: entity identification was supposed to be done "from the information points", but the points did not exist.

This is more than a technical glitch. It is the empty page of that old scorebook — and I learned forty years ago that an empty page is never "zero". It is either rain, or abandoned overs, or a delivery I could not understand how to record. On the night of the 2026 Eden Gardens semifinal my sheet developed the same kind of blankness — 120 for 8 in 34.1 overs, and then the crowd. I logged the stoppage minutes, the over where the mood broke, but my sheet had no column that could write "unfinished match". The scorebook does not argue; it waits for the dust to settle. But before the dust settles, someone has to decide what gets recorded and what gets left out.

Consider the three plausible causes behind that empty table. First, the content-extraction stage received an empty body — paywall, a JavaScript-rendered page, an anti-scraping wall. Second, the body arrived but the paragraph selector failed and returned only the page shell. Third — and this is the one that matters most — the extraction model's output was truncated or failed to parse, and the system silently returned an empty list. That third option is the trap, because an empty result travels downstream as though it were a valid one. In Indian cricket journalism I have watched this happen, and not only in software: as the beat shifted from print desks to rights-deal reporting, Ranji coverage quietly thinned, and no column ever noted what was being lost.

The rule is simple: an empty result should never be passed downstream as a valid output. If the domain label has been found but the information points are zero, the pipeline should return "incomplete input", not success — exactly as a scorer does not write "N/A" when he fails to record a ball, but writes the reason in the margin. In cricket analysis this distinction is decisive. When I sit in Bangalore reading an xG table or a net-run-rate ledger, my first question is not about the number but about the source. Which match log produced this figure? Which over-by-over sheet? Which dated report? A number that cannot be cited is not analysis; it is opinion. And I keep a separate column for opinion.

One fact belongs here that often gets left out. Cricket's biggest data problem is not scarcity — after the IPL and BCCI revenue expansion, measurable events have never been more abundant. The problem is that the scope of measurement has not spread across domestic first-class and women's cricket as widely as broadcast language suggests. Since 2026 the WPL's broadcast promises sit in plain view, and they should be audited after every season the way I audited Roger Binny's first spells in 2026-84 — 2.1 an over in the opening spell against an overall economy of 3.4. That gap tells you what the number is measuring. When someone says a bowler's "average" is poor, my first question is: the average of which phase? Under which field? Against which batter?

From here comes the counter-intuitive angle every analysis needs but which is often hidden under the label of "data". Correlation is not causation. A batter's average and a team's wins can rise in the same season without the average having caused the wins. It is the error made daily in commentary boxes — turning a fine innings into "the reason" for a result when the scorecard shows it arrived on the fourth day with the match already settled. My old notebook taught me: count the ordinary before explaining the miracle. Until the ordinary spells, the ordinary catches, the ordinary run rates are recorded, the celebration of an extraordinary innings is only a story.

That is why the document with the empty table unsettled me. It was not merely saying that data was missing; it was not saying what had been lost. Good analysis is not the one that displays the most numbers — it is the one that states clearly which number is absent and what that absence means. When a report says "this season's figures", one should ask: how many deliveries were recorded this season? How many matches? Which platform? Which database? Without those answers, the number is decoration.

For some years I have tracked the part of Indian cricket that broadcast language calls destiny — net run rate, review counts, retention caps, auction purse arithmetic, and the number of domestic match-days a board announces versus actually schedules. The counter-intuitive discovery usually lives in the columns nobody reads on air. That is my margin ledger. Look at auction purse arithmetic and you find that large signing-on fees for free agents sidestep the financial scrutiny that applies to transfer fees. Transparency in sport is not only about how much was spent but under which heading the spending is hidden. The same logic applies to the cricket figure presented as an "average" whose sample is never shown.

I should be clear that I am drawing no conclusion about that pipeline or the organisation behind the document. All I have is the document in front of me, and it says only this: domain label present, everything else absent. An honest reading says the next stage of analysis cannot begin until minimum input arrives — either the original text, or a completed stage one with 3 to 10 information points, the author's stance, the article type (news / comment / match report / data analysis / interview), and a publication date. The demand is not unreasonable. It is the same demand I make of a scorecard: is this the first innings or the second? What was the pitch like? How many overs were lost? Without those, a score is only a number.

A long career has taught me something the data world often forgets. In 2026, when I took my first step at The Indian Express, Chandigarh, there was no data terminal. In June 2026 I listened to Kapil Dev's World Cup final on All India Radio, and that winter I talked my way into a part-time scorer's chair at Chinnaswamy. I did not know then that what I was doing would be called "data analysis". I only knew that Kapil's 175 at Tunbridge Wells existed as a single scorecard line, and that was a failure worth correcting. Forty years on, the same question returns — this time as an empty data table. Until we know what was not recorded, we cannot know what we saw.

So the next time a piece of cricket analysis tells you "the numbers say", ask the question I have asked from that old chair at Chinnaswamy: which scorebook did this number come from, and which page of that scorebook is blank? If the answer is silence, you are looking not at analysis but at dressed-up figures. And on the day that blank page is filled, you will find that the true story of the match was hiding there all along — in the margin, where the ink had thinned.