Where the Numbers Go Quiet: Indian Cricket's Empty Dataset and the Real Cost of Selection

Aditya Gupta | Hyderabad

---

On a Wednesday night I ran a first pass on an article and the pipeline returned an empty set. No title, no source, no information points, no entities. In the trade it is called a null return, and the rule says treat it as a failure and run it again. I ran it again. Same result.

The interesting part is not the script stalling. The interesting part is that a large share of what I have read under the label "analysis" across forty years of cricket writing was, in fact, written on exactly this empty set — only the emptiness was set to a rhythm. When the model sees nothing, it writes "strength of intent." When the sample is three matches, it drafts a comeback story. When it has no idea about the pitch, it calls it "a surface full of character."

My job is to hold a simple line: what cannot be measured must not be written in the grammar of measurement. An empty dataset is itself a conclusion — provided the writer has the nerve to publish it.

---

Context: an information economy split in two

Indian cricket runs two entirely different data worlds, and the distance between them is the subject here.

The first world is the IPL. From every delivery come release point, speed, seam position, swing, bat angle, shot direction, initial fielder positions. Tracking cameras generate ball-by-ball raw material any analyst can re-run. The rights money sits here too: the media rights for the 2026-27 IPL cycle sold for ₹48,390 crore — roughly north of a hundred crore rupees per match. That figure is public and checkable against any source.

The second world is the Ranji Trophy. More than thirty teams, well over a hundred group-phase matches a season, and for most of them the public record stops at a scorecard: runs, balls, fours, sixes, order of dismissals. No tracking, no pitch maps, no line-and-length data. Television and streaming coverage collapses to the knockout stage.

There was a time I read this gap as a budget problem. In 2026, when the BCA–Mafatlal Bowling Scheme was being built and Frank Tyson was shaping Abey Kuruvilla, Paras Mhambrey, Sairaj Bahutule, Salil Ankola and Nilesh Kulkarni in the nets, I watched from sixteen or seventeen yards to see whose seam moved on which morning and whose did not. That information never entered a scorecard. It lived only in the heads of the people standing there.

That is the real problem. Cricket's most valuable data is the data recorded by the cheapest possible method — namely, not at all.

Method (version 1.3): This piece rests on three public sources and one private log — (a) the declared media rights value of the 2026-27 IPL cycle, (b) the Ranji season structure and the reach of available coverage, (c) my own model logs from the 2026 FIFA U-17 World Cup and the 2026 IPL in the UAE. Where a number comes from my log rather than a public document, it is labelled as such. The model does not include the variables I could not measure: the preparation history of the pitch, the curator's decision, the closed room where selectors sit.

---

Core analysis

What gets measured, and what does not

One IPL innings generates more data than an entire Ranji season. The comparison sounds harsh, but it is not a metaphor — it is structure.

So when we read a number like "900 runs in the season," we do not know which bowling that came against. Opposition quality is not even a column in the public data. Which means that in a runs-versus-runs comparison we may be weighing one batter against Group C attacks and another against Group A attacks, and we will never be told.

I have logged this blank under three names:

• Wicket variable unknown. How much a surface changed across four days is not in any public file.
• Weather variable unknown. Dew, wind direction and humidity show up in bowling figures, but the variable itself is absent from the figures.
• Opposition variable unknown. There is no public index of bowling strength.

From experience: a clean dashboard is not evidence; it is a question wearing a suit. A dashboard that hides its empty cells is not analysis, it is interior decoration.

The base-rate question: selection on what data?

Putting together India's second XI is the hardest job in world sport. Eleven take the field, but thirty or forty names sit behind them with domestic records good enough to play international cricket. Every selection call is therefore a data decision, whether or not anyone calls it one.

That is where the base rate problem lands. Suppose an opener makes 900-plus runs in a season. The question is not whether the runs were made. The question is: how many came against top-four bowling, how many on spin-friendly surfaces, how many in the first ten overs against the new ball? None of those three questions can be answered from public data. By my own log, batters from sides like Mumbai and Bengal have often topped four-season charts, but the decomposition that would explain that height was never published.

The effect shows up directly at the auction table. By the 2026 auction the purse had reached ₹120 crore per team, and a large share of that goes to players valued mostly on small T20 samples. One more distortion compounds it: the Impact Player rule, in place since 2026, has changed every denominator in every comparison. A 2026 strike rate and a 2026 strike rate cannot sit on the same scale, because batting depth, bowling workloads and innings planning have all shifted.

When I build auction tables, I stamp the rule era on every comparison. Where two eras do not overlap cleanly, I refuse the chart. That is not stubbornness, it is arithmetic.

When the model published its own limits

In October 2026 I stepped away from ten IPL seasons for six weeks to consult on tracking data at India's first FIFA tournament — the U-17 World Cup. Fifty-two matches, six venues, Kochi to Kolkata. Alone in a Guwahati hotel room I built a possession-value model from scratch.

It ranked England's Rhian Brewster — eight goals, Golden Boot — as the tournament's most efficient shooter two rounds before the final. That felt good, but it was not the point. The point was that I published the weights, gave the spreadsheet away free, and wrote down what the model could not see — defender positioning, goalkeeper reach, pitch quality.

The open model worked because it admitted what it could not see. A closed model makes the mistake a closed selection room makes: it treats its own blindness as invisible.

That U-17 data taught me something else. A model without a story is just arithmetic. When I printed Brewster's number alone, nobody read it. When I wrote down the circumstances — which minute, at what score, against which defensive shape — the number began to speak.

Three years later, during the 2026 shutdown, I applied the same method to a variable nobody was isolating: the crowd. I pulled all sixty IPL matches played in the UAE — Dubai, Abu Dhabi, Sharjah — plus the ISL season played behind closed doors in Goa.

The result: at neutral venues, home advantage collapsed into statistical noise, which is to say zero. And Sharjah's short boundaries pushed scoring roughly 1.9 runs per over above Dubai's. That 1.9 figure is version 2.1 of my log; the original draft recorded it per ball, which was wrong. The correction is on record.

The conclusion was that home advantage is mostly scheduling and pitch allocation, not the magic of a ground. Two editors sent it back. One digital outlet ran it. That night a new file opened in my log — rejected pieces. It has fourteen entries now, three of which later proved correct. I mention the count because rejections and errors both deserve to be counted, otherwise a model becomes self-congratulation.

The most useful variable I logged that season never made it into a spreadsheet column. I counted the empty seats and found a variable no spreadsheet could hold. And when the crowd vanished, the model kept its shape and lost its meaning.

That lesson applies to the current tournament cycle. Every major tournament assumes atmosphere is a constant. It is the least constant variable on the board.

---

The contrarian angle: transparency is also a claim

Now the part where I have to doubt my own argument.

The easy route is to say: bring more data and everything is fixed. What I have actually seen in the nine years since 2026 is the opposite. Indian cricket's data desks won the tactical argument long ago. What lost is the commentary box, which still sells words like "intent" and "character" — claims that cannot be measured, dated, or retracted.

But here is the more uncomfortable conclusion: transparency is not a virtue in itself. It is also a claim, and like any claim it should be falsifiable. There are analysts who publish every weight they use and still reach wrong conclusions. Being openly wrong is better than being quietly wrong, but it is not the same as being right.

The second contrarian point is harder. This information gap is not an accident. No group-phase Ranji coverage, wall-to-wall knockout coverage — that is a choice, not neglect. Knockouts produce narrative. Narrative is cheap to sell. Three months of group cricket produces evaluation, and evaluation is expensive to sell, because evaluation can be proven wrong.

Here I have to repeat my own warning. Four decades of franchise and board relationships taught me that when a source hands you something inside, their incentive slowly becomes your frame. So every piece I file states who spoke and what they gain. The media rights number is good news for the BCCI, and the same number explains why Ranji coverage has shrunk. Both conclusions come from one fact.

And the biggest trap of all — one I have fallen into — is treating a model as prophecy. Thirty-five years of building them creates a fluency that reads like foresight. A T20 knockout is a one-off; extracting a single number from it and calling that a forecast is a category error. I no longer write "will." I write: here is the range, here is the sample size, and here are the conditions under which the estimate breaks.

I keep a separate list of variables I cannot quantify. It includes the curator's name, the ticketing manager's estimate, and the steward who counts empty seats. Talking to them is not outside my model. It is the part of the model I have not yet converted into a number.

---

Takeaway

Next time you open a Ranji scorecard and see 900 runs, ask one question before the number: how much of this came against top-tier bowling? If the answer is not in a public file, you have not read a conclusion. You have read an empty set.

Over the next two seasons I will be watching three signals: whether domestic group-phase data ever becomes public; whether auction valuation tables start carrying the rule-era stamp; and whether Impact Player usage gets priced against actual output or against the name.

And my own empty set? I am not tearing it up. I am keeping it on file, with a version number. Because the day I stop showing my blanks as blanks, I become the thing I have spent forty years criticising.

Are we ready for cricket writing that can occasionally print: our data on this match is insufficient? I think the reader is ready. The question is whether the people on the other side of the desk are.