The Dot-Ball Sequence: A Hidden Middle-Overs Pattern Across Bangladesh's Last 23 Innings

I have re-watched a single innings four times over the past three weeks — once on the scorecard, once on the conditions sheet, twice on the ball-by-ball timeline. Seven wickets fell, and those seven are now the centre of every conversation. But scrolling the ball-by-ball file, my eye caught something else: 47 dot balls across the 180 deliveries between overs 11 and 40, against a tournament baseline of 38 on the same surface, in the same wind, under the same conditions. The wickets were a symptom. The disease was the dot-ball cluster — not continuous, but in pockets, exactly where the innings was supposed to take root.

Twenty-one years of watching matches tells me our stories about Bangladesh batting collapses almost always end on the same sentence: they could not absorb the pressure. Is that sentence true? Or is it a conclusion we reach because we stopped reading at the first layer of the data?

The numbers didn't break the model; they exposed where the model was blind.

Method note: rules first, narrative second

When I launched Expected Truth from Khulna in late 2026, the first lesson was that the index must be written before the story. Otherwise any explanation can be retrofitted to the data, and that is not analysis but staged evidence.

Sample: 23 Bangladesh ODI innings between January 2026 and December 2026, filtered to matches where Bangladesh either chased 240-plus or set 240-plus. Nine venues, six different bowling attacks, six pitch profiles. The filter exists because middle-overs strategy in a 180-run innings measures compulsion, not craft.

Two metrics. Phase Leverage Index (PLI) = [actual phase run rate / conditions-adjusted expected run rate] x [1 - (phase dot-ball rate / tournament dot-ball rate)]. Recovery Efficiency (RE) = run rate in the three overs after a wicket / run rate in the three overs before it. Conditions adjustment uses four variables: venue first-innings average over the last three seasons, day-night split, spin-revolution proxy, outfield speed. I have capped myself at six variables. Put seven or eight into ODI middle-overs and what comes out is not a model but a model's jewellery. Revision rule, locked in advance: if the rolling PLI over five innings moves beyond two standard deviations of baseline, I re-weight variables — I do not swap the sample or add variables.

Core: three phases, three different teams

Across 23 innings, phase PLI reads powerplay 1.08, middle overs 0.79, death overs 0.96. Overall dot-ball rate is 36.9 percent, but between overs 11 and 40 it climbs to 41.3 percent. Same measure for India's middle overs: 1.14. Afghanistan 1.02. Sri Lanka 0.94.

So the finding is not the lazy one that Bangladesh bat badly in the middle. It is that Bangladesh overperform expectations in the powerplay and hand the entire surplus back in the middle overs. There is no saving account for surplus runs. A football parallel is fair here: Abahani Limited Dhaka in 2026 scored 34 goals from 26.8 xG, a +7.2 overperformance, and held it because their set-piece structure was repeatable. Cricket has no set-piece for powerplay overperformance. Overperformance is a loan, and the repayment date is the first ball of the 11th over.

Dot balls do not arrive at random. In this ball-by-ball sample, 49 percent arrived in clusters of two or three within a three-ball window. One dot ball pulls another behind it. More precisely: 61.4 percent of Bangladesh's middle-overs dot balls came in spinners' overs, although spinners bowled only 47.8 percent of deliveries in the sample.

Here is where my model was blind. In version one I treated dot balls as a batsman output — defensive shots, settling in, the wagon wheel shut. Version two showed the reverse: strike rate in the ball immediately after a dot rises roughly 34 percent, usually through a boundary attempt, and wicket probability in the following two overs is 1.6 times baseline. The dot ball does not take wickets; it forces the shot that does.

Recovery Efficiency is the most uncomfortable number. Bangladesh average 0.62. India 0.91. Afghanistan 0.84. In 75 percent of these innings the wicket fell between overs 12 and 26, and 68 percent of those wickets came within two balls of a dot delivery. Blaming temperament cannot explain why the blame always lands in the same over-window.

Conditions matter but do not explain the gap. Since 2026, first-innings averages at Mirpur have fallen to 237 from 261 across the previous three seasons. Splitting the nine venues into spin-friendly (Dhaka, Chattogram, Sylhet) and batting-friendly (Pallekele, Dambulla, Hobart), Bangladesh's middle-overs PLI is 0.74 and 0.87 respectively. Both sit below baseline. India on the same spin-friendly surfaces: 1.09. Sri Lanka: 0.96. No conditions theory covers a 113 percent leverage gap.

The second pass also caught something the first pass missed. I had only measured dot balls from the batting side. In fact 62 percent of Bangladesh's middle-overs dot balls were balls the batsman actually hit, saved by fielders. I call this Direct-to-Fielder Rate: the share of shots travelling straight to a fielder between five and 25 metres from the rope. Bangladesh: 31.7 percent against a 27.2 percent baseline. Mehidy Hasan Miraz's middle-overs economy in this sample is 4.4; Taskin Ahmed and Mustafizur Rahman concede 8.9 at the death. Good numbers, weak correlation with results — when you are bowled out for 185, a 4.4 economy is decoration.

The data ledger: why the number cannot be edited later

The biggest difference between my 2026 model and my 2026 model is not technology but verifiability. I now treat every ball-by-ball entry as a ledger record: timestamp, bowler, batsman, shot zone, conditions hash. Once written, an entry cannot be quietly changed — only corrected by a new entry, with the correction permanently visible. This matters because official scorecards still contain gaps, particularly around wides, byes, leg byes and shot classification. Cross-checking these 23 innings against the CricSultan database, three innings matched on runs and balls but diverged on shot classification. Years ago I buried such divergences under the word approximate. Now I log them in a ledger where even a correction looks like a correction.

Contrarian: the explanation I refuse to give

First objection: correlation is not causation. Dot-ball rate predicts results strongly, but both may be outputs of a third variable — opposition spell quality, or early moisture on the surface. In three innings of this sample Bangladesh lost despite a 4.9-plus middle-overs run rate; in two they won at 4.1.

Second objection: my own bias. I have not spent ten consecutive minutes in a dressing room. If a batsman says that spinner on that pitch had no easy route, I can complicate his claim but not disprove it.

Third, and largest: the question may not be the middle overs at all but the powerplay. If a 1.08 leverage is built on fragile risk, then middle-overs caution is simply the bill for the powerplay, not a middle-overs failure. That would invert my entire framing.

Takeaway: pre-registered for the next five innings

If Bangladesh's rolling middle-overs PLI across the next five ODIs holds at 0.92 or above, my pattern thesis is supported. Between 0.85 and 0.91, partial revision: I increase the weight on conditions. Below 0.85, I change the framing and re-test powerplay fragility. Secondary trigger: if post-dot wicket probability falls below 1.3 times baseline, that counts as counter-evidence. Third trigger: if Direct-to-Fielder Rate drops below 29 percent, the fielding variable leaves the model.

Expected truth is not a verdict; it is a probability statement that will either survive five matches or fall flat. Either way, I will write down which one happened.