When the Ledger Comes Back Empty: The Indian Cricket Data Nobody Ever Counted

The file arrived with seventeen fields. All seventeen said the same thing — N/A. Article title: N/A. Source: N/A. Type: unclassified. One-sentence summary: blank. Information points: none. Entities involved: "identify from the information points above" — while there were no points above. My first reaction was not that the pipeline had broken. My first reaction was that this is exactly what a Ranji Trophy scorecard looks like in March. Balls recorded, overs recorded, runs recorded, and everything that actually happened on that field sitting outside the box.

I did not delete the file. I put it in a folder called 'Incomplete'. There are thirteen files in that folder now. Nine of them contain the same seventeen N/As. This is not a technical failure. It is an occupational habit — and the habit is that we count only what the scorecard script was built to count.

Context: the boundary line of counting

Cricket is the most counted game in the world. A Test match contains roughly 2,400 deliveries and each one produces a recorded entry: who bowled, to whom, in which over, for how many runs, with what outcome. No other sport has that density. Football gives you ball-by-ball only if you build it yourself; basketball gives you possession-level data in a limited frame; tennis gives you every point but with very little context. Cricket's advantage is that every unit is discrete, countable, and joinable to the next.

That advantage has produced a delusion. The delusion is that whatever can be counted is being counted. In fact the recording system is old and its architecture was designed by people who had no professional need to know where the slip fielder stood. So we have a dataset that tells us how many runs came off a ball, and never tells us who was standing at slip, which fielder was thirty yards away, or how much of that over's effort the bowler had already spent in the previous four.

The most useful thing I have learned in eleven years is that cricket's real data lives where the broadcast camera does not go. The commentary box is the last room the data has not entered. The panel tells you a bowler has 'found his rhythm' — I go and look at his release point across the last four overs. The panel says a team is 'under pressure' — I look at what the required rate was across the previous six overs and how many wickets were left. Half the time the panel is right. The other half it is describing what the eye can hear.

This piece starts with that empty file and ends in three places: a ground in Pune, a laptop in Adelaide, and a Russian newsroom where I was made to stop covering cricket and build an expected-goals model.

1,247 deliveries, and the pattern that arrived later

February 2026. I was freelancing from a one-room flat in Kothrud, thirty-three years old, with no permanent media accreditation. For that Test at the MCA Stadium in Pune I bought my own press pass. India were bowled out for 105 and then 107. Australia won by 333 runs. Steve O'Keefe took 12 for 70. Steve Smith made 109.

The next morning every report carried the same word: rank turner. Batting was impossible on that pitch. The ball was turning three times a day. I read those pieces and said nothing, because all I had at that point was a scorecard from two innings, and a scorecard does not tell you whether the ball turned or the batsman stood on the wrong line.

So I started counting by hand. All 1,247 deliveries of that match, one by one, from release to the keeper. I filed each ball into three columns — line, length, and what it would have done if the batsman had not played a shot. It took three weeks.

When the count was finished, the pattern that emerged was not in any report: fourteen of India's twenty wickets fell to deliveries that would have hit the stumps on any surface — straight, or not turning enough, with the batsman coming inside the line. Only six wickets could reasonably be attributed to the pitch. And the turn itself was concentrated in two sessions: the last hour of day one and the first hour of day three. The rest of the time the pitch was just a pitch.

This does not mean it was not a turner. It means it was not a whole-match turner; it was a two-session turner, and the rest of the match belonged to batsmen's line choices. That distinction vanishes in the commentary box, because the commentary box connects wickets to pitches, and connecting wickets to pitches is not always true.

That year I also noted something nobody wrote: how many balls O'Keefe bowled that were simply left alone. That figure is in no scorecard. It exists only for the person who sat down and counted.

128 deliveries, four days of waiting, and a silence

December 2026. Adelaide. I was on my balcony in Pune at 4:30 in the morning with a laptop. No press box, no crowd, no travel. India made 244 and then 36. Australia won by eight wickets.

That night several colleagues filed within an hour. I filed nothing for four days. It was not a matter of principle. It was that I had no count to offer, and without a count I could only repeat what the screen was already showing.

So I coded all 128 deliveries of the second innings. Each wicket was classified three ways — mode (caught, bowled, lbw, run out), line (stumps, fourth stump, fifth stump, wide), and ball count (how many deliveries that batsman had faced).

The result: nine wickets for 27 runs inside 12.1 overs, and six of those nine to deliveries on a fourth-stump line. Fourth stump is the line where the batsman must decide — play or leave — and a one-second delay in deciding becomes a wicket. This was not the story of batting suddenly collapsing. It was a line that Josh Hazlewood and Pat Cummins kept hitting and a line that Indian batsmen kept postponing judgement on.

The most interesting number in my ledger was not the wickets. It was the balls that were left alone. In the first four overs Indian batsmen left eleven deliveries. Across the next six overs that number fell to three. As wickets tumbled, batsmen stopped leaving — and a line that could have been survived by shouldering arms became wickets by playing at it.

Here I have to file an objection against my own method. In my coding every wicket is a fault — the ball did something, the batsman did something, the result was a wicket. But what happened in Adelaide that morning was larger than tactics: fourteen hours earlier a team had been bowled out for 244, the captain was flying home the next day, and eleven men walked out carrying that news. The ledger records the ball; it cannot record the silence. I took four days because I knew what I had was half a truth, and printing half a truth as a whole truth is the oldest habit in my profession.

Russia, 2026: where I had to go elsewhere to learn my own trade

June 2026. The whole newsroom was on the World Cup and my cricket desk went dark. I was reassigned to football, a sport I had never covered. I built an expected-goals model on StatsBomb's open dataset across all 64 matches.

The model rated Croatia's 2-1 semi-final win over England a 48/52 coin flip. Croatia had scored in the 109th minute and had dominated the whole of extra time. My model was calling it roughly even.

It took three weeks to understand that the model was not wrong — the model was incomplete. xG without game state is a lie. A team leading 1-0 deliberately sits deeper after the sixtieth minute, and its shot quality falls, but that fall is its strategy, not its weakness. I rebuilt the model with game-state splits.

I am not telling this story to write about football. I am telling it because Russia taught me the thing nobody in cricket was teaching me: without context, every statistic starts to believe it is the whole truth. Cricket does not even have the excuse of missing game state — required rate, wickets in hand, overs remaining, all of it is printed above the scorecard. And still we publish averages as if match situation did not exist.

The Ranji ledger: the account nobody reconciles

My career formed in the era when the IPL auction became a national event. That era also produced its inverse story, and the inverse story is the Ranji Trophy — the season where scouting data is thin, coverage is a scorecard wire, and the actual selection decisions get made.

I read IPL retention lists backwards. I look first at which domestic performances were paid for, then at which domestic performances were never measured at all. The pattern across recent seasons is plain: a thousand Ranji runs and a thousand Ranji runs are not the same thousand runs, because one was made in the elite group and the other in the plate group — and the scorecard prints both in the same font.

An honest example is hard to supply, because group-wise breakdowns are not published in one place. What is available is this: take every match a team played in a season and average it out, and you get a number that no selection meeting has ever used. What a selection meeting actually needs is a different question — how often did that batsman face the new ball against pace that season, and how often did he rebuild against spin. That count is still not printed anywhere.

The advantage of reading a retention list backwards is that you start seeing what was never measured. When a player is retained, there is a measurement behind it — usually his own short IPL record, sometimes international form. When a player is released, there is a measurement behind that too. But the player who never appears on any list has no measurement behind him, and that is where Indian domestic cricket's deepest data blindness sits.

There is a second side to this that has to be written down: there is no reliable public attendance record for domestic cricket in India. When I say India is the last country where Test matches fill grounds, a good editor will ask me how many, at which ground, on which day. I often do not have the answer. A story repeated often can still be true, but repetition is not evidence.

The milestone versus the match situation

Fifty-over cricket has an odd fixation on the hundred, and that fixation says less about cricket than about how we read scorecards. A batsman is 88 off 47. His team needs 118 in 12 overs with five wickets in hand. The commentary box is discussing whether he reaches three figures. It is not discussing how he should bat the next six overs.

For me an innings is priced by three things: the required rate at that moment, the wickets available, and the phase. A hundred is not a fourth input. A century can be the outcome of winning a match, and it can be the outcome of losing one — and the scorecard calls both by the same name.

I am not arguing the hundred does not matter. I am arguing that if an innings is discussed only in terms of how much it made and never in terms of what it won, we have filed it in the wrong drawer.

Contrarian: what the ledger cannot see

Now the part I have to write in every piece, because I cannot leave it out and leaving it out would be dishonest.

My ledger can tell you six of those Adelaide wickets fell to fourth-stump deliveries. It cannot tell you what was moving through eleven heads that morning. It cannot tell you whose ankle was sore. It cannot tell you how hard a spinner's fourteenth over of the day cost him in the heat. Pressure, fatigue and fear are recorded in my dataset as zero, and zero does not mean nothing happened — zero means nothing was recorded.

There is a second problem I will declare before anyone else does: sample size. My Pune count is one match. One match means one. If the same team had read the same pitch differently the following week, my pattern would have been that match's pattern, not the pitch's identity. So whenever I publish a number, I publish my error rate beside it. It looks strange. It is not strange — it is the honest weight of the claim I am making.

The third problem is the largest. A man who reads ledgers for a living starts to believe he is the equal of the ledger. That is the moment the accounting becomes the whole story and cricket becomes a metrics map. I have spent a decade counting what the commentary box gets wrong, and I know how easy that itch is to scratch. But the audit is aimed at the claim, not the claimant. If I do not state the panel's reasoning fairly and then let the count do the embarrassing work, I am not analysing — I am winning a trophy for myself.

And one last objection I have filed against myself. On Test cricket in India I have a sweet exception: 'India is the last country where Tests sell out.' I like hearing it. But if I have not checked the attendance and the broadcast numbers, I have failed to apply my own test to a different part of the match. The same logic obliges me to write the Ranji irrelevance story, and I have to write that too.

Takeaway: the next time you open a scorecard

What came out of that empty file is a rule I wrote in one paragraph and pinned to the wall. The rule: before you write a number, decide which question it settles. If it settles no question, it may be elegant, it may be rare, but it does not belong in this paragraph.

Next season I will try to count three things I have never properly counted. One: the deliveries batsmen leave alone — a left ball is a statement, and it has no public ledger. Two: how many overs a bowler bowled consecutively in a single session of a domestic match, because workload has no scorecard. Three, and this is the hardest: the matches I watched during the day and never wrote about. There are thirteen files in my incomplete folder and nine of them are empty.

I do not know which file comes back empty next. I do know that hundreds of domestic matches will be played and finished and their data will return to the same seventeen blank fields, unless somebody sits down to count. The match ends; the data keeps playing in the archive — and the only question is whether anyone ever bowled to it.