36 All Out Is Not an Outlier — It Is the Forensic Trail We Refused to Read

19 December 2026, Adelaide. Bio-bubble, empty stands, and 36 all out in 21.2 overs. I was in Delhi — the press box door had been shut by the pandemic, so I had nothing but the scorecard. When a team is bowled out for 36, the first reaction is 'this was an aberration.' I thought so too. Then I opened the ball-by-ball feed and began coding every dismissal by line, length, shot type, and balls faced. Four hours later the picture in front of me was not psychological — it was structural. The top order's average stay was under 20 balls, and the pattern had been logged in the data for eleven months. 36 all out was not an outlier; it was a forensic trail we refused to read.

This piece is an attempt to make that trail public. Because until we treat 'mental fragility' as an explanation, those same eleven months will recur, and we will be surprised again.

The Birth of the Data Desk in Cricket, and My Bias

In 2026, when I was 50 and had spent eighteen years writing scorebooks — what editors called 'the numbers corner' — I talked a Delhi digital outlet into funding a three-person data desk. The condition was this: a live win-probability curve for every IPL match. On 21 May 2026, at the Rajiv Gandhi International Stadium in Hyderabad, Mumbai Indians defended 129/8 against Rising Pune Supergiant. I logged all 240 balls myself; my model had Pune at 87% at the 12-over mark — and the chase collapsed into a one-run defeat. That piece ran the curve, not the narrative.

Since that day I have banned the word 'momentum' from my copy and instituted a mandatory 'decision point' paragraph — the over in which the match actually turned. Editors who asked for 'feel' got a table. That bias is still with me: I trust the dashboard, but I audit the story it tells.

At the 2026 Bengaluru mega auction I ran a value-over-replacement model built on three seasons of death-over data, regressed hard to the mean. The model flagged Jaydev Unadkat's ₹11.5 crore buy by Rajasthan Royals as the worst value on the board, and KL Rahul's ₹11 crore move to Kings XI Punjab as the smartest. Within six hours of the final gavel I published a 200-player buy/sell table. A franchise official asked the room, 'Who let the stats lady in?' That season Unadkat's economy climbed past nine.

That context matters because to understand Adelaide's 36, you must accept that every bid at the auction table is a confession about what the market cannot measure — and the worst-measured thing in cricket is top-order dwell time, the duration that never appears in a highlight reel.

Context: The Part of the 2026 Series Nobody Read

It is essential to remember how India played the first match of that series. In the first Test (Adelaide) India collapsed for 36, but the pink-ball Test's peculiarity — the evening session under artificial light, when the ball swings and seams more in the first hour of night — was a factor I flagged separately in my coding. In my coding, 10 of the dismissals came in the evening session, where pink-ball seam movement was sharp and Indian batsmen's footwork was losing balance on drive-length deliveries.

But session change alone does not explain why 36. Session change is an external factor; 36 is an internal pattern. I pulled the top-order data from every Test innings of the previous eleven months (2026-20 season, India's home and away Tests). The result was clear: the Indian top order's balls-faced-per-dismissal average had fallen by up to 40% on overseas seaming pitches, and the contribution of the night-watchman and lower order kept rising — meaning the top order's job was shifting to the lower order. That was not consolation, it was warning: when top-order dwell time falls, one day it will approach zero. In Adelaide that day arrived.

Moreover, the empty stadium of 2026 was a natural experiment. I was using that period to test whether home advantage is crowd-driven or pitch-driven. Even in empty stands the pink-ball seam movement was just as sharp; the difference the crowd's absence made in favour of bowling was less psychological pressure than the absence of 'set-shot' feedback — a batsman's rhythm is built by crowd reaction, and in an empty stand that rhythm breaks. That insight will matter later.

Core: Coding Every Dismissal and the Chain That Led to 36

I coded the full 21.2 overs of Adelaide's innings across four dimensions: (1) line, (2) length, (3) shot type, (4) balls faced. Key findings, without decoration:

First: 24 of the 36 runs came from the three batsmen who faced more than 15 balls. The remaining eight batsmen contributed 12 runs. The problem was not that everyone fell early; it was that even those who stayed did not rotate strike. The dot-ball percentage was 71%, among the highest for any Test innings that season. The pile of dot balls was the fuel that built pressure, not any single magic delivery.

Second: there was a clear cluster in the manner of dismissal. 6 of the 10 dismissals came on 'drive-length' deliveries — balls outside off stump that tempted the batsman to drive, most of them caught behind or at slip. That was not a technical flaw but a 'flaw of intent': batsmen were forced to score because runs were not coming, and they played shots at balls outside their scoring zone.

Third: the wickets fell in clusters — three slides, each of 3-4 wickets. First cluster: 3 wickets in the first 6 overs after the opening stand. Second: 4 wickets in the next 7 overs. Third: 3 wickets in the final 8.2 overs. This cluster pattern is my strongest evidence that this collapse was not random — it was a chain reaction in which each wicket built pressure for the next. The model did not predict the one-run final; it explained why we should have seen it coming — and here too, the model could not predict 36, but it could flag the risk of 'cluster collapse,' and I had that flag eleven months earlier.

Fourth: the empty stand's effect was on rhythm, not psychology. I compared: in home Tests of 2026-19 the Indian top order's strike rotation (rate of taking runs after leaving a ball) was 34%; in Adelaide it was 19%. In an empty stand a batsman gets no feedback of being 'set' — neither applause nor the echo of pressure. This is technically inverse evidence of 'crowd-driven home advantage': the crowd's absence would have hurt the host team too, but for India the damage was doubled because its top order was already weak.

Now the part I consider most important — this collapse was not 'mental fragility,' it was 'ball-level urgency.' I noticed that when India were 30 for no loss in 15 overs, batsmen were still 'surging' — playing shots at balls not worth hitting but worth defending. This is not psychology, it is 'run-rate pressure' that comes from a target-oriented game. When you are bowled out for 36, every batsman thinks 'I must score at least something,' and that thought gets him out on a drive-length ball.

Here is the moment everyone missed: the 11th ball of the innings. India were 4 runs, and a batsman left a ball outside off stump that seamed in. In the next over he drove at the same line-length and was out. That was not a 'lack of discipline'; it was 'decision inconsistency' — leaving the first time, playing the second time, with no new information in between. That is the evidence I demand instead of 'mental fragility': decision patterns at ball level. If we do not measure it, we will talk about 'the mind' after every defeat and be surprised again next time.

Contrarian: Correlation Is Not Causation — and the Real Culprit of '36' Is the Selection Room

Now the part where I must criticise my own work. Reading the data above, the easy conclusion is: 'the top order breaks on overseas seaming pitches, so change the top order.' But correlation is not causation. The fall in top-order dwell time across eleven months was a symptom, not a cause. The cause was the selection room's decision not to change the top order's structure in those eleven months — the same batsmen, the same order, the same roles, even as the data screamed.

Here I must record an uncomfortable truth about my profession. Working in India, the board controls my credentials, the press box and the whisper network. At 59, after five beats, I have relationships to protect. But when I look at Adelaide's 36, I want this on record: the responsibility for this collapse belongs to the analysts and the selection room, not to fate. The data was available eleven months earlier; whoever did not see it chose not to see it.

And the second contrarian point: those who call '36' a 'one-day aberration' forget that India won later in that same series — but that win did not erase the structural cause of 36. A win is not a solution to a problem; it temporarily covers it. That is the same mistake we make when we call it 'confidence.'

Third, what stings my profession most: the columns that called 36 all out a 'crisis of Test cricket' were written on the same day IPL retention data came out. The truth of the media footprint is that the Ranji Trophy is almost invisible outside the knockouts, while every IPL retention call gets a column. I will not say in nostalgia that Ranji should get the coverage it had in the 1990s — that coverage will not return, and pretending it will is a lie. But this must be said: the IPL's real casualty was never Test cricket; the casualty was the relevance of domestic first-class cricket. When domestic coverage thins, no one is left to track structural signals like top-order dwell time — and then 36 arrives suddenly, and we call it an 'aberration.'

Takeaway: What Must Change Before the Next Test

My desk's rule is: every collapse analysis must end with a 'what changes before the next Test' paragraph. Here are three things I demand on data, not emotion:

First, 'balls-faced-per-dismissal' for the top order should be made a formal metric and published before every overseas tour. When it drops below 20, that is a 'red flag,' not a 'mood.'

Second, a separate preparation protocol for the combination of empty stands and pink ball. My data says strike rotation falls by up to 15 percentage points in empty stands — that is trainable, if we measure it.

Third, and hardest: a 'data veto' in the selection room. When an eleven-month trend says the top order's structure must change, no one should have the right to ignore it in the name of 'experience' and 'form.'

36 all out was no outlier. It was the forensic trail we refused to read, and we will refuse to read it again if we keep treating 'mental fragility' as an explanation. The next time a team collapses inside 40, the first question should not be 'how?' but: 'Why did we not have that eleven-month data — or if we had it, who refused to look?'

My scorecard is open again this year. When a team next collapses inside 50, I hope I will already have a timestamped probability — and if my model misses, I will show that miss in the same column. Because credibility does not come from access; it comes from the ledger you publish every week.