Bangladesh's T20 Middle Overs: What 47 Innings of Logged Data Say That the Scoreboard Doesn't

June 16, 2026, at Arnos Vale in St Vincent. Bangladesh were bowled out for 106 in 19.3 overs. Nepal collapsed to 85, Bangladesh won by 21 runs and walked into the Super Eight. In the press conference the questions circled around bowling; the next morning's headlines told a story of brave defence. In my ledger that innings is the lowest-scoring win of the cycle. At eleven at night I sat down to update the phase map and found exactly what had been there six months earlier: a boundary percentage below ten in overs 7 to 15, for the twenty-fourth innings in a row. At two in the morning I wrote a line in the margin — the win arrived, the pattern celebrated a birthday. Nobody actually loses on the field; a drought simply gets longer.

This piece is about that drought. What came out of a log of forty-seven innings is not a bowling story. It is a story about batting method, and method changes very slowly.

1. I left Rajshahi for a digital desk in Dhaka paying eighteen thousand taka a month. That season I charted all 66 matches of the domestic football league by hand — shot location, body part, defensive pressure, goalkeeper position. In week six I rebuilt the whole sheet in Python. My expected-goals table showed Abahani Limited Dhaka had outperformed their xG by 11.4 goals; the real table showed Abahani as champions. Nobody in Bangladeshi football had printed those two numbers side by side. That day I stopped writing sides 'deserved to win' and started attaching a number to every claim, with a methodology note under every column.

I carried that habit into cricket in 2026. From June 2026 to February 2026 I logged 47 Bangladesh T20 innings by phase — powerplay overs 1-6, middle overs 7-15, death overs 16-20. For each phase I recorded run rate, boundary percentage, dot-ball percentage, the over in which each wicket fell, and the over in which the number five batter walked in. As a baseline I used global men's T20 phase averages for 2026-2026. For every innings I calculated a phase surplus: Bangladesh's actual runs minus the runs expected at the global rate from the same number of balls faced. A positive value sits above the baseline, a negative value below it.

What I left out matters too. I did not weight for opponent strength, I did not classify pitch type separately, and the confidence intervals on 47 innings are wide. Making a big claim off a small sample is the easiest trap in my trade, and I know it.

In the first six overs Bangladesh's run rate is 7.4 against a global average of 7.6 — the gap is essentially zero. The dot-ball percentage is 42 against a global 39, slightly high but nothing alarming. In four of every five innings the powerplay more or less survives. The trouble begins in the seventh over, and there is nowhere to hide it.

Between overs 7 and 15 Bangladesh's run rate is 6.3 against a global average of 7.9. Boundary percentage is 10.2 against 14.4. Dot-ball percentage is 45.6 against 36.1. Across those nine middle overs the side runs roughly a run and a half per over below its peers, and almost one ball in every nine passes without a run. On a tournament schedule, that shortfall is about 13 to 14 runs per innings. That is the difference between two wickets and an over in hand.

The death overs, oddly, are fine — 9.4 an over in overs 16-20 against a global 9.0. This is the uncomfortable part of my model. The claim that Bangladesh fold in the last five overs is not supported by this log. The failure story is written in the middle overs, not at the end.

That good death-overs number is conditional, though. When two or more wickets fall in the powerplay — 14 such innings — the picture inverts: middle overs 5.4, death overs 7.8. The late-innings charge is a product of wickets in hand, not independent proof of skill. Once wickets fall, that reserve empties and there are no balls left to spend.

The bowling-side cross-check sharpens the picture. In overs 7-15 Bangladesh's spinners concede 6.6 an over against a global 7.4 — the strongest phase the team has. The fourth wicket falls at an average of 15.8 overs, against a global 14.6. Bowling was keeping matches alive while a batting method problem hid behind the spin numbers. This is the cricket version of a losing winner — the scoreboard delivers a result, the phase map does not.

Bangladesh won 11 of those 47 innings. In seven of the eleven the phase surplus was negative. Most of the wins came with a negative batting process, pulled up by bowling or by the opposition's fielding and death-bowling errors. For anyone judging a side purely on results, those seven numbers carry weight.

Franchise economics sit underneath this. Every auction is a ledger, and every rumour has a decimal point. Batters who can clear the rope in overs 7 to 15 see their price jump; batters who reduce dot balls in the powerplay sit far down the list. The league rewards a certain type of batter, and the national side slowly casts its batting method in the same mould. That number is not made on the field; it is made at the draft table.

The easy explanation is that the number five has a poor strike rate. My holdout window says the opposite. I split the innings in two — those where the number five walked in before the eleventh over, and those where he walked in at the thirteenth or later. In the first group his strike rate is 118; in the second, 141. A 23-point gap. The names have changed — Towhid Hridoy, Jaker Ali, Mehidy Hasan Miraz have all batted at five or six — but the median entry over has not.

Here lies the correlation trap. The number five's entry over is not a measure of his ability; it is a proxy for how many balls the top order has burned. In innings where dots pile up at the top, the lower-order batter arrives to rebuild; in innings where boundaries come early, he arrives with licence to launch. The hunt for a finisher is therefore posting letters to the wrong address. The right address is the last two overs of the powerplay and the first three balls of the seventh.

One heuristic, offered carefully — it is borrowed from football data and cannot be transplanted directly into cricket. After the 2026 shutdown I tracked 306 matches across five leagues and found the home win rate fell from 43.2 per cent to 33.6 per cent in empty stadiums, with home xG down 0.11. In hypothesis language: a neutral venue erases the edge that a home ground gives through the toss and through familiarity. The approximate cricket translation is that pitch knowledge counts for little at Arnos Vale or in Dallas, and Bangladesh's batting method leans harder on familiar conditions than most. That remains a testable hypothesis, not a logged event.

One line against myself. I had pre-registered the hypothesis that death overs were to blame for the tournament failures. That hypothesis is rejected; the death-overs run rate sits above the global average. Based on my years of watching matches, the eye almost always turns to the final over, because that is where the drama lives. Numbers do not look at drama. If the log refuses my story, the fault is not the log's.

For the next cycle I will watch two signals, and both concern entry overs and dot balls. First, whether the powerplay dot-ball percentage drops below 40. Second, whether the number five walks in after the twelfth over in 60 per cent or more of innings. If both happen together, the middle-overs drought can invert, because it would mean the top order is banking balls instead of burning them. One of the two would bring partial improvement. Neither would mean the 2026 story is reprinted word for word in 2026, with only the cover image changed.

The last line in my ledger has not been written yet — it waits for the first powerplay of the next tournament. If Bangladesh shed a single dot ball across those six overs, I will be able to write that cricket changes so slowly nobody notices. But the spreadsheet did not lie, and the model does not look at the flag.