Ledger of a Collapse: 38 Matches Say Bangladesh's Middle-Order Breaks Begin at Over 13, and 42 Percent Share One Source

Rajshahi, the last week of February. A laptop on the balcony table, a cup of tea going cold, and one frame frozen on the screen — not a photograph, a timestamp. Fourth ball of the 13th over. The match was still open. The scoreboard read 89/3 and the equation demanded 10.2 an over.

The ledger said something else entirely. In that bowler mix, that field setting, those conditions, Bangladesh's expected runs across the next 24 balls came to 31.6. They scored 11. The shortfall of 20.6 runs is roughly a sixth of a T20 innings, evaporated inside a single frame window where the batters did not change, the laws did not change, and only the over number moved.

A few years ago I would have covered that gap with one word: pressure. I no longer do. Pressure is a feeling, and feelings cannot be audited. Frames, balls, field coordinates and over numbers can.

Method: why I open the table first

In 2026, aged 44, teaching kinesiology in Rajshahi, I coded an open-source xG model for the Bangladesh Premier League. I logged all 132 matches — every shot, every pressing sequence, every coordinate. Abahani Limited Dhaka's title run finished 8.9 points above expected points. Sheikh Jamal Dhanmondi Club's Nabib Newaj Jibon scored 15 goals from 11.2 xG. I delayed publication by three weeks to verify every shot coordinate.

The Rajshahi xG ledger taught me that small samples still leave fingerprints.

The architecture built for football ports directly onto cricket. An over is a phase, a ball is a shot, a field setting is a block height. Since then I begin every cricket piece with one question: which metric is contradicting the scoreline?

Sample, stated plainly, because a number without a sample is decoration:

• 38 men's T20Is for Bangladesh, 1 March 2026 to 21 February 2026.
• Conditions normalised: Mirpur's slow, low deck; Chattogram's true bounce; the flat, larger Dubai and Abu Dhabi ovals; the sea-breeze swing windows of Sydney and Melbourne.
• Excluded: rain-shortened segments where DLS revised the target — discussed separately, because that rule is the largest trap of all.
• Variables: phase-xR, Dot Ball Pressure Index (DBPI), Wicket Cluster Index (WCI, defined as three or more wickets in 24 balls), boundary dependency ratio, and the bowler workload cliff.

After 37 years of watching grounds and tables, one thing holds: cricket data does not lie, but data does not speak alone. Call it as a witness, not a judge.

The geography of collapse: 42 percent enter the same alley

Across 38 matches Bangladesh produced 19 wicket clusters. Eight of them — 42 percent — began between the last ball of the 12th over and the second ball of the 14th. That is not coincidence. It is the hinge where the post-powerplay accumulator phase ends, the fielding side recalls its pressure bowlers, and the set batters rotate out.

| Phase | Bangladesh phase-xR | Actual RPO | Deficit | |---|---|---|---| | 1–6 (powerplay) | 8.94 | 8.61 | −0.33 | | 7–12 (middle) | 8.42 | 7.88 | −0.54 | | 13–16 (pressure window) | 8.61 | 7.42 | −1.19 | | 17–20 (death) | 10.78 | 10.31 | −0.47 |

The table argues for itself. Bangladesh are roughly level in the powerplay and roughly level at the death. The structural loss accrues from overs 13 to 16 — 1.19 runs per over. Across 38 matches they batted in that phase an average of 3.7 times, so roughly 4.4 runs per match are lost by default. In tournament cricket, 4.4 runs is one extra match-altering decision resting on the bowlers' shoulders.

Bangladesh's middle order does not lose the slog overs. It loses the over immediately before them.

I keep dot balls in a separate ledger

Clusters pull analysis toward emotion because a wicket is visible. The dot-ball index is clumsier and less photogenic. But before every collapse, something happens that is not a boundary. It is silence.

In the 13–16 window Bangladesh's DBPI is 41.8 percent. Opposing bowling units average 33.2 in the same phase. In seven matches where DBPI exceeded 45, Bangladesh clustered in six. Where DBPI stayed below 36, clusters came in only four of nineteen.

Generating turnover requires rotation; rotation requires reading release time and field spread. Those are separate skills. Blending them turns analysis into blame and denies a good fielding side its credit.

Thirty-four percent of wickets come from one matchup

Left-arm orthodox spin. Across 38 matches, 34 percent of middle-order dismissals came from that angle, and 23 percent of those were premeditated strokes, not forced errors.

When opponents deploy it is more revealing. In 14 cases the bowling change arrived precisely in the over when the set batter's strike rate peaked above 140. Opposing captains were not reading batters; they were reading phases.

The workload cliff

Bangladesh pace economy in overs 17–20 is 9.8, respectable. In overs 11–15 the same unit sits at 7.1. The gap is workload, not quality. Bowlers used in the 11–15 pressure role return for 17–20, and the spell breaks there. Across the sample the post-17th-over economy jump averages 2.7 runs per over — structural risk rather than personal failure.

Fingerprints of a small sample

When the stadiums emptied in 2026, the numbers finally spoke without an echo. Home advantage fell from 0.42 to 0.18 goals per game; referee stoppage bias dropped 31 percent. Dataset behaviour shifts with an audience.

In the domestic ledger, young batters with a 118 strike rate in their first ten innings settled at 129 across a career. Those who faced 40-plus balls at least once in that window settled at 137. The difference is exposure, not talent. Five matches is not a career — but how many balls you faced in five matches can forecast one.

Where prices rise, and where value actually forms

Every transfer is a hypothesis wearing a deadline and an agent.

A 22-year-old Bangladeshi batter's price rises 35–40 percent after three good tournament innings, while the runs he influences per innings stay almost flat. Tournament inflation is simple: price tracks visibility, value forms in the invisible work — surviving dots from overs 13 to 16, breaking a set bowler's spell.

— Root: 2026 Russia World Cup France.

At Russia 2026, 5.8 of France's 14 goals came from set-piece xG, with a PPDA of 12.8 — a controlled mid-block trap. Some called it destiny. The data said a set-piece-dependent model can be devastating in a short bracket without proving permanence. One bracket can crown a team without validating it. The parallel is plain: two good group-stage DBPI performances tell you nothing about a knockout, where bowling plans, field formations and dew all change at once.

The contrarian turn: am I selling correlation as causation

My strongest claim — that the middle order lacks match awareness — is incomplete and falsifiable. Control for required run rate.

| Required rate | Matches | Cluster rate | |---|---|---| | Below 8.5 | 12 | 18% | | 8.5–10.4 | 15 | 33% | | Above 10.5 | 11 | 51% |

If clusters were purely temperamental, the distribution would stay smooth after controlling for rate. It jumps from 18 to 51 percent. What we call a test of nerve is, in significant part, a mathematical obligation.

Rival hypotheses deserve airtime. Bowling change: opponents switch on matchup; nine of fourteen clusters carry this signature, which would reclassify the metric as a measure of bowling success. Dew and conditions: seventeen clusters came with dew above 50 percent. Toss and field protection: six showed an abnormal batting-share collapse.

Limitations, stated plainly: 38 matches is small; I have no ball-tracking data, so I cannot separate bat-plane error from delivery deception; domestic field-coordinate data is often missing.

My model expires after four series. If the deficit in that phase drops below 1.00 and cluster rates halve, I am wrong, and I will say so.

Takeaway: where the next round gets watched

Three variables stay on my slate. The first three balls of the 13th over — who bowls, how high the block stands, whether the reverse sweep has begun. The No. 5 batter's DBPI across his first 12 balls: above 42 percent means a collapse is loading. Whether the left-arm orthodox angle arrives as early as the 11th over.

Bangladesh's middle order may not be weak. It may be answering the wrong question — how to score in the slog overs, when the break happens one over earlier. — Root: Data Monk archetype | Scenario: personal essay or reflective piece on data work.