The Empty Sheet: The Indian Cricket Record That Was Never Logged

Hook

Eleven fields. Nine blank or N/A. The tenth carried an instruction — "identify entities from the information points above" — while the eleventh, the information points themselves, was an empty list. Title: N/A. Source: N/A. Author stance: N/A. Publication date: N/A. What survived was a single label: cricket_world. Nothing more. No sport, no competition, no timeframe, no claim.

I opened that sheet three times on the afternoon of 14 February 2026 in Bengaluru. Three times, the same emptiness. My first reaction was that the pipeline had failed — the scrape had broken, something had gone wrong upstream. The second reaction, which lasted longer, was that this emptiness is not an accident. It is the default condition of Indian cricket's record-keeping. I have been reading domestic scorecards for eleven years. Most of them look exactly like this: a whole match was played, and the part of it that was actually worth understanding was never written down anywhere.

Sixty-four matches, one hostel room, and a spreadsheet nobody will ever see. In 2026 I hand-coded all sixty-four World Cup matches because I did not trust black-box event feeds. Six years later another sheet sits empty in front of me. The difference is that this time the emptiness is not my fault. It belongs to a structure in which a match gets played, a report gets filed, and the number never gets recorded.

Context: what an "information point" actually is

Definition box — information point: A single verifiable atomic claim. It needs four elements: an entity (who, which team), a number or event, a time (date/season/over), and a source (scorecard, board statement, press conference, or my own coding sheet). If any one of those is missing, it is not a claim. It is a comment.

An empty information-point list never means nothing happened in the world. It means what happened did not take a shape that a machine or an analyst could pick up and carry forward. Cricket shows this gap more starkly than most sports, because cricket is probably the most counted game ever invented — and yet its largest portion sits uncounted.

Take an ordinary Monday. Day four of a Ranji Trophy match. Two hundred people in the ground, sixty of them schoolchildren and twelve local reporters. The day ends. The scorecard says: 224/6, 92 overs, one century, three fifties, two catches, one run-out. That is complete and correct information. It is equally complete and correct that this information cannot tell you how that batter survived the third spell, or at which over the spinner changed his line, or on which ball a fielder took two steps in the wrong direction.

There are three tiers of record:

| Tier | What is recorded | Availability in India | The question it answers | |---|---|---|---| | Tier 1 — scorecard | Runs, wickets, overs, boundaries, catches, strike rate, economy | Ranji, Vijay Hazare, Syed Mushtaq Ali, Duleep — near-complete, decades deep | "How much happened" | | Tier 2 — ball-by-ball | Line and length, shot type, field position, footwork | Complete in internationals and the IPL; patchy and fragmentary domestically | "How it happened" | | Tier 3 — tracking | Release point, seam position, backlift, fielder's starting position, sprint speed | Effectively zero in domestic cricket | "Why it happened" |

One example settles the difference. Tier 1 says a fast bowler took 3 for 34 in 18 overs at 4.1 an over. Tier 2 says 61 of his 108 deliveries landed on an off-stump line outside the batter's arc, and 22 hit the pad. Tier 3 says his release point was 2.18 metres in the first spell and 2.11 in the third — his action was dropping, and the wickets came anyway.

Without the third tier you can buy that bowler. You just cannot know what you are buying.

The BCCI has started live-streaming domestic matches over recent seasons, and that is a genuine improvement — at least the games are visible now. But visible and recorded are different things. A stream can be watched; it cannot be queried. I have hours of Ranji streams watched from Bengaluru, and I have no artefact of those hours that I could hand to anyone else. What I saw lives inside me. What I logged lives in my sheet, and my sheet is not shared.

Core: the evidence chain, link by link

The lesson of the codebook

In 2026, aged nineteen, in a hostel room in Bengaluru, I hand-coded all sixty-four matches of the World Cup. That came to 1,584 shots, each with xG, team PPDA, and distance covered. A twelve-page codebook followed, because the biggest question was not who won. It was why two different people should code the same shot the same way.

I re-checked 10% of entries — roughly 158 of them. The discrepancy rate came in under 5%. That is a personal observation, unreplicated, because the sheet was never published. What survived was the method: before quoting any number, write three things — sample size, competition context, definition.

Which is why an empty information-point list is not a technical failure to me. It is a familiar sight. That list came from a place where context exists, where an event exists, but where the number does not. And an event without a number is invisible to analysis.

The first note I wrote on the sheet read: "A pipeline failure and a cricket failure look identical. In both, you can read the headline and not the evidence."

The empty-stadium study: a benchmark that was never a benchmark

In 2026, aged twenty-one, I joined the lab study that grew out of the codebook. The comparison was clean: sixty IPL matches in the UAE in 2026 against sixty home-and-away matches in 2026. Three things were tracked — home-win rate, phase scoring, wicket probability.

Result: home-win rate fell from 55% to 47%, and death-overs scoring rose by 0.4 runs per over.

Those two numbers are easy to quote. They are hard to quote correctly. Sixty against sixty is a small sample. "Home" meant a home ground in 2026 and no ground at all in 2026. There were crowds in 2026 and none in 2026 — except crowd sizes varied match to match in 2026, and I did not have that figure for many of them. Travel fatigue, pitch familiarity, toss — all three are dissolved into that difference.

What I learned from that study was not about cricket. It was about the boundary of a claim. You can say: "Across sixty matches against sixty, we observed an eight percentage-point fall in home-win rate." You cannot say: "The absence of crowds destroyed home advantage." For that you need matches where crowd size is a controlled variable, and in the IPL it never will be.

I was the only woman on a five-person panel at that symposium, so I brought numbers instead of nerves. A senior coach asked whether I had "actually watched the games." I put the codebook on the table. The answer was not satisfying — no answer is satisfying when the question is not about your method but about your presence. From that moment, though, every piece I wrote carried a limits section.

The auction memo nobody asked for

In 2026, aged twenty-three, working as a junior analyst at a Bengaluru cricket data firm, I wrote an unsolicited memo before the IPL mega auction. The basis was 2026 death-overs data. One thing stood out: Indian seamers with death-overs economy under 8.5 and control percentage above 60 were going, on average, roughly ₹1.2 crore below value.

Definition box — control percentage: deliveries that landed in the batter's playable zone and from which the batter did not extract runs without undue risk — in my codebook, logged as "ball on intended line, outcome dot or controlled single." This is not an official statistic. It is my coding definition, and I attach it every time.

I sent that memo to ten franchises. None replied. My firm used it in a client briefing, and a digital editor saw the deck. Publicly the memo was ignored; internally it became precedent for auction valuation.

One clarification matters here, because sports journalism usually does the opposite: the auction did not prove my memo right. An auction is an event, not a validation. What matters is that the threshold was written down in advance — economy 8.5, control 60%, and a minimum death-overs condition. When the outcome arrived, the prior document already existed. Had the 8.5 bowler gone expensive that season, the document would still read the same. You judge a process by the process, not by the result.

The unlogged ledger

Here is where I want to write a number, and I will not, because I do not have it. Nobody publishes how many deliveries in Indian domestic cricket are bowled without a ball-by-ball record. That is the most important sentence in this piece — what is not counted does not show up as a shortfall either.

What I do have is my own sample, and its inclusion rule is strict:

| Element | Status in my sheet | |---|---| | Seasons covered | 2026 to 2026, incomplete | | Grounds covered | Only those I could attend — Bengaluru, Chennai, Mumbai, Rajkot, and two more | | Missed matches | Those I could not watch; no complete log of them | | Coding depth | Tier 2 (ball-by-ball), not Tier 3 | | Verification | 10% re-check per the personal codebook | | Permissible conclusions | Local patterns, indicators of player-level change; not estimates of national rates |

That table is my real identity, more than any finding. The stadium was empty. I stayed ninety minutes and counted anyway. But the value of the count is not that I stayed. It is that I wrote down which matches I could not see.

Format changes break datasets

The Duleep Trophy was restructured in 2026 from a zonal format to four teams — India A, B, C, D. Administratively that made sense. From a data standpoint it is a cut. Pre-2026 Duleep numbers and post-2026 Duleep numbers are two different products, because squad composition, competition standard, and selection logic all changed. Anyone averaging across the two is producing a wrong number — not deliberately, carelessly.

In the IPL, the Impact Player rule arrived from the 2026 season. It changed the role of death bowling, because there is now an extra specialist on the field and bowler spells are fragmented. Which means: death-overs economy before 2026 and death-overs economy after 2026 are not two values of the same variable. They are values from two different games. I corrected my models. How many others did, I do not know — and that is the actual problem.

A patch note is just a transfer window with better documentation. Cricket does not write patch notes. It changes the rules and quietly carries the old data forward.

There is one more cut, and no rule created it. The 2026-21 Ranji season was not played because of the pandemic. An entire first-class season is missing from every player's career. A player who emerged in 2026-20 and played in 2026-22 has a two-year hole that no form analysis can reach.

The argument that cannot be had

Ranji is India's premier first-class competition. Vijay Hazare is the 50-over format. Syed Mushtaq Ali is T20. Each means something different. Weighing a first-class spinner's 55% control rate against a T20 spinner's 55% control rate is like putting a fast bowler's sprint speed and a batter's sprint speed in the same table.

And that is what happens, because Tier 2 records do not exist. Where records are missing, the basis of selection shifts — onto three things: zonal identity, the coach's memory, and the strongest impression from the most recent match.

Wasim Jaffer is the highest run-scorer in Ranji Trophy history. That is a Tier 1 fact and fully verifiable — the runs can be pulled from any scorecard. But there is no route, from that same public record, to knowing how many of those runs came against third-spell spin, or in which season his cover drive changed. Without Tiers 2 and 3, we know the total and not the detail.

Sarfaraz Khan was picked for India's Test side after seasons of relentless scoring for Mumbai. Here is a case where Tier 1 numbers were so heavy that ignoring them became difficult — and that is the most telling part. A player being selected on the sheer weight of Tier 1 output means Tiers 2 and 3 were not available to argue with. The pile was the only advocacy.

That is the central finding: selection arguments in Indian domestic cricket degrade not through bad reasoning but through incomplete records. The absence of an argument is the biggest risk we keep inside the system.

Cheteshwar Pujara's durability shows up in a scorecard. The technical architecture of that durability — balls hitting the pad, backlift height, changing drive angles — is a Tier 3 question, and we have no domestic answer.

A threshold is only worth something if it is written first

My codebook's hardest rule: every evaluation states three things up front — the threshold, the rejection condition, and the sample limit.

| Component | My threshold | Why | |---|---|---| | Death-overs economy | Under 8.5 | Above that, wicket-taking becomes the excuse | | Control percentage | Above 60% | A bowler's real currency is control, not lethality | | Minimum death overs | 60 | No ranking on samples smaller than 20 overs | | Rejection condition | Under 60 overs, no rank — "insufficient data" | Avoid both ranking and zeroing |

The last row matters most. Faced with an empty sheet we reach for two bad reactions: treat it as zero, or call it unknown and end the argument. Neither produces a number. What works is: "This question is currently unanswered; write down what process would answer it."

Contrarian: the empty sheet was never the problem — assuming the sheet is the game is

What follows is against my own interest.

One: correlation is not causation, and data journalism's commonest disease is putting both in one sentence. In the 2026 study, death-overs scoring rose 0.4 runs per over. That does not establish that scoring rose because crowds were absent. It is equally possible that pitches differed in 2026, that bowling workloads differed, or that travel restrictions changed spell management. I quoted the 0.4. I did not claim its cause. That gap is the difference between professional work and promotion.

Two: more data will not improve selection. Structured and public data will. If Tier 3 tracking were installed at every Ranji ground tomorrow but stayed inside one franchise, selection arguments would get worse, because inequality would rise. A record's value is in its distribution, not its size. Until the inclusion rule is written, a bigger dataset is a bigger claim, not a bigger truth.

Three, and this is the uncomfortable one: I have reproducibility theatre of my own. I attach a methods paragraph to every piece. It reads rigorous. It is unreplicable, because my sheet is not shared. A methods paragraph is cheap to write; publishing the sheet is expensive. I have a twelve-page codebook and no downloadable version of it. Until I publish it, I have to label my claims "personal observation, unreplicated" — and I do, but every time I write it, I notice the formatting implying a verification that does not exist.

Four: the easiest move is turning the empty stadium into moral superiority. "The stadium was empty and I counted anyway" is my origin line and my most quoted one. Repeating it in every piece is easy. The moment the writing becomes about being unseen rather than about the cricket, it is over. The count should carry the moral weight, not the recounting of it.

Five: treating a threshold as physics. My 8.5 cut-off is a tool, not a verdict. The qualitative input my model cannot see is a coach saying, "His action changed after November." That is nowhere in my data, because I have no Tier 3. On what evidence would I defer? Two conditions: if the change appears in release-point data across three or more matches, or if the bowler's own workload log matches the change. Until then, the coach's eye is "unverified, but worth considering" — and I will label it exactly that rather than passing it along as a throwaway nets line.

Six, and this falls straight out of the empty sheet: read a zero as a zero, not as a mystery. When no report of a match turns up, do not assume a large hidden truth is sitting there. The likeliest explanation is that the scrape broke, or the piece did not exist, or whoever wrote it had no time to count. A transfer rumour is a story. A transfer ledger is a confession. We do not have the ledger. Do not hand us a story in its place.

Takeaway: the next-round signal

Three things can be watched over the rest of this season, and all three can be written down in advance.

One, whether an inclusion rule gets published. If the next table of domestic numbers does not state which grounds, which seasons and which matches were missed, it is not analysis. It is advertising.

Two, whether data gets re-based after format changes. The new Duleep structure and the Impact Player rule both cut through the spine of comparability. An analyst still valuing 2026 on 2026 death-overs numbers is borrowing a number, not earning one.

Three, whether a threshold goes public before an auction. My memo stayed internal. Next time I will write it outside, with a date on it, so that when the auction answers six months later, the question is not whether I was right. The question is when I wrote it down.

Before you quote the control percentage, ask who sat and counted. And when the answer is "nobody," write that down — because an empty sheet is also a finding. We have simply never learned to read it.