The Quiet Ledger of Middle Overs: Bangladesh's 47 Percent T20 Question

Hook: The number that stays awake after the broadcast ends

The spreadsheet began to hum, and I knew the broadcast was over. Twenty-fourth of June, 2026, twenty minutes to four in the morning, a Hackney flat lit by one laptop fan. Bangladesh had lost to Afghanistan at Arnos Vale four hours earlier, and with it their tournament. The highlights had already moved on. I had not. I was staring at a column I had named the Dot-Pressure Proxy, and next to Bangladesh it read 47.3.

That number means Bangladesh's batters played more than one dot ball every two deliveries between overs seven and fifteen across nine matches. Forty-five middle overs. I had no commentary box that night, no colleague to nudge and say I told you so. I had three CSV files, one Python script, and a silence you only hear inside a data table. There is a monastery inside every dataset, and its silence is not empty.

Seven years earlier I walked out of a London radio studio after an on-air fight about Burnley's "lucky" sixteenth place. I had pulled up their 2026-17 expected goals: 42.1 for, 44.8 against, a minus 2.7 differential that looked mid-table, not relegation fodder. My producer called it spreadsheet sorcery. I quit that week and started a weekly xG column. Since then every lede opens with a number, not a scene.

Context: three phases, one empty room

A T20 innings splits into three parts. Overs one to six are the powerplay, seven to fifteen the middle, sixteen to twenty the death. Under ICC playing conditions only two fielders may stand outside the inner circle during the powerplay; from over seven onwards, five may. That single rule sets the economics of the format. Boundaries are comparatively cheap early, risk is licensed late, and the nine overs in between - forty-five percent of an innings - belong to arithmetic rather than emotion.

Analysis culture mostly ignores that middle block. Powerplay strike rates make highlight reels. Death-over sixes trend. A forty-ball rebuild in overs ten to fourteen makes nothing at all. The player who survives there is invisible; the player who stalls there is unexplained.

Bangladesh know this invisibility well. In 2026 they beat West Indies and reached the Super Eight. Seventeen years passed. In 2026, in the United States and the Caribbean, they reached it again - beating the Netherlands and Nepal, losing to South Africa, watching the Sri Lanka fixture wash out. That is not a small achievement. But three straight Super Eight defeats to Australia, India and Afghanistan shared one repeated failure, and it was not bowling. It was the empty middle.

Core: dissecting one metric

Let me be frank about method before anything else, because this is my own model and not an official statistic. I logged every ball of the tournament by hand: batter, bowler type, over number, runs, dot or not, boundary or not. Then I weighted three components into what I call the Dot-Pressure Proxy. Fifty percent weight on dot-ball rate. Thirty percent on strike rotation - the ability to take one, two or three off any given delivery. Twenty percent on a silence score, measuring consecutive boundary-less balls. The higher the total, the more an innings seizes up in the middle.

Bangladesh's figure was 47.3, among the three worst of the ten full-member sides in the tournament.

Why is a dot ball more expensive in the middle? Opportunity cost. A dot in over three can be recovered. A dot in over fourteen means the power hitters never receive the platform. Broadly, one middle-over dot forfeits somewhere between ten and twelve runs of downstream expectation, because that delivery could have been a six at the death. This is why a middle-over dot and a death-over dot look identical on a scorecard and are not.

I do not trust the eye test until it can survive a scatter plot. So I split the middle into three bands.

Overs seven to nine are the aftershock of the powerplay. The field spreads, spin arrives, and Bangladesh slowed most sharply here. Every match showed the same shape: discomfort against left-arm spin, and near-total stoppage of rotation when the surface offered slow bounce.

Overs ten to twelve are theoretically the best scoring window, because bowlers are still saving overs and fielders still attack. My count put Bangladesh at one boundary roughly every twenty-eight balls in that band, nearly double the tournament gap. This is where Liton Das's innings most often died. He can absorb boundaries, but he cannot manufacture a single, so the pressure of the over never leaves him.

Overs thirteen to fifteen Bangladesh actually handled well. Mahmudullah Riyad's experience carried innings through that stretch, and the strike rate there is clearly healthier. Bangladesh's middle-over problem is not the death approach; it is the six overs immediately before it.

Two contrasting models sit alongside. India's middle overs run on rotation - singles, twos, unbroken pressure on the bowler. Boundaries may not come, but dots do not either, and the innings arrives at the death prepared. Afghanistan invert it: their strength is middle-over bowling, especially leg-spin. In 2026 they beat Bangladesh exactly that way, choking boundaries through the middle and forcing a chase outside Bangladesh's natural rhythm.

A personal memory attaches here. In 2026, reporting for The Daily Star, I interviewed Soumya Sarkar as a rising star; that piece became my first verifiable byline when Prothom Alo picked it up. He told me he did not count overs, he watched the ball. Nine years later, scrolling Arnos Vale data, I thought: T20 now demands that you watch the over rather than the ball. That is not a player's fault. It is the format's cruelty.

Look at the bowling side and the picture inverts, and this inversion is the most neglected part of the story. Running the same model as a symmetry check, Bangladesh's spin attack dragged opponents' middle-over dot rate down to roughly parity. Young leg-spinner Rishad Hossain finished as Bangladesh's leading wicket-taker in that tournament, with Taskin Ahmed and Mustafizur Rahman strangling rotation through the middle. The bowling unit was keeping Bangladesh in matches. The batting unit was spending that currency on piles of dots.

Methods travel across sports. In football I have tracked for years how a goalkeeper whose shot-stopping is quietly declining gets repriced upward the moment he can kick long. Cricket has a cousin: the all-rounder whose primary skill is eroding while the secondary skill guarantees the shirt. The metric changes clothes, not logic. At Russia 2026, the hosts' group-stage PPDA of 8.7 was the most aggressive pressing by a host nation in tournament history. Spain completed 1,005 passes in Moscow and lost. I ran the PPDA numbers again, and the flat in Moscow started to feel real. Use one metric as a scalpel, then turn the blade on it three paragraphs later.

Contrarian: correlation is not causation

Now the blade on my own throat. The 47.3 is real. It proves very little on its own.

First objection: pitch. The 2026 drop-in surfaces in the United States were slow, bounce was irregular, and the ball gripped quickly. In that environment, dot balls measure conditions more than skill. Teams on flat decks naturally post lower middle-over dot rates. Sometimes I am measuring Bangladesh's batting. Sometimes I am measuring the groundstaff.

Second: sample size. Forty-five middle overs is not a dataset. If five or six rotation attempts per innings fail, that is variance, not deficit. A confidence interval around 47.3 will happily swallow 43.6. Anyone who computes 43 is not wrong; they are using a different definition.

Third, and largest: selection. A batter given the middle-over brief was often not chosen for the powerplay. Bangladesh have no deliberate division of labour between top and middle order, so whoever walks in at seven is frequently conscripted rather than designed. That is a system flaw masquerading as a player statistic.

Now the ethical kill switch. I will build a model for six days and delete it in one afternoon if it starts pointing at a human being. The 47.3 is not a person's name. Last year a young cricketer messaged me asking what his middle-over number was. I replied that his number was not his identity. That night I closed the sheet, because a column was generating fear in someone who had done nothing but take a job nobody else wanted. The first rule of data journalism is that people must never become data.

There is a structural cousin to this, clearer to me since I was appointed one of three BCB advisors in 2026, overseeing digital and media affairs. Franchise economics increasingly imitate football's loan-with-obligation model: a large organisation lets a smaller one do the development, talent matures on somebody else's balance sheet, and the profit lands where no academy exists. Traces of that mechanism are entering Bangladesh's domestic game. The damage does not appear in a dataset, because it happens off the page.

Takeaway: what to watch next phase

In the ghost games the crowd disappeared, but the pressing lines left fingerprints. My arithmetic suggests Bangladesh's T20 problem is not an absence of boundaries. It is the absence of singles while boundaries are absent. The first is a ceiling on talent. The second is a gap in planning.

So I am pre-registering one hypothesis in my public project ledger. If Bangladesh lift their middle-over strike rotation rate by five percent - without adding a single boundary - their last-five-over scoring should rise by roughly eight to ten runs. One condition: keep the same batting order for at least four matches, or the sample is meaningless.

I have five half-built dashboards in this flat and a podcast pilot eight months unfinished. This method, though, is complete, and it is not mine. Anyone can pull ball-by-ball data, change the weights, and run the same vacuum. The answers will differ; the road will not. I have drawn one line on one scatter plot. How long it holds is a question for February's middle overs, when the scorecard and one number will sit side by side: 47.3 - a limit on batters, or a failure of preparation?