Zero Information Points: Where Cricket's Data Chain Breaks

Eleven fields. Not one fact.

This week a cricket file reached a data desk in Chennai. From the outside everything looked fine — the label was attached, the category was set, the schema was complete. Then the fields began to open. Title: not available. Source: not available. Article type: unclassified. One-sentence summary: blank. Author stance: undefined. Information points: empty list — zero.

A file that had called itself cricket could not return a single fact that could be repeated, checked, or placed beneath a claim. The desk had two routes. The first was to fill that emptiness with a plausible-sounding story — a big match, a big name, a big number — and write it with confidence. The second was to call zero zero.

The second route was taken. That is the subject of this report.

A single innings scorecard breaking is not news. But when the lowest link in cricket's information chain — the extraction layer — comes back empty-handed from an entire file, that event is itself a signal. It says that the "data revolution" this sport has celebrated for a decade rests, in places, on foundations thin enough that one parsing failure can take it to zero.

And in cricket zero is never neutral. Zero is a claim. When a database says nothing about a player, the world fills that space with memory — and memory, as I have seen across nineteen years of work, is far more confident than data and considerably less precise.

How a Fact Is Made

In the summer of 2026 I left a broadcast production job and joined an eleven-person digital sports startup in Chennai. There was no data desk. I was told that during the IPL season I could build whatever I wanted, provided I built it.

So I built it. Sixty matches, ball by ball — fourteen thousand four hundred deliveries. A record for every ball: which bowler, which batter, which over, what outcome, how much pressure. On that data I built a metric I called "death-overs exposure" — an index weighting required rate against wickets in hand.

The final was in Hyderabad. Mumbai Indians made 129/8; Rising Pune Supergiant stopped at 128/6. One run. The trophy was decided by one run, or so the scorecard says.

My table was saying something else. Across that playoff season, the finalist's numbers six and seven absorbed 38 per cent of its death-overs balls between them, while its top three faced fewer pressure deliveries than any other qualifier. The one-run margin was being decided forty overs earlier, in the part of the batting order the camera never visits.

Roughly nine hundred thousand people read that piece. The franchise's analyst called — first to argue, then to offer a job. That season I was the only woman in a thirty-seat press box at Chepauk.

The story matters because it shows one thing clearly: the metric that worked did so not because it was clever. It worked because beneath it lay fourteen thousand four hundred separate, checkable facts. Facts first, story after.

Now reverse it. Imagine that at the end of that season someone had asked me which team had the worst death-overs record, and my table had come back empty. I would have had two options. Say I did not have an answer. Or reach into my memory, my bias, my likes and dislikes, manufacture an answer, and present it as data.

The second option is chosen everywhere. And that is where the biggest problem in cricket's analytical culture begins.

The Most Expensive Zero: Format

Among the eleven fields in that empty file was one called format. Blank.

Ask anyone who writes about cricket which single missing field does the most damage, and the answer should be format. Test, ODI, T20, The Hundred — the grammar of the numbers is different in each. A strike rate of 140 in T20 is the ordinary equipment of a competent opener. In Test cricket, 140 is an event — innings-changing, but an exception. Same number, two entirely different meanings. Read that 140 as a Test figure and you have committed cricket's most common analytical offence: cross-format conflation.

Economy rate works the same way. Nine an over is roughly acceptable in T20 today. In ODIs it is expensive. In Tests it gets you removed from the attack. On averages, a Test average of 35 and a T20 average of 35 are two different professions, two different players.

I have watched this happen dozens of times. A young batter scores at a strike rate of 150 in the Syed Mushtaq Ali Trophy, and within a week something is written about "questions over his Test technique" — as though T20 strike rate were a measure of red-ball method. The two have no arithmetic relationship. An aggressive T20 approach is often an advantage in Tests; plenty of modern Test batters are the proof. But to reach that conclusion you need Test data, not T20 data.

An empty format field looks like a technical glitch. It is actually the loudest warning in the whole report: the moment you do not have a format, you do not have a single comparable number. Benchmarks cannot be set. And without benchmarks, every analysis is an opinion in good clothing.

The Definition of a Fact

The schema contained a field called information points. That is the heart of the process.

An information point is a discrete, verifiable fact. Something you can pass to another person, match against a second source, count or weigh.

Take an example. "The captain said the team is balanced." Not a fact. A quotation. Useful, perhaps, but not fuel for analysis. "That team's numbers six and seven absorbed 38 per cent of its death-overs balls this season." A fact. Countable, repeatable, challengeable.

An empty information-points list means the file contained nothing of the second kind. Only fragments of the first, and even those incomplete.

Here is an uncomfortable truth the industry does not say often enough. A very large share of cricket media coverage is of the first kind — statements, quotes, mood, the temperature of the day, the air in the dressing room. All of it is readable and some of it sells. None of it is analysis. It is reporting. There is no sin in either, but calling both by the same name is where the damage starts.

The desk I worked on had a rule I still think is right: a minimum of five to eight hard information points. Below that, the file does not move. The number is not arbitrary. With three facts you can make an observation. With five to eight you can make a pattern. Without a pattern, analysis is just commentary.

We are forgetting that distinction. After every big match, hundreds of "analyses" appear containing not one verifiable fact. Everything sounds right. Nothing can be checked. That is exactly the condition the empty file created — except there the desk wrote "zero" honestly, and here the story is told with confidence.

The Inverted Pyramid

That file's emptiness may be a technical event. It also points at a structural problem, and the structural problem is real.

Indian cricket's data pyramid is upside down. At the top sits the IPL, where every ball is logged, every ball's speed measured, every shot named, every fielder's position mapped. Sixty matches, fourteen thousand four hundred deliveries, and at least ten variables on each.

Below that sits the Ranji Trophy. Thirty-eight teams. More than a hundred matches in a season. Tens of thousands of deliveries. How many of those are coded ball by ball? Very few. Most seasons output at scorecard level — who scored how many, who took how many, who batted how long.

The difference is not only of quantity but of kind. A scorecard tells you a batter made 38 off 74. It does not tell you how many of those 74 were against the new ball, how many against spin, how many came while wickets fell at the other end, how many came with a day and a half to save.

The innings most valuable to the team is the one least visible in the data.

Women's domestic cricket is thinner still. The WPL is smaller than the IPL, and below it reliable scorecards are not universal, let alone ball-by-ball data. In associate cricket it sometimes happens that an international is played and its entire statistical history sits on a single page of a single website.

That asymmetry is itself a selection structure. A player without data never enters the selection argument. He is not seen, not sold, not remembered. A player with dense data — meaning a player who has been through the IPL — automatically appears to stand ahead of twenty others, whether or not he does.

This is where a line from one of my own pieces comes back to me: I stopped counting trophies and started counting habits. The problem is that we only have the tools to count habits for the people in front of the camera.

The Box's Real Limits

There is one party in this argument I want to be precise about, because I think it has been treated unfairly.

The commentary box.

I spent nine years in it. I know what happens there. You are live. You have no feed. A ball arrives and you have one ball's worth of time to react. You cannot say, "hold on, let me check the database." The viewer needs words in that moment. What is available in that moment is memory — and memory does work there, because thirty years of experience is faster than any spreadsheet.

That does not make the box right. It makes the box constrained, and calling a constraint laziness is a mistake.

But there is something else I cannot ignore. The box is not only constrained by time; it is constrained by structure. It still sells milestones. "That's his two hundredth match." "That's his fiftieth half-century." The information is accurate, and it is not analysis. It is a count converted into emotion.

In my database I have seen, repeatedly, what decides matches: wickets in the powerplay, run-rate pressure through the middle, the exposure of numbers six and seven at the death, the timing of a dropped catch, the state of the ball in the first ten overs after a toss. Nowhere on that list is a player's fiftieth half-century.

The box sells it anyway, because the audience buys it. And there the problem is ours, not the box's. We told the audience that the count is the story. Had we told them thirty years ago that a number six's performance under pressure is also a story, the box might be selling that instead.

Selection: A Data Problem, Not a Talent Problem

In that empty file the list of players was also empty. No names. No teams.

That was a property of the file. It is also a picture of a real Indian problem. India's second XI is the hardest team to make in world sport — the line has become a proverb, and proverbs are usually true, but nobody explains why.

The reason is not that India has too much talent. The reason is that the selection machinery has been distorted.

Franchise scouting is now doing zonal selectors' work. The change happened slowly, which is why nobody noticed. Once a player had to score in Ranji, then in Duleep or Irani, then for India A, and only then did the national door open. At every stage a different eye watched him.

Now a large part of that process is decided by advertising and franchise data. And franchise data is incomplete, because it only measures the conditions franchises play in. Twenty overs, a certain pitch, a certain crowd, a certain pressure. Test pressure is a different species — it arrives three days in, when the legs stop answering and the scores are still level.

One more layer has been added: the Impact Player rule. Introduced to the IPL in 2026, it makes the game more watchable, no doubt. It has also broken role definitions. An all-rounder no longer needs to be a whole all-rounder. He can bowl four overs and bat eight balls and be entirely useful.

Great for a franchise. Dangerous for a national side, which still needs people who can bowl fifty overs and bat two hours. We are preparing a generation under a rule that does not apply in international cricket.

Back to the empty list. If selection is based on data at any position, and data exists for only one kind of cricket, selection will bend toward players from that kind of cricket. That is not prejudice. It is arithmetic. But the result looks like prejudice, and it is usually read as a shortage of talent.

The Contrarian Case: An Empty Table Says More Than a Full One

Now to the point I disagree with most colleagues about.

Many would say that publishing a report with zero information points is a failure. The desk should have worked harder, found sources, rebuilt the file, and then written. I agree it should be rebuilt. Not instead of publishing, but alongside it.

Because an empty table is information.

It tells you a joint has broken somewhere in the chain. It tells you the extraction layer failed — a parser, a format, a human check that did not happen. Had that failure been hidden beneath a good story, it would never have been fixed. It would have happened next week, and the week after.

This is where an old lesson returns. Working with live xG taught me to watch the second screen and the room. The second screen tells you what the number is saying. The room tells you what people are hearing. When the two diverge, the real story is there.

With that empty file, the second screen was showing zero and the room was saying nothing. That was the loudest signal of all.

There is a further illusion in this industry. We believe more data has made analysis better. My experience says something else: more data has made analysis faster. Better is a separate process, involving doubt, questions, and refusal.

A model is only as honest as the questions it refuses. A model with no option to say "I don't know" is not a model. It is a guessing machine wearing a calculator's coat.

Our industry has many such machines. They do the most damage precisely when they look most reliable.

What the Data Cannot Answer

I have to be clear about something, because a data-first identity creates pressure to have a number for everything.

It does not.

About Indian domestic cricket, particularly Ranji and women's domestic cricket, our data is thin. I have said this before, but it has a practical consequence worth stating plainly: this data cannot tell you who the best player is. It can tell you whose record is best documented. Those are different things, and treating them as one is the error that has distorted our selection system.

For instance: one batter makes 900 runs in a Ranji season, another 750. Is the first better? Perhaps. But if the first played six innings against weak Plate Group bowling and the second played four in the quarter-final and semi-final, the scorecard is misleading you. We have no systematic way to supply that context, because we never coded those balls.

The same applies faster to women's cricket. The WPL has changed a great deal, but the layer beneath it — the senior women's domestic tournament — still does not produce the data at the level we want to analyse it. That does not mean do not analyse. It means declare the sample size and the coverage gap first.

I learned that lesson expensively. Once I built a claim on a small sample — twelve innings over eighteen months — and published it. An analyst wrote to me and broke the claim with my own data. He was right. The sample was too small and I had not stated the limit. In a Chennai startup I learned that empty tables make loud enemies, and that day I had no excuse.

What a Real Fix Looks Like

If the problem is that the chain can break, the answer is technical discipline, not good intentions.

First: make format a mandatory, non-nullable field. No file without an explicit Test, ODI, T20 or Hundred label should move. That one rule would remove the industry's most common analytical error.

Second: install a validation gate. A file with fewer than five information points, or an empty title, or no source name and date, should never reach the analysis layer. It sounds like good journalism. It is actually good engineering.

Third: make source quality a field that cannot be left blank. We deserve to know whether a claim came from a board release, a journalist, a general media report, or an account whose business is traffic. Those four do not carry equal weight, and showing them as equal is dishonest.

Fourth, and hardest: the nerve to publish zero.

I know that advice comes from someone who once reached nine hundred thousand readers. But that piece did not travel because of a clever sentence. It travelled because beneath it lay fourteen thousand four hundred balls. Without those balls it would have been worthless — and I hope I would not have written it.

Why Football's Vocabulary Does Not Travel

There is one more trap here that I want to avoid myself, because my background pulls me toward it.

I have spent a few years working on live xG models in football. That vocabulary is native to me: expected goals, progressive passes, packing, half-spaces. When I return to cricket, I want to use those words.

That would be wrong, and the reason is mathematical.

Football's uncertainty structure differs from cricket's. A football match usually contains two or three goals and the value of a goal is direct and linear. Cricket has no ceiling on runs — a side can make 129 and win, another can make 180 and lose. Cricket's real currency is not runs; it is wickets and balls.

So the metrics that work in cricket are of a different kind: phase run rate, wicket probability, required-rate pressure, DLS-adjusted targets, net run rate scenarios. All of them sit on a finite resource — deliveries. Football's finite resource is time, and time is counted uniformly.

When a cricket analyst borrows xG-style vocabulary, he is often concealing the limits of his own model. Not from malice. From a search for language that feels easy. But cricket's uncertainty is strange enough that saying it simply makes it wrong.

The Gap Between Club and Country

That file contained no league, no contract, no fee. Yet this is precisely where part of cricket's information chain breaks — between club and country.

We run two entirely separate data worlds. Franchise cricket, where everything is logged, every contract value is public, every player is valued. And international cricket, where central contract terms are secret and selection reasons are sometimes never given.

Between them sits an empty space, and in that space lives the question asked after every franchise auction: is this fee worth this player?

My answer has always been the same: wrong question. Commercial value and sporting value are different things. A high auction price does not tell you how good a player is internationally. It tells you how many alternatives that franchise had for that role, at that auction, on that day. It is supply and demand arithmetic, not cricket.

Treating the two as one is the same mistake football's transfer market makes every year. The transfer market speaks in fees; the spreadsheet asks who pays later. In cricket the question sharpens, because an auction ends in a day and its effects run for three years.

The Real Risk

The biggest risk in this story is not cricketing. It is analytical.

The risk is a claim that cannot be proven false.

If I write that a team's middle overs are weak, that is an opinion and it can be argued. If I write that the team scored 14 per cent below par at 6.2 an over through the middle, that is a claim that can be checked — and if it is wrong, I have to concede.

The first kind of sentence is easy to publish. The second is risky. Our industry is running toward the first while calling itself data-driven.

Being genuinely data-driven means restricting yourself to claims a colleague could break. It is uncomfortable. It is slow. And it is lonely, especially in a room where everyone else is applauding the first kind of sentence.

What to Watch Next

Three things, over the coming months.

First: when you read any major cricket analysis, count how many verifiable facts it contains. Fewer than five, and it should not be called analysis.

Second: notice which institutions disclose their sample size and which hide it. Trust the ones that disclose more — even when you dislike the conclusion.

Third: watch domestic coverage. Next time a young player appears in the national side and you ask where he came from, check whether you can find ball-by-ball data from his Ranji season. The day that becomes available, India's selection system will change — without a committee, without a report.

Sixty matches, one feed, and a growing list of doubts. I do not want that list to get shorter. It is the list that tells me I am still working.

The day it empties, I too will have eleven fields and not one fact — and on that day I hope nobody tells me the data has been checked.