The Table That Looks Complete: Indian Cricket's Data Blackout and the False Selection Debate

Eight columns. All eight filled. Every cell had an answer. And every answer stopped at the same word — N/A.

Last week a report landed on my desk: an eight-dimension analysis of the cricket domain. Format and match reading, player technique and data, team landscape and ranking, league and commercial ecosystem, rules and governance, risk side, public narrative and expectation, industry transmission — all eight sections present, all eight standing in full structural dress. Title field: blank. Source field: blank. Article type: unclassified. Information points: zero.

In eleven years of reporting I have not seen a table that looked this complete and was this empty. This was not an article with thin information. This was an article whose information was never extracted at all — and the system passed it through quietly, because technically nothing was broken. The technical name for it is a schema-valid empty payload: correct skeleton, zero content. I read it as a map of the part of Indian cricket we do not want to look at.

Indian cricket's data is a pyramid, and the pyramid is upside down. At the top sits the IPL — eleven seasons from 2026 to 2026, roughly 1.1 million deliveries, and on every delivery the batter, the bowler, the line, the length, the field placement, the type of dismissal. I cleaned that dataset by hand in the lockdown summer of 2026 in Bangalore, aged 21, with the final semester of my degree running in the background. In the same stretch I pulled the event data from the Bundesliga's 81 behind-closed-doors matches to test home advantage. The home win rate fell from 43.2% to 33.3%. It was the first time I had a sample large enough to be wrong in public.

Below the pyramid sit the Ranji Trophy, the Vijay Hazare Trophy, the Syed Mushtaq Ali Trophy. There the ball-by-ball record is partial. Some matches yield a scorecard, some do not. Some yield a scorecard but not a ball-by-ball log. There is no record of which line the ball landed on, which shot the batter played, where the fielder stood. What is not logged does not enter the model. What does not enter the model does not enter the selection debate either. That is where the real problem begins.

Before I trust a table, I ask who assembled it and why. That report turned the question back on me in its own shape. It showed that when the publisher, the URL, the byline and the timestamp are all lost, source grading is closed off permanently. Text can be re-fetched. Provenance cannot. In journalism that is the most expensive gap there is.

Now the real lesson of that report, the one it did not write itself. Its most useful passage was not any cricket comment — it was its operating rule: make no claim on empty input. When information points are zero, nothing can be written about averages, strike rates, economy rates or rankings, because every figure must trace back to an information point. That discipline is something I need in my own work every day and do not always have.

It breaks most often when I write about the Ranji Trophy. Because the data there is half-present, and half-present data is the most dangerous kind — it looks like complete data. Take one season's average and you have a number that will support almost any argument. But the number does not tell you how many matches were logged and how many went missing. What looks like a trend is often a missing column in disguise.

So I state the sample size and the confidence before the finding, never after. The practice is dull to write and duller to read. Without it, every number becomes a claim nobody can check. A 40-ball phase split is technically computable — the tool sits open on my desk — but judging a bowler on 40 balls is like forecasting the weather from two overs. When the sample is too thin to carry the conclusion, refusing to build the model is itself a conclusion.

Let me clear the vocabulary, because the whole argument rests on it. A phase split means breaking an innings into three parts — powerplay, middle overs, death overs. A bowler's single economy figure is built by merging all three, and that is its greatest weakness. A bowler who goes at 7.2 in the powerplay and 11.4 at the death has an overall number that is good for nothing: it buries two different bowlers under one name.

Expected runs means: on that ball, on that line and length, with that field, how many runs are scored on average. Delta means the gap between runs conceded and runs expected. That gap tells you whether the bowler won or lost the situation, not whether a wicket fell. Match-up means a specific batter's record against a specific bowler, in the same format, in the same phase. These four terms are my native register, but they need saying once in plain language, because the audience that actually argues about this sits in regional broadcasts and YouTube comment sections.

Here is one fact that stands without any model at all: the IPL holds a record of roughly 1.1 million deliveries, and the domestic tournaments of the same country have not logged a tenth of that. The asymmetry is not technical. It is an investment. Where the broadcast money is, the camera is; where the camera is, the logger is; where the logger is, the model is. Ranji does not have a camera in every match. So it has no model. So its arguments run on memory and rumour instead of numbers.

This is where the bench-strength debate shows its real face. Every season ends with the question of who is ready from domestic cricket. The answer arrives as runs — averages, hundreds, wickets. But in what situation those runs were made is logged nowhere. Making 120 not out in the first innings with 400 on the board and making 60 chasing 150 in the fourth innings both land in the same column. The selector reads that column. The column will never tell him who stands up under pressure.

I am not saying the selectors are wrong. I am saying the table handed to them is incomplete, and a conclusion drawn from an incomplete table is not a selection error but a collection error. India's bench-strength debate is really a data-collection problem dressed up as a selection problem. I have been writing that sentence for three years and the evidence renews itself every season.

On the other side sits the auction — the most quoted and least audited number in Indian cricket. Every season prices are announced in crores and every price is a claim: this player is worth this much. How is the claim audited? By setting value-per-crore against phase-split output — what the player delivered per over in the powerplay, the middle and the death, and what was paid for it.

When I ran that exercise, the pattern that emerged was not in the prices but in the reasoning behind them. Franchises repeat the same mistake every cycle: treating age as a proxy for capacity. A 22-year-old sells high because he might grow; a 31-year-old sells low because he might decline. On paper the logic is clean. What the model cannot measure is dressing-room chemistry: who makes the other ten better, who lightens the captain's workload in the fourth over, who calls the meeting after a loss. Transfer-market models overrate youth potential and underrate dressing-room chemistry — because the first is a column and the second has not yet become one.

A model is a story with its seams showing; I map the seams. When a player's price does not match his phase-split output, there are two possibilities: the model is incomplete, or the franchise knows something the model does not. In both cases the correct response is the same — read the price as a question, not an answer.

Now the least-written part: the matches that never finished. Rain-washed innings, DLS-adjusted targets, qualification resting on net run rate, the empty-stadium fixtures of the COVID years. In the table these are a single row. That row changes someone's fate. DLS is a formula, and every formula stands on an assumption — about how the rate of scoring shifts with wickets lost. The assumption is built on averages, and the average counts matches that were never played. The ghost games left a paper trail; the archive kept it breathing.

I learned that lesson the expensive way in 2026, from the 81 behind-closed-doors matches and the fall in home wins from 43.2% to 33.3%. Part of that drop was the crowd and part of it was something I did not understand then. With 81 matches, every percentage point carries weight, and I gave it the full weight. That was my largest public error and also my most useful one.

Now the counter-angle, without which this piece is unfinished. We assume the empty table is the failure. My experience says the failure is something else — the empty table is only a symptom. The real failure is that we have built a framework so complete that it starts to look like analysis. Eight dimensions, eight headings, clean formatting, and no evidence underneath. Any reader could finish it believing a deep analysis of cricket has taken place, when what happened is that an empty payload put on a full suit.

That risk was recorded by the report itself, and I credit its honesty: structural completeness is the primary hazard. Because schema validation passes. In any automated system that payload goes forward with a green light, and everything downstream — dashboards, alerts, editorial triage, selection reports — treats the emptiness as truth. An empty payload is most dangerous because it passes validation.

The second counter-angle is less comfortable. The report stated plainly that no integrity or anti-corruption risk was flagged, and stated in the same breath that the absence was not a clean bill of health but an artefact of missing data. That is the most necessary and least written sentence in journalism. Where there is no record, there is no risk — that is the easiest and falsest conclusion anyone can reach.

The third: fixing the parser will not fix the problem. The report itself said the fault is easy to diagnose and cheap to repair — run the extraction again and most analytical value returns. That is technically true. But if the same pipeline runs tomorrow on a Ranji match, it will come back empty again, because there is no content there. Repairing a pipeline does not manufacture content. Someone has to put up a camera, hire a logger, and pay for it.

There is one more thing I attach to this from my own record. In 2026, in Bengaluru, I hand-coded 1,842 shots from 90 league matches across a full ISL season — freeze-framing broadcast replays to log shot location, body part, assist type and defensive pressure — because no public expected-goals model existed for Indian football. I built the xG ladder by hand, then let the numbers argue. That ladder put Bengaluru FC top on a 14.2 xG differential while warning that they were 11 goals above expectation. They won the league by six points and lost the final 3-2. The warning was published two days before the final.

One thing became permanent from that exercise: where there is no data, no model gets built — and where no model gets built, someone has to sit down and code by hand. Ranji does not need IPL-grade data. It needs three things logged on every ball: line, shot, outcome. Three columns. The model will build itself.

Which leaves the question that will return every season. The next time someone says domestic cricket lacks talent, ask which table the claim came from and how many rows that table actually held. The next time a price is debated, ask where the phase-split output behind it is. The next time a report looks complete, count its empty columns.

I do not chase the noise; I build a shelf where it can be checked later. That report's shelf was empty, but it did one job — it showed where the shelf is broken. The next job is not to join it back. It is to gather things worth putting on it.