No Camera, No Coordinates: Mapping the Record Famine in Indian Cricket

14 June 2026. The third tier of M. Chinnaswamy Stadium. A seating chart and a stopwatch in hand. India had declared on 474; Afghanistan folded for 109 and 103 across two innings, and the match was over inside two days by an innings and 262 runs. The scorecard said history; my notebook said humidity, footfalls, and patience.

That notebook holds 204 deliveries from Rashid Khan. Sixty-one of them were googlies. And of those 204, only 14 landed on a line that genuinely threatened a right-hander's stumps. The legend was larger than the landing zone. I could write this because every ball of that Test had a camera on it, a public ball-by-ball log, and all I had to do was sit and count.

That same week, somewhere else in the country, a domestic match was played for which nobody thought a camera was necessary. That match has no pitch map. No release height. No venue baseline against which anyone could say whether 240 was a lot or a little that day. And that is my real concern — the pool from which Indian cricket selects is played on grounds where even the first level of record is incomplete.

Three levels of record, and the absence of the third

In cricket, "record" is not one word. It is three different things, and they need to be separated.

Level one — scoring. Runs, wickets, overs, extras. This exists almost everywhere, in every tournament, at every ground. This level is like law: without it, the match has no existence.

Level two — logging. Which ball to which batter, in which over, with what outcome. This level is tied to broadcast — where the commentary team sits, an extra row of scorer sits with it. It is not entirely missing from domestic tournaments, but it is uneven: some matches are streamed, the rest remain a scorecard.

Level three — mapping. Where the ball pitched, how high the release point was, which way the seam was, how late the batter's trigger movement came. This level exists only where ball-tracking cameras are installed — that is, international cricket, the big leagues, and a handful of domestic finals.

This three-tier structure matters because the Indian selection system largely runs on the first level, leans on the second, and never reaches the third — even though the third is the language the game is actually spoken in.

18 June 2026, The Oval. Pakistan made 338 for 4; India were bowled out for 158 in 30.3 overs, a 180-run defeat. I was seventeen, a first-year science undergraduate in Bangalore, watching a 2 a.m. stream. I downloaded the match file and froze every one of Mohammad Amir's 96 deliveries frame by frame, measuring lateral deviation off the seam. The ball that dismissed Virat Kohli, caught behind for 5, moved 4.3 centimetres off the seam at 141 kph. I filled 34 notebook pages and missed a Monday lecture. The final was decided in a space smaller than a thumbprint — and that space was visible only because the ICC had three cameras and a frame rate that night.

A comparison here, and by my own rule, only one. English county cricket is not a direct yardstick for the Indian domestic structure, but it offers one lesson: wherever coverage was built beneath the broadcast contract, the language of selection debate changed within two seasons. In India that change has so far arrived only at franchise level, where the people sitting at the auction table independently fund their own scouting departments.

Between 19 September and 10 November 2026, sixty IPL matches were played in front of empty stands in the UAE, and Mumbai Indians beat Delhi Capitals by five wickets in Dubai. I was twenty, midway through a master's, watching every ball on one screen. Without crowd noise the stump mic became a data source: I transcribed 214 field changes called out loud, 63 percent of them between overs 7 and 15, and hand-coded 13,600 legal deliveries, tracking dew, toss and second-innings scoring rates. An empty stadium has a different kind of noise: the hum of what should be there.

Which is not to say data only comes from cameras. It is to say data comes when someone is paid to sit and listen. Nobody is hired to transcribe field changes in a Plate Group Ranji match. The record is missing there not because the cricket is missing, but because the log-keeper is.

The scorecard is a compression algorithm

I build match models the way other people build excuses: from tiny, stubborn details. And every model I have built has run into the same wall — the wall of the scorecard.

Take a bowler's figures: 4 overs, 0 maidens, 28 runs, 1 wicket. That is a sentence that knows nothing about the match. It does not record where the yorker landed, whether the slower ball had been picked early, whether the batter's trigger movement was late or early, or whether the one wicket came from the quality of the delivery or the haste of the batter.

Take two bowlers with identical economy of 7.00. The first bowled his eight balls on back-of-a-length and two of them slipped onto middle. The second bowled four full and two full tosses that the batter mistimed. In the scorecard they are identical. To a coach they are opposites, and the coach will pick the first next game — because he has no record of the second one's full tosses.

The entire selection argument rests on a document that deliberately discards information. The scorecard never claimed to be a summary of the match; we added that claim ourselves, because it is cheap and available everywhere.

The second layer is that this compression costs most in domestic cricket, because domestic cricket is where the competition is densest. India's second XI is the hardest team in world sport to make — and precisely where the competition is densest, the evidence base is thinnest. That is not a paradox, it is the architecture of the system.

The real problem is the absence of a venue baseline

Much of my work is building venue baselines. Chinnaswamy's true bounce, Wankhede's sea air, Chepauk's turn — for all three I have years of international data, which is why I can say how far an innings' total deviated from that ground's rolling par. This is the discipline that lets me write the sentence I like most: the final was decided in a space smaller than a thumbprint.

For domestic grounds, that baseline does not exist. A five-wicket haul at Lahli, where the ball seams, and a five-wicket haul at Visakhapatnam, where the pitch is flat, enter the same column of the scorecard. Without par scores, without deviation, no signal emerges — only counting.

When you have no venue baseline, you inevitably fall back on raw numbers, and raw numbers always point to the player who got the most opportunities. That is arithmetic, not prejudice. The selector is not biased; he is sitting at a desk with one column on it.

One more thing must be added here, and it is the least-recorded part of my notebook. The most decisive data in domestic cricket is what nobody ever logged — who bowled how many overs in a row, who travelled how far in the last three weeks, whose pace dropped three kilometres an hour by the eighth week of the season. All of it exists, none of it is in a public table.

Women's cricket: the largest data desert

Since the Women's Premier League began, every match has been broadcast, and that is a genuine change. The five founding teams now hold three seasons of data that no Indian woman cricketer previously had — ball-by-ball, legal-delivery counts, phase-wise strike rates.

Directly behind that sits a gap. Pre-league domestic women's matches survive largely as scorecards. The consequence is straightforward: the players now doing their best work at 28 to 32 have career baselines built on a decade of records that contain only runs and averages.

Two labels have grown out of that gap, and I find both suspicious. The first: late bloomer. The second: inconsistent. Both come from the same deficiency — when you do not have 400 balls of a player's line and length across the last seven seasons, your only comparison is her average in the season that was best covered. That measures coverage, not form.

I am holding this as a hypothesis, not a conclusion — and it is cheap to test. Take one player's 2026-22 domestic season, rebuild it ball-by-ball from available video, and compare its strike rate with the strike rate the scorecard gives you. If the two differ by more than ten percent, our definition of inconsistency was wrong. If they don't, I will withdraw the point.

The auction: a scarcity system, not a talent ranking

Here I need to be plain, because sentiment flows freely around this subject.

The IPL auction does not evaluate talent. It runs a scarcity system: the purse, the retention maths, the overseas cap, and above all which roles are limited and which supply is infinite. When every team wants two finishers and the market holds twelve people carrying the finisher label, price is set by scarcity, not by ability.

And this is where the record famine starts setting prices. A confident 12-ball cameo in a televised playoff has been seen by everybody; the same player's 200 balls of consistency sit in untelevised domestic matches nobody watched. What was visible sold expensive. What was invisible sold cheap.

This is not an injustice, it is an arbitrage — and the franchise that builds its own logging department will earn directly from it. I say this without romance: scouting has become the de facto replacement for zonal selectors, and work that public money has not funded gets done with private money. That is not good news; it is an admission of a gap.

The real blind spot: the demand for more data

Now the part where I have to disagree with my own camp.

One note is loudest in this whole conversation — we need more data. I think the diagnosis is wrong. Our problem is not scarcity, it is misuse. We are measuring what we have incorrectly, and repeating that incorrect measurement at a faster tempo.

Three examples, all ordinary.

First, the impact score. Every league has produced a composite "impact" number that folds runs, strike rate, wickets and fielding into a single figure. The problem is not that such a number is bad. The problem is that most of them are published on 30-40 ball or 12-15 over samples, with no error bar. Show me two players within 0.3 of each other on impact whose real difference becomes obvious next season, and I will accept that the score explains a decision. I know what happened to xG in football; that model cannot tell you why a coach made a change in the 70th minute. Cricket's impact score is on the same road.

Second, matchup splits. "This batter strikes at 168 against this bowler." The sentence is incomplete. Over how many balls? At which venue? In which phase? Against which match state? When I add those questions, a great many 168s collapse into 168 off 34 balls — which is noise.

Third, and most serious: the body. Injury information stays inside medical confidentiality, and it should; I have no right to anyone's medical record. But the cold consequence is that the most decisive variable in a phase profile never enters the dataset. When I build a death-overs economy table, it has no column for how much sleep that bowler got in those six weeks or which bone was under load. And clubs disclose only what suits their stock — I have watched this across four seasons and never once seen it run the other way.

The most important stat was the one nobody had a column for.

None of which makes me anti-data. It means the sequence of the data project is wrong. We are adding ball-tracking first and leaving workload transparency to the end. That is like recording a patient's temperature every hour and never noting whether he has a fever.

Before the next match, one test

I do not write predictions without error bars, so here is the expectation, stated plainly.

Take six matches at one ground in the next domestic season. Before them, build that ground's rolling par score — the mean of its last twenty innings, with a band above and below. Then log each innings' deviation from that par. My claim is that this deviation will predict the results of those six matches better than the raw totals do. If the claim fails in four of six, my venue-baseline framework is useless for that ground and I will throw it out.

The second test is more interesting. Of the 214 field changes I transcribed in the empty stadiums of 2026, 63 percent fell between overs 7 and 15. That number may be a product of UAE dew and closed roofs rather than of the game. Next time a match is played without a crowd, count the same way from the stump mic. If the middle-overs share falls from 63 to around 45, then my 2026 finding was a finding about weather, not about field placement.

Complaining that the archive is empty is easy. Harder is measuring the archive we do have, honestly. Next time you open a domestic scorecard, look one line below it — whatever is not written there will be the basis of the next selection.