The Empty Cell: The Cricket Data No Scorecard Ever Records

This morning a spreadsheet opened on my screen — 6,214 rows, 41 columns, and eleven of those columns entirely blank. The only thing written anywhere was a single line: cricket_world. No headline, no source, no publication date, no one-sentence summary, not one information point. The pipeline behind it returned the name of a domain and left every other field empty.

I first saw a sheet like that in 2026. Back then I assumed something upstream had broken — a crawler returning null, an extraction job timing out. I know better now. The most honest scorecard in Indian cricket is the one with eleven blank columns in it. An empty cell is not a bug. An empty cell is the design.

I did not close the file. I printed it and wrote names onto it by hand — names that never made it onto a broadcast graphic, whose first-class record sits in a row somewhere in an Excel sheet nobody opens. This article is about those eleven columns.

---

Context: how this data actually gets made

Cricket is called the most recorded sport on earth. Every ball is counted, every run accounted for, every wicket timestamped. That sentence only holds for the cricket a camera is pointed at.

The domestic data stack in India rests on three links: a broadcast camera, a ball-by-ball logging operator, and the software that turns that log into a scorecard. Remove the first link and the other two fall on their own. The Ranji Trophy currently runs 38 teams and well over a hundred matches a season. Where there is no camera, there is no ball-by-ball data. Where there is no ball-by-ball data, there is no economy rate, no strike rate, no dot-ball percentage, no record of a bowler's length. What survives is the skeleton: runs, wickets, overs.

The consequence is that selectors and newsrooms end up making the same kind of decision in the same room. They watch the player everyone else is watching. This is not a conspiracy. It is a convenience system. Data that is easy to obtain becomes the basis of the decision, and data that does not exist is never missed.

It took me two full seasons to see this. Early on I thought the problem was data quality. It is actually a problem of the sample frame, not the sample size — we are measuring the wrong population and then arguing about the precision of our measurement.

---

Core analysis: three sheets, three lessons

Sheet one — 2026, Rashid Khan and the order of deliveries

In IPL 2026 Rashid Khan was 18 years old, bowling for Sunrisers Hyderabad, and finished the season with 17 wickets at 6.62 an over. I was an economics undergraduate in Ahmedabad. I spent six weeks hand-coding all 60 matches of that season from ball-by-ball logs into one file, chasing a single question about a single bowler.

Did Rashid's wicket come from the leg-break or the googly? The commentary box gave the same answer every time — flight, googly, batsman didn't pick it. Commentary watches the wicket ball. I started counting the three deliveries before the wicket ball and measuring their order: which ball was the leg-break, which was the googly, and in what sequence. I called it sequence entropy — how much uncertainty a bowler manufactures within a single over.

The finding was that the wicket was not coming from the leg-break or the googly. It was coming from the order. Rashid was bowling the unexpected delivery on the third ball, and the pitch was not doing the work. The wrist was.

I wrote 2,400 words on it. It drew 4,100 reads and one direct message from a three-person analytics start-up in Mumbai. I have never opened a spin analysis with a wicket since. I open with a delivery sequence and a number, and editors have learned to expect a spreadsheet attached to the copy.

Sheet two — 2026, value per crore, and the name nobody called

Nine months later I was the third employee of that start-up on ₹28,000 a month, building a value-per-crore model for the January 2026 IPL mega auction in Bengaluru. The logic was simple: convert three seasons of domestic and international output into one common unit, decay it by age, divide by the crore.

The sheet ranked an 18-year-old above three capped internationals. His name was Shubman Gill. My flag came from his 2026 Under-19 Asia Cup reports, where his shot-selection data existed. The franchise I was assigned to bought experience instead. Kolkata Knight Riders took Gill for ₹1.8 crore, and he made 203 runs in his first season.

My model was right. Nobody called.

That sheet sat on my hard drive for three years, waiting to be right. That same year I changed how I published auction work — two columns, model price and market price, plus one line naming who was overpaid. It made me quotable, and for two seasons quietly unhirable inside the franchises I had graded, which is how I ended up in independent consultancy rather than a staff desk.

On auction night I learned the thing that now governs every model I build: value per crore is a question, not a number. Auction night is where models go to be humbled by human beings, and I have come to like the lesson.

Sheet three — 2026, the empty stadium and a coefficient of 0.04

For IPL 2026 I was embedded with Kings XI Punjab in the UAE — 94 days inside a Dubai hotel bubble. The season was 60 matches and none of them had a crowd. I built a home-advantage coefficient weighing toss, pitch type, travel load and crowd density. In a normal season it sat at 0.31. In 2026 it fell to 0.04.

When the stadium emptied, home advantage left with the crowd.

But the real sheet that season was the super over. My simulation ranked our captain's chosen bowler fourth. We played two super overs — against Delhi Capitals on 20 September and against Mumbai Indians on 18 October. We lost both. We finished the league level on 12 points with two other teams, and net run rate decided who sat above whom.

I said nothing in the review meeting. What I would have said ran to 40 pages, in a memo nobody read, which I later posted anonymously. Since then one full sentence has lived in my writing, and I will stand behind it any day: the decision was correct and the outcome was not. Those are two separate rooms, and locking them into one is the most common error in this trade.

That season I also stopped treating the crowd as atmosphere and started treating it as a measurable variable. I logged the silence, and the model finally started telling the truth.

---

The contrarian angle: an empty cell is not a data failure

There is a trap running through all of this, and I have fallen into it. It is the assumption that missing data is a technical gap that a better crawler can fill. The truth runs the other way. An empty cell is not a failure of data production; it is a failure of data demand. We do not collect what the market does not buy. There is no log of day-four Ranji bowling because nobody asked for one — not the viewer, not the sponsor, not the selection committee on paper.

The second trap is mistaking the model for the ground. My value-per-crore sheet does not know the moisture in the pitch. It does not know whose action changed this year, whose eyes are going, which coach's trust factor applies to which player. A model is a claim, not a guarantee. An analyst who does not print his model's blind spots is, functionally, a PR agent.

The third trap is reading correlation as cause. Rashid's sequence entropy moved with his wickets, but that does not make sequence the cause of the wicket. It is possible the sequence existed because the batsman was already under pressure — meaning the real variable was the wicket, not the order. I still cannot say this with full confidence. The day I can is the day I should stop writing.

The fourth trap is the most expensive: testimony as decoration. A scout at that start-up once told me, "A domestic batsman's average doesn't tell you his price, it tells you his flat pitch." It sounds good, which is why it is dangerous. I dated the quote, named his incentive — he was selling scouting reports — and tested it against the record. He was right in some cases and wrong in others. A scout's line printed unchecked is not my reporting. It is his sales copy.

And the cheapest trap in domestic cricket: an average built on six innings. My rule now is that the sample size appears inside the same sentence as the number. If the innings count is in single digits, say so before the adjective. The Ranji frame is long and the sample is thin, and hiding that ratio from the reader is a form of deception.

I trust a number most on the day it makes me uncomfortable. Printing the figure that cuts against your own thesis is the cheapest and rarest act in this profession.

---

Takeaway: what to watch next season

When I go back to that blank spreadsheet now, I do not see a broken pipeline. I see the domestic bowler whose six seasons sit in a row nobody has read. I found him in row 4,112, and nobody in the room knew the name.

Watch three signals next season. First: of the bowlers with the highest dot-ball percentage in the domestic season, how many get bought at auction — bought and picked are two different things. Second: for teams whose home-advantage coefficient stays below 0.20 once crowds return, their home record will be explained by some other variable, and that variable is usually death bowling. Third: eighteen months after the auction, work out each squad's output per crore. I will publish that table after a full season, not on auction night, and I will label the delay openly.

The question was never how many columns in the scorecard are filled. The question is who is going to read the empty ones next.