Zero Rows, Loud Rooms: The Quiet Failure Mode Inside Cricket's Data Desks

Two-forty in the morning in a Chennai office with eleven people in it. One screen was on, the death-overs table loaded, the filter set — overs sixteen to twenty, ranked by a pressure index. The result came back as zero rows.

The first reading was that no entry in that match satisfied the condition. The second reading took forty minutes: the feed had died forty minutes earlier.

In cricket analytics the expensive mistake is not the wrong number. A wrong number gets someone out of their chair, scorecard open, asking questions. A missing row gets nobody up. It becomes a blank space, and the blank space gets filled by memory. Cricket's memory is milestone-shaped — "his fiftieth hundred, in his two-hundredth match" — and that sentence is the database's oldest enemy, because a count of hundreds does not tell you what happens in the next innings. It tells you how many opportunities a batter got, at which position, in which era.

Zero is an output, and this profession has spent decades treating it as a finding.

A report landed on my desk some weeks ago with exactly this shape. Title field: N/A. Source field: N/A. Format: unclassified. Information points: an empty list. Twenty-one pages in which every analytical position carried the same sentence — insufficient information.

That document did nothing wrong. It did what an honest model is supposed to do, which is refuse to answer a question it cannot see. The failure sat one layer above it: the pipeline that was supposed to deliver the raw material came back empty-handed, and no alarm fired at any stage.

I have watched matches for years and read scorecards for longer, and the structural weakness of Indian sports journalism is not censorship. It is that nobody leaves a hole empty. Where data is absent, atmosphere is added, and atmosphere always speaks in the language of the past.

Seventeen years ago I left broadcast production for an eleven-person digital sports startup in Chennai that had no data desk. I was thirty-four. Things start from the other end there: blank sheet first, questions second, argument with the scorecard third.

Through IPL 2026 I logged all sixty matches ball by ball. Fourteen thousand four hundred deliveries, four columns per ball: over, the batting side's wicket state, the required rate at that moment, the batter's position. Those four columns produced one metric — death-overs exposure — that weighted required rate against wickets in hand. The formula was simple. The claim was not: pressure is not an experience, it is an arithmetic quantity, and it can be counted the way a bowler's economy is counted.

The final that season was played in Hyderabad, and Mumbai Indians beat Rising Pune Supergiant by one run. Commentary kept the one-run margin, because a one-run story is easy to write. The database said something else.

That one-run gap had been set forty overs earlier. The playoff side handed thirty-eight percent of its death-overs balls to its numbers six and seven, while its top three faced fewer pressure deliveries than the top three of any other qualifier. The most experienced batters batted under the least pressure; the least experienced absorbed the most. The one run was interest on that gap.

Nine hundred thousand people read that piece. The franchise's analyst called — to argue first, then to hire. And in a thirty-seat press box at Chepauk that season I was the only woman.

Both things come from the same place. In that box I learned that empty tables make loud enemies — the table with no food on it is always the loudest room. And the database started empty too; the first real finding arrived as a fight, on a phone call, not in a meeting.

The argument that won that year was not technical. It was that calling a team's number six a "finisher" is a label, not a measurement. A decade on, the fight is no longer about whether the numbers are right. It is about who controls the story at the moment of delivery. The data desks have won the tactical argument; the commentary box is the last holdout, and it holds out under real constraints rather than bad faith.

Here is the turn. Indian cricket's data volume has tripled in two years and the quality of selection argument has barely moved, because every pipeline we build is designed to catch wrong data and none of them catches absent data.

Wrong data shows itself. An average that jumps to a hundred and twenty will stop somebody. But when a match was never entered, the average still looks clean. It does not lie; it goes quiet about a performance, and silence carries no tag.

The second failure is format ambiguity. "Good" is not a universal word in cricket; it is a format-relative number. A strike rate of 130 is a composed innings in T20 and a different species in a Test. An economy under seven is strong in T20; the same bowler will not hold a long spell in a red-ball game. A report without a format label is standing all its numbers on borrowed ground.

This is why DLS and net run rate are the first things I test. Rain-revised targets and tie-breaker NRR both manufacture numbers that mean nothing without context. One missing match can shift NRR far enough to change a league table's conclusion. A missing row does not just break analysis; it breaks the logic of a ranking.

The third failure is source blindness. A board release, an agent's leak, a fan account's screenshot and a senior reporter's tweet look identical in a database unless publisher, author and timestamp are recorded. Without a timestamp, a claim stays true forever — and the claims that live longest in Indian cricket are usually built on a team structure four years out of date.

The fourth, quietest failure is entity loss. A marquee name, a board decision, a suspension, an NOC dispute can pass through a pipeline without a sound. Sports journalism runs the opposite rule: what is not visible is assumed not to have happened.

The Ranji Trophy is the standing proof. The IPL's real casualty was not Test cricket — India remains the last country where Tests sell out. What lost its audience and its coverage infrastructure was the domestic first-class pyramid, and its data is just as uneven. Forty wickets in the Plate Group and forty in the Elite Group look identical on a scorecard while the batting they came against belongs to two different worlds.

Women's domestic cricket is thinner still. Whole seasons go without a ball-by-ball record. Half the game we claim to model has no sample to model from.

So say it plainly, against a decade of pressure to have a number for everything: some questions cannot be answered by data, and building the list of those questions is as much of the job as filing the report.

The hardest team to make in world sport is India's second XI, and that is a selection problem rather than a talent problem. Three institutional distortions do the work. Franchise scouting now performs the zonal selector's function, so the geography of selection has moved. The Impact Player rule warps role definitions, because a batter is no longer a number six but a second-innings specialist, and his all-round value drops out of the arithmetic. And decisions still get made on reputations the current data no longer supports.

In this transfer window the same problem wears new clothes. The conversation is about retention, and retention's real story is purse arithmetic, not names. Right to Match, the remaining auction purse and contract length — the gap between those three tells you whether a franchise built a strategy or chased the loudest voice in the market.

Numbers arriving from an agent are usually the next move in a negotiation, not the deal. The transfer market speaks in fees; the spreadsheet asks who pays later. The figure in the headline is gross; the effect is the net number left after purse, quota and contract duration.

Readers need a filter they can run themselves. Three tiers. A board or franchise statement: a fact, even when incomplete. An agent-sourced or management-sourced claim: a state of conversation, true only as long as the bargaining holds. An aggregator or social echo: volume, not information. A story where all three sit in the same paragraph reads best and teaches least.

NOC arguments belong in the same window. A board-issued No Objection Certificate and central-contract terms now decide a player's availability, and that is the thinnest layer of data we have. League numbers and international numbers are recorded separately, built on separate samples, then added together at selection time as if they came from the same vessel.

The objection arrives immediately: garbage in, garbage out is nineteen-sixties computer science, what is new? What is new is that software engineering solved wrong data with validation, checksums and range checks, and never solved missing data inside an editorial workflow, because no engineer sits there. Sports desks assume data is raw material that simply arrives. It is a supply chain, and supply chains break quietly every week.

The second objection concerns the commentary box, and I have to hold myself back there. Calling the box obsolete is cheap. Its limits are real: live, no second feed, one ball to react. Under those three conditions running toward the milestone is rational, because the viewer already knows it and it can be said in a single ball. Argue with the argument, credit the person.

Live xG taught me to watch the second screen and the room — what the first screen says, and what the people in the room are moving toward. The cricket box rewards the same reading. Give it a tool that shows, by the third over, how a pitch's middle-overs run flow is turning toward the death, and phase-based commentary arrives on its own. The problem is equipment, not talent.

The third objection applies to me directly. My first real finding arrived as a fight, and that origin trains you to treat consensus as proof of value. The fix is to state the base rate before overturning it.

Not every empty result is a failure. Sometimes zero rows means nothing qualified. The distinction is between "no result" and "no data", and it cannot be seen without metadata: last-updated time, expected row count, entity count. A dashboard without those three columns cannot tell you whether the zero is a fact or an accident.

A model is only as honest as the questions it refuses. My model no longer promises an answer to every question. When entity count is zero, when the format label is missing, when the source has no name — the output should go empty. That is discipline, not weakness. The lesson from a database built from nothing was never that an empty start is shameful. The shame is pasting an old story into an empty space.

Sixty matches, one feed, and a growing list of doubts. The list is still growing, and it is no longer only about numbers. It is about the three checks we skip while telling the story of a game: the count, the format, the time.

Three signals to track in this window. Watch whether every central-contract and NOC decision ships with an availability calendar or only a headline. Watch whether retention announcements publish the purse accounting or only the list of names, because the latter makes the next trade impossible to forecast. Watch whether phase-level domestic and women's domestic data goes public; the day all three become routine, selection arguments move from reputation toward measurement.

Until then, one habit. The next time a scoreboard says nothing notable happened, ask when the feed was last alive. If the answer is that nobody knows, assume a milestone will be pasted into that space by morning, and that it will feel as true as the oldest memory in the room.