What the Scoreboard Hides: Bangladesh's T20 Middle Overs in a 118-Match Spreadsheet

Chittagong, a T20 evening. Chasing 165, the No.3 batter finished on 52 off 44. It was the highest score of the match on the scorecard; the broadcast called it a "responsible innings". The dugout applauded. The team lost by nine runs. The last five overs needed 56, six wickets were in hand, and the man who faced 44 balls was still there at the end.

Back at my laptop was a seven-year-old Excel file: the 2026 BPL season, all 66 matches, hand-charted, later rebuilt in Python. Next to a 44-ball fifty of the same shape was a number in red font: −8.7. The model said that innings cost the side 8.7 runs against the match-state-adjusted expectation, not added them.

The scorecard and the spreadsheet are both telling the truth. The only question is which one you call a good innings.

Context: one sentence, one World Cup, one pipeline

For a decade, Bangladeshi T20 conversation has run on one sentence — the top order scores, the middle overs stall. At the June 2026 ICC Men's T20 World Cup in the United States and the Caribbean, Bangladesh reached the Super Eight for the first time, then lost to Afghanistan, India and Australia. Their bowling economy sat in the tournament's top four; their batting rate sat near the bottom. The first crack appears right there: the problem is not uniform across departments.

When I joined The Daily Star sports desk around 2026, the lesson was simple — a claim without a number next to it is not a report, it is a comment. The 2026 episode was the next step. At 24 I left Rajshahi for a Dhaka digital desk paying BDT 18,000 a month and hand-charted every ball of that BPL season — shot location, body part, defensive pressure, keeper position. By Week 6 I had rebuilt the sheet in Python. It surfaced two numbers nobody in Bangladeshi football had printed side by side: Abahani Limited Dhaka were 11.4 goals ahead of their expected-goals baseline, and they were champions. I stopped writing "deserved to win" and started attaching a methodology footnote to every column.

That habit now runs on cricket. The base here is a ball-by-ball dataset of 118 T20 matches involving Bangladesh between January 2026 and December 2026 — BPL 2026 and 2026, bilateral T20Is, and ICC events. Every ball carries seven tags: bowler type, line-and-length bucket, the over the batter entered, shot intent, field setting, outcome, and the ball's baseline wicket probability. After losing my job in 2026 I built my own scraping pipeline, and since then one rule holds — every claim ships with code and a reproducible dataset link.

This is not football xG. Expected runs in cricket means what an average ball of that bounce, line, field and shot zone is worth, and combining it with par score and wicket resources gives a match-state valuation. Not everything on a cricket field is measurable, but what is measurable can stop being lied about.

The core: it is rhythm, not rate

I define the middle overs as overs 7 to 15, because that window hides more unclaimed runs than any other. Bangladesh's run rate there is 7.38. The top eight sides in my dataset average 8.21. That is 0.83 runs per over, roughly seven and a half runs a match — not small in a format decided by two or three.

But the gap is not evenly spread, and that is the real finding.

The middle-over problem is not speed, it is rhythm. In the three balls after hitting a boundary, Bangladesh's dot-ball rate is 41.6 percent. Across the rest of the dataset it is 32.4 percent. We hit the four or six, then stand still for three balls. In those three balls the bowler finds his length, the fielders push in, and the captain's field settles. The next boundary is then four balls and ten runs away.

Ball by ball, the pattern sharpens. In overs 7 to 15, Bangladesh's singles-per-ball rate is 0.42 and boundary rate 0.103. The boundary arrives; the next ball goes back to the bowler. On a long sample the same dot-clustering picture repeats.

The most neglected truth is the last line of that picture: Bangladesh bowl well in the middle overs. At the 2026 World Cup, Rishad Hossain's leg-spin asked a completely different question in that phase; Nasum Ahmed's left-arm orthodox and Mehidy Hasan Miraz's control often kept opponents under six an over across ten. In the six matches where the overs 7–15 bowling economy sat in the sixes, Bangladesh had at least four wickets down before the 16th over in five of them.

So where does the defeat come from? Overs 16 to 20 produce a run rate of 9.78, which is competitive. But the side enters that phase with two new batters at the crease — one just out, the other starved of strike for seven overs. The issue is not the absence of a power hitter; it is the absence of a batter.

Then there is a cost no broadcast graphic shows: fielding. In my sample Bangladesh drop 1.9 catches a match, each worth 4.3 runs. In a phase where every ball is expensive, eight or nine runs simply fall out of the hands.

The No.3 batter arrives at 4.1 overs across the sample, and at 5.8 for Bangladesh. That is a delay of nearly two overs — twelve to sixteen balls. The best batsman finishing dinner before walking out is habit, not strategy.

At player level, two of the sample's top three Bangladesh batters sit below a 125 middle-over strike rate, and the team average without them is 119. That eight-to-ten-run gap compounds every match, and under tournament pressure it surfaces as 128 instead of 140.

The spreadsheet did not lie. We asked the wrong question.

The contrarian angle: correlation is not causation

The most repeated explanation of the last four years is that Bangladesh lack power hitters, so the middle overs stall. The dataset does not dismiss that outright — power is genuinely short. But the biggest controllable variable is not power; it is entry point: who comes in, when, and with whom.

Dot-ball percentage correlates with defeat at 0.61. Correlation is not cause. When I control for the composition of the batting pair — how much their shot zones overlap, whether the same delivery ties both down — the dot-ball effect falls to 0.19. Dot balls do not hurt by themselves. What hurts is two similar batters occupying the crease together, letting the bowler know exactly which length shackles both.

I have fallen into that trap twice. In 2026 I wrote a column treating middle-over dots as the single cause, and a women's bilateral dataset broke the claim. In 2026 I reached a verdict off seven matches of one tournament. The rule I write by now: build the hypothesis on one half of a season, test it on the other. The 2026 World Cup Super Eight was a pre-registered holdout. In two of three matches the model flagged in advance that Bangladesh would score under eight an over between overs 7 and 15. Both times it happened. It failed in the third, because the opponent bowled nothing but spin — the conditions changed and my checklist had not accounted for it.

One outside example, offered strictly as a heuristic rather than an equation. On 27 June 2026 in Kazan, Germany lost 0-2 to South Korea while recording 2.31 expected goals to South Korea's 0.78. Football xG and cricket expected runs are not the same system — ball speed, pitch behaviour and field settings differ entirely. But the question is identical: when the scoreboard and the process disagree, which do you trust? South Korea's win did nothing for them in the next round. Bangladesh's 44-ball fifty will do nothing in the next window unless the method changes.

The second contrarian point is incentive. Every auction is a ledger, and behind every rumour sits a decimal point. In the BPL auction, teams pay for average, because average means not-out, and not-out means the individual's balance sheet looks good. The batter who makes 52 not out off 44 raises his price; the one who makes 41 off 26 and gets out lowers his, even though the second innings gave the team more runs. Under that incentive structure, discipline does not survive.

Limitations, plainly. The 118-match sample involves only Bangladesh, so the comparison base is thin, especially in direct bilateral matches against top sides. BPL ball-tracking data is limited, so my line-and-length buckets come from scorecard and video coding and are not exact. Dew and pitch behaviour are large variables I have captured only through session-level qualitative flags. Strike rate itself is never a neutral measure. With those four caveats, this is an estimate, not a verdict.

Takeaway: the number I will watch next cycle

Next tournament cycle I will not watch strike rate. I will track two things. First, from over 12 onward, the ratio of dot balls in the three deliveries after a boundary — that single number is the true health reading of a middle phase. Second, the average over in which the No.3 batter arrives. If the second figure falls from six to four-and-a-half across two seasons, and the first eases from 42 into the mid-30s, I can predict the results in advance — though I will not need to, because the spreadsheet will say it first.

The question, then, is not about power hitters. It is whether Bangladesh's selection system will tolerate a batter with a low average and high impact. If the board cannot answer that, we will still applaud 52 off 44 in 2030 while somebody in the next room sits quietly, staring at −8.7 in red font.