Blank Columns, Firm Verdicts: Where Cricket Analytics Hides Its Biggest Error

January 2026, Bengaluru. The second day of the mega auction. On my laptop sat the squad-age curve of all eight franchises, and beneath it a workload column — the number of balls each fast bowler had sent down across three seasons. Shane Watson, thirty-six years old, sat at the top of that column. That evening Chennai picked him up for four crore rupees. From the chair beside me came the verdict: "Look at the age, that's money wasted." I did not argue. I had a column that stayed blank on everyone else's screen, and four months later, on 27 May 2026 at the Wankhede, that column was putting 117 off 57 balls on the board.

I learned the game in a fan forum, then spent years proving that the fans were half right. I do not write that line lightly. The voice next to me in that auction hall was not wrong — thirty-six is a risk. It was incomplete. And cricket's analytical machinery is slipping on exactly this spot today: turning an assumption into data, and in the hurry to fill a gap, turning an opinion into evidence.

Context: the part of the pipeline nobody sees on a screen

Every ball in modern cricket generates a row. Bowler, batter, line, length, release speed, swing, shot type, outcome. Add Hawk-Eye tracking, add field-placement coordinates, and you have a spreadsheet any team would envy. That is the part that looks full, and that is precisely why nobody asks what is missing.

The columns that never fill are these: a bowler's aggregated workload across three formats, travel days, hours of sleep, the wear on a pitch by the third day, a batter's confidence after something private has happened, a fielder's shoulder that took a dive last season. These are not statistics. They are conditions, and matches are built on conditions.

Our industry has developed a habit I consider damaging. When three columns in a scouting brief are left blank, the meeting does not stop. Someone with the loudest voice fills them in from his own prejudice. By the next morning that prejudice has become a model output. Nobody labelled it an input. Nobody questioned it.

Franchise scouting has quietly absorbed the function zonal selectors once performed. The result was not cleaner selection. The result was that data on second-string players became thicker than data on first-choice ones, and selection meetings produced confusion rather than comfort. When every player carries 400 T20 innings behind him, the metric is no longer the data — the metric is which column you choose to weight.

Core: how blank input becomes a firm output

Watching matches over the years, I noticed a pattern that holds well beyond cricket. When an analytical framework receives incomplete input, it does not leave the space empty — it fills the space, then presents the filled part as its conclusion. That is not a mathematical error. It is a reporting error wearing mathematical clothing.

Split it in two. One half is the data. The other half is the documentation. In cricket the documentation is far weaker. We forget to record what we did not record. An analyst who honestly says the sample is too thin to decide gets marked down as weak in the room. The one who serves a full verdict on half the data gets marked down as decisive. That is an incentive structure running backwards, and it is the real problem.

The Watson ledger: what the model could not see

In that 2026 room I had two different things. One was the age curve, which was on everybody's screen. The other was a ledger — a column beside every signing recording what role the team actually needed from that player, and whether the workload of that role was already loaded onto his body.

Watson was never an age case. He was a role case. Chennai needed a top-order batter who could hit the ball in the air during the powerplay and bowl two overs if required. A thirty-six-year-old fast bowler is a large risk. A thirty-six-year-old batter who bowls four overs a month is not a risk — he is very cheap insurance. At four crore.

The column left blank at my table was not 'age'. It was 'role fit'. And that is cricket analytics' original sin: we fill the easiest columns (age, strike rate, economy) and leave the hardest ones (role, workload, place in the dressing room) empty. Then we call the verdict built on the filled columns decisive.

The auction model worked because I stopped pretending the bidder had no pulse. The 27 crore paid for Rishabh Pant at the Jeddah mega auction in November 2026 is the cleanest example. No valuation model will call that number reasonable. But the model was not measuring what the bidder was measuring — he was buying wicketkeeper, captain and face in a single transaction, and in that market the supply of that combination was one. Demand one, supply one. The arithmetic of price breaks exactly here.

The real document at a selection meeting

We treat selection as a talent question. I treat it as a documentation question. The hardest team to make in world sport is India's second XI, and it is hard not because talent is scarce — it is hard because we have no reliable column to separate four equal claimants.

Consider what you have to distinguish a domestic spinner with 40 wickets from one with 32. Red-ball bounce, pitch wear, the quality of the batter at the other end, humidity, the fielder's hands — half of those columns appear in no domestic scorecard. Yet a name is pushed forward in the meeting, and the reason is later filed as "form".

Bench strength is a selection problem, not a talent problem, and selection is really a recording problem. Until we write down what we measured and what we did not, we have no defence for picking one of four. We will only look decisive.

I trust the model most when it explains the silence in the room after a loss. A model that cannot explain the silence is only repeating the scorecard.

The Impact Player: a column that changes mid-match

The Impact Player rule arrived in 2026 and shook an entire analytical structure. Previously you picked eleven and then distributed four overs within that eleven. Now you pick eleven and can add a twelfth person in the middle of the match.

The first consequence was not the one the coverage described. The consequence was that the pre-match plan lost value. A coach who treats the Impact Player as a tool — pulling two overs out of him in the field and then converting him into a batter — is rewriting the plan twice inside a match. A coach who treats him as a lottery ticket calls the name in the tenth over, when the pressure is at its peak.

Ninety minutes gave us a plan; extra time showed us who could still listen. In cricket those ninety minutes are the first ten overs and extra time is the last five. The first plan reveals the coach. The second hour reveals whether the captain is still listening.

Sample thresholds: three innings, five overs, one chase

Here is the least discussed and most damaging habit. A verdict built on one match. In my career I have heard it dozens of times — this captain was late changing his bowlers, so he cannot hold his nerve. Or — this batter cannot play the short ball. Both said from one innings, one chase, one sample.

Before making any claim about adaptation I set my own thresholds: number of innings, types of decisions, and the situation in which the decision was taken. Until those thresholds are crossed I do not write that a captain can or cannot still listen. Cricket's most seductive and most trusted sample is a single chase. It is also the most deceptive.

On 18 June 2026 at The Oval in London, I later opened up what happened in 2,400 words. Pakistan reached 338, India folded for 158, the gap 180 runs. Mohammad Amir took 3/16 — Rohit Sharma, Shikhar Dhawan and Virat Kohli, all three in that new-ball spell.

The real story of that match was not that someone suddenly became bad. The story was that the data on trigger movement already existed, and it showed all three batters committing to the front foot early, which makes the first fifteen overs difficult if the ball is new and moving a little. The data was there. The blank column was the moisture in the pitch that day and the morning air. That blank column decided the shape of the match.

Let me be clear that I do not grade decisions by their outcomes. I have thirteen seasons of auction data and the result always arrives first. That is the great temptation — to call every overpay a blunder and every pick a masterstroke in hindsight. I built a rule against it: publish your valuations and role-fit notes before the gavel falls, then grade yourself publicly against your own numbers after the playoffs. It is uncomfortable. It is the only way to know whether you are analysing or writing a forecast backwards.

Contrarian: the error belongs to the question, not the analyst

Now the part where I have to write against my own side.

The standard complaint from an analytics desk is that the model got it wrong. In my experience the model is rarely wrong. The error sits in the question put to it. If you ask "what is this batter's powerplay strike rate", the model answers honestly. The trouble starts when the room assumes that answer also tells you how he will bat under pressure. Those are two different questions. The model never answered the second one.

Data analysts have moved into the dressing room, and their conclusions are often cut off from the actual rhythm of the match. I have watched a spell build on the field — the bowler in rhythm, the pitch slowing, the batter sweating — while the desk is recommending a spinner on matchup grounds. Both things can be true. But the desk has no column for the rhythm of a match, and filling that column is the coach's job, with his eyes and his gut. I am not saying the desk is wrong. I am saying a desk that does not recognise its own blank column starts calling its opinion data.

The second thing, and the one that troubles me most, is that we dismiss the crowd's read as noise. The crowd often gives the wrong reason and arrives at the right conclusion. The forum writer did not say Watson's workload could be managed; he said the team was short of something and this man would cover it. That was a half-right diagnosis. I took it and tried to measure the other half. People who wave away the crowd's read as rubbish lose the most valuable signal of all — the one that shows up late in the data and is felt early at the edge of the field.

Takeaway: the first question in the next meeting

Next time a scouting report or a selection model lands in front of you — before the next IPL, in the first meeting of the next domestic season — do not read the conclusion first. Ask which column in it is blank. A report that does not name its empty column is not telling you what it has assumed. And the hidden assumption does the most damage.

In cricket we usually argue about whose number is bigger. The argument we should be having is about what we failed to record. Until that argument happens, every confident verdict will have a blank column standing behind it — and that column speaks in the final over.