The Scorecard With a Blank Column: Asia's Real Cricket Data Problem

The scorecard had eleven names, two innings, one hundred, four wickets — everything was recorded. What was not recorded was the part from which a match is actually understood: the ball-by-ball file. No line, no length, no speed, no pitch map. One blank column, and behind that blank column, an entire season.

In 2026 I was logging deliveries for a London scoring vendor when an odds desk hired me as a junior analyst and handed me the IPL's death overs. Across sixty matches that season I coded all 1,140 legal balls bowled in overs 16 to 20 by length, slower-ball variation and batter intent. I kept the ledger because memory is a bad auditor — and today that same ledger tells me that the difference between where data exists and where it does not is often not a difference in the cricket, but a difference in the camera.

Asia's professional cricket data infrastructure is not uniform. It divides into clear tiers, each with its own rules and its own ceiling.

The first tier is franchise cricket — IPL, PSL, BPL, LPL, ILT20, SLC. Every ball is recorded, ball speed and pitch point are tracked, even field placement is logged. In 2026, when the IPL was played in the UAE — 19 September to 10 November, sixty matches across Dubai, Abu Dhabi and Sharjah, stadiums empty — this entire tier behaved like a controlled experiment, because everything was on file.

The second tier is international cricket. Ball-by-ball data exists, but tracking coverage is not present in every series. One series has Hawk-Eye; the next does not. One ground has a pitch map; the next has only a score.

The third tier is domestic red-ball — Ranji Trophy, Duleep Trophy, Irani Cup, Dhaka Premier League, Quaid-e-Azam Trophy. The scorecard exists; the ball-by-ball record has a large hole in it. Some matches yield a full ball-by-ball file, some only over-by-over, some only a scoreline. I will not put a percentage here, because I do not have the file that would earn that percentage — and printing decimals without a denominator is precisely the dishonesty my ledger was built to catch.

The fourth tier is associate and emerging cricket — Nepal, Oman, UAE, Malaysia, Hong Kong. Here the thing called data is often a spreadsheet at the end of a tournament, and not always that.

The architecture itself is not a scandal. It is a difference in money, priority and infrastructure. The scandal begins when the selection room, the auction table and the commentary box start looking at all four tiers through a single door.

The real question is not how much data exists. It is which players the data covers and which it does not — that is, coverage bias.

A simple example. In that 2026 ledger I tested one thing: in the death overs, which predicts the closing total better — economy rate or yorker frequency? Economy rate lost, because economy rate is itself an outcome, not a cause. Yorker frequency won. When Jasprit Bumrah landed four or more yorkers in an over, Mumbai Indians conceded 6.1 an over in those overs; otherwise 9.8.

Notice how I wrote the first of those numbers — 6.1 an over, from 84 balls. Attaching the denominator is a habit, and the habit was born of fear. Without a ball count, 6.1 becomes a statement; with a ball count, it becomes a limit. A third of the claims in cricket journalism go wrong at exactly that gap.

Now move the same question to domestic red-ball. What is a particular fast bowler's average in the sixth and seventh overs of a Ranji Trophy spell? The question cannot even be asked, because the denominator for it does not exist. Where there is no file, there is no finding — and where there is no finding, opinion forms. Opinion then circulates in the shape of data, except it is not data, only a sentence.

This is where the televised sample walks in. The selector watches the match that is on television, and television shows the match where its business is. A bowler who sends down thirty balls at a showpiece city ground therefore appears more often than a bowler who sends down three hundred at a ground nobody televises — appears, in the sense of being visible, not in the sense of being better. This is not a conspiracy. It is simply the base of the sample being displaced.

The empty stadiums of 2026 became the reverse proof of that displacement. When the grounds emptied, my home-field coefficient collapsed. I re-ran the 2026–2026 data as a control against the sixty matches of IPL 2026, and the model confessed that home advantage had been a crowd-and-travel variable, not a soil variable — the multiplier fell from 1.06 to 1.00. In the same season, dew pushed chasers to 34 wins in 60. That number is still in my sheet, and I no longer write a sentence about a home record without first asking whether the match was played in front of a crowd.

That experiment was possible only because every ball was on file. You cannot run the equivalent on empty Ranji grounds, because it was never recorded which over was bowled under which conditions. So the problem in Asian cricket is not too little data; it is unequal data — and unequal data is more dangerous than missing data, because it hands you confidence.

At the auction table the inequality shows up as price. A player's valuation is set below his actual skill and above his visible skill. A domestic season in which he scored quickly across four televised matches weighs four times as much as twenty untelevised ones, because the material for comparison exists on only one side. No conspiracy is required for this. Only the thing every market does: treat the most easily available information as the most reliable information.

In a betting market it becomes even more direct. If you hold fifteen years of international data and three years of domestic data, your model quietly overrates the international player — not because he has more experience, but because he has more balls on file. That is a bias, but it arrives dressed as a statistic. Written to three decimal places.

I have another example of this, one I stopped myself. In August and September 2026, India lost a five-Test series in England 4–1. That summer I logged every deliberate leave outside off stump by India's top three. A twelve-point gap appeared between the Trent Bridge win and the three defeats. Then I counted my confounders: five matches, one win, shifting light, an ageing Dukes ball. I spiked my own note. That five-Test sample remains unpublished; a hunch is not a finding.

I want to apply that standard to domestic data — but in the opposite direction. I am saying that where I have five matches, I stay quiet; therefore where a selection committee has twenty innings, it should stay quiet too. The trouble is that it does not, because there is no coverage column on its sheet forcing it to.

Now the part where I have to disagree with myself.

My first instinct says more data will produce better decisions. My experience says that is wrong.

When tracking data is added to domestic cricket, it is not added at random — it is added at the grounds where installing a camera is cheap, which is to say the big cities. The result is that you hold more data, distributed even more crookedly than before. The model speaks with greater confidence, and its confidence is less justified than it was. This is the trap I catch myself in repeatedly: treating the ledger as complete when the ledger is only the part that happened to get logged.

Second. This entire complaint of mine may simply be a complaint about my own convenience. The boring hypothesis — and I give it equal standing with the interesting one — is that selectors are doing their job well. They are at the ground. They see the change in a bowler's pace, the change in a batter's trigger movement. I do not have that. I have only the log. Assuming that what cannot be measured does not exist is the oldest arrogance of the measuring method.

Third, and this one is technical. When we build models for domestic red-ball cricket, we routinely import England or Australian county workload frameworks — balls bowled, spell length, recovery days. That model was calibrated on cold, slow, soft grounds. Running it in Chennai humidity or Multan heat is using an instrument whose number answered a different question. I have the mirror case too: years of logging venue-by-venue results in Indian domestic and franchise cricket produced a home-field coefficient that behaved nothing like the pitch lore, and when the stadiums emptied it had to confess.

All three arrive at the same place: the question is not how much data exists. It is whether the data is published with its coverage column and its ecosystem label attached.

So what next? I have no prediction, only a proposal. The Asian Cricket Council and the relevant boards should publish a public coverage index — for every domestic tournament, every season, what percentage of deliveries has a complete ball-by-ball record. I am not betting on the index being accurate; I am betting on it being published. The day it appears, every domestic average will show its true base beneath it, and some claims will collapse on their own.

Meanwhile, next season, when you watch an emerging fast bowler, ask one question — across how many balls was that pace measured, and how many of those balls were on television? If there is no answer, the absence of an answer is the answer.

A correction is not a defeat; it is the first honest line of the next model. And right now the only correction I have is this: I would rather show that scorecard's blank column as blank than dress it up as filled.