Dot-Ball Cartography: Why Bangladesh's ODI Chases Fracture After Over 34

I watch the second innings of an ODI from the western gallery at Sher-e-Bangla National Stadium with a different set of eyes. My scorecard is not the board; it is a ball-by-ball log — which delivery was a dot, which one was a single, which one pulled a fielder two steps sideways. Across six years I have logged 41 ODI run chases ball by ball. In 23 of them, the required run rate had crossed 9.00 by the end of the 34th over. Five of those 23 were won. The comfortable explanation writes itself: high rate, forced risk, wickets fall. The data does not say that. In 14 of those 23, six or more wickets were in hand. In 15, a batter who had already faced at least 25 balls was at the crease. The shortage was not wickets and it was not deliveries. It was a structure for rotating strike every over. And that structure does not collapse at over 34 — it was being assembled quietly from over 12. Context I built the first version of this from the wrong end. In 2026, in a bedroom in Rangpur, I logged every shot of France versus Argentina and built a crude xG model. It taught me one thing that matters here: football xG is per-shot probability, and there is no clean transplant into cricket. Football's state is continuous before the ball reaches the net. Cricket's state is discrete; the expected value jumps from delivery to delivery, and the batter's stroke quality multiplies the input variables several times over. What I run in cricket is expected runs added — wicket probability times run probability, weighted by match state. It is not xG. It is the cousin that measures the workload xG measures, and I state the gap between the two at the top of the piece rather than at the bottom. Dataset: 41 ODI run chases, January 2026 to December 2026. Twenty-six involve Bangladesh; fifteen involve India, New Zealand, Australia and Sri Lanka. The era window is a constraint, not a flourish. Adjustments to the two-new-ball regulation and the application of fielding restrictions have shifted powerplay scoring since 2026. Pooling seven years would corrupt any powerplay-level conclusion. Context integrity note: crowd, weather and travel sit on a separate layer from tactical metrics. The 2026 empty-stadium window remains the closest thing I have to a controlled natural experiment — across 83 Bundesliga matches played behind closed doors, home win rate fell from 43.2% to 33.7% and goals per game from 3.1 to 2.7. Part of home advantage is crowd-driven, not purely travel-driven. So whenever a cricket number appears here, venue-level effects are stripped out first. Core Analysis Powerplay launch-pad deficit Across the 41 chases, the average powerplay score was 40.3 with 1.2 wickets lost and 2.1 dots per over. Split by side: India 52.1, New Zealand 49.8, Australia 48.4, Bangladesh 38.6. That gap is 13 to 14 runs. The objection arrives quickly — what is 14 runs? In this dataset, average first-innings totals sit at 278. A 14-run deficit means the middle overs must be played at 8.3 an over instead of 7.8. Inside that half-run, the tolerance for dot balls roughly halves. At 8.3, you can afford about 20 dots across 120 balls. At 7.8, the allowance is 24. Four fewer powerplay runs costs you four dots of breathing room later. Recalibration one: a chase is not lost in the powerplay, but the powerplay decides how much risk the middle overs can carry. Dot-ball compression, overs 11 to 40 Dots per over by phase: 2.1 in overs 1-10, 3.4 in overs 11-20, 3.6 in overs 21-30, 2.8 in overs 31-40, 2.4 in overs 41-50. Boundaries per over: 1.02, 0.61 across 11-30, 0.74 in 31-40, 0.98 in 41-50. The lowest-entropy window is overs 11 to 30. Over those twenty overs the distribution of possible outcomes narrows — fours fall, dots rise, singles barely move. Holding a chase together through that window demands roughly 6.4 an over. At 0.61 boundaries per over, that is one boundary every other over. Everything else is footwork: ones and twos. To make 128 in twenty overs, about 80 must come in singles and twos, meaning 80 of 120 deliveries must produce a run. The remaining 40 are dots. Forty dots is the real weight of the match. I compute a dot-compression index for every chase — total dots in overs 11 to 30 divided by total boundaries. The median across the 41 chases is 5.7. All eleven chases above 7.0 were lost. Eight of the nine below 4.0 were won. I borrow a technique here from an earlier life, with the seams marked. Tracking Italy's pressing structure at the European Championship, I used PPDA — passes allowed per defensive action — and learned that pressing is a ledger, not chaos. Cricket has no literal PPDA; there is no clean equivalent of a defensive action. But the structural job is identical: pressure imposed per unit of time. When I count dots per over in overs 11-30, I am reading the same ledger with a different denominator. Where the mapping breaks: in football, pressing is a team decision; in cricket, a dot is often the child of field placement and pitch behaviour, not the batter's intent. The set-batter strike-rotation paradox Conventional wisdom says a set batter scores faster later, because the wicket risk is banked. My data points the other way. Batters who have faced more than 25 balls carry a dot percentage of 41.0. Batters under 25 balls carry 38.4. Being set does not reduce dots by three points — it adds them. Two mechanisms are at work. First, field setting. With a set batter at the crease, the opposing captain pushes mid-on, long-on and deep point back. The one- and two-run channels close, rotation stops, and the only remaining option is a boundary at wicket risk. Three or four dots follow. Second, spin matchups. Middle-over spinners bowl over the wicket into the pitch, push the batter deep in the crease; the result is fuller deliveries, less rotation, more dots. One number I keep returning to: Bangladesh converts singles into twos at 21% in overs 11-30, against 29-31% for New Zealand and India. Eight to ten percentage points sounds small until it is projected over 120 balls — that is twenty to thirty runs. Death-over entropy Dots drop to 2.4 and boundaries rise to 0.98 in overs 41-50. The easy read: the match is decided in the last ten. My model warns against that read. Both sides hit more boundaries in the death. First innings average 78 in overs 41-50; chasing sides average 71 — a seven-run gap. But look at dispersion. Winning chases show high variance in the last ten overs; losing chases show low variance. Winners go up and down. Losers glide smoothly at six or seven an over. Smoothness in the death is a symptom, not a virtue. In ODI cricket, the real death-over currency is not wickets — it is boundaries. Falling below one boundary per over in overs 41-50 caps you at roughly six an over. Needing 78 means needing at least 1.25 boundaries per over at a rate of 9.00. A side that cannot reach that density loses, wickets in hand or not, however much fight the eye reports. The flip over I ran the model across every chase to find a boundary: the over in which win probability first drops below 50% and does not return. The line is not sharp, but it is clear. Above 3.2 dots per over across overs 11-30, win probability sits at 27%. Below 2.8, it sits at 64%. When that ratio is set early, the flip arrives at 34.6 on average. Once required rate passes 9.00 after over 34, only 21% of those chases were won in this dataset. So I will make a claim and register it in advance. If dots per over in overs 11-30 exceed 3.2 and the powerplay score is under 45, I will price Bangladesh's win probability under 20%. I will test that indicator across the next four bilateral series. Contrarian Angle Emotional analysis puts the blame on batters — lack of intent, sleeping through the middle, no courage on the big stage. I do not accept that reading, and I have no reason to. A batter who reaches 41% dots after facing 25 balls is not lazy; he is trapped inside an optimisation problem where every safe option in front of him carries a very low expected return. The real drivers sit outside his will. One is the powerplay platform — a 14-run deficit asks the middle overs for something impossible. The other is the opposing spin matchup, which controls the dot-compression index almost directly. Across the 41 chases, whenever two spinners combined for 16 or more overs in that phase, the chasing side's run rate fell to 5.9. Which brings the necessary caution: correlation is not causation. Part of the elevated dots at Mirpur in overs 11-30 is pitch slowness, equally true for both sides. In 14 of the 26 domestic matches, the first innings also exceeded 3.2 dots in that phase. Some of it is environment, some opposition, some your own structure — the sum of three. There is a claim that pitches ease late and chasing is therefore preferable. My log does not support it. Chasing sides score at 7.1 in overs 41-50 when the target exceeds 270, and 7.8 below 270. On a good surface, the advantage runs the other way: the ball is older and the spinner's grip has gone. The eye test has a formal role, and I will name it. The eye is a strong hypothesis generator and a weak judge. Many times in the gallery I have felt the shift beginning; the model has said it begins only when the current mix of dots and singles changes. When the two disagree, I publish the disagreement, not a ruling. In seven of the 41 chases they disagreed outright; the model was right four times, the eye three. Those three are now the most useful inputs to my next recalibration, because a variable is missing from my feeder set. A model is a monastery: you enter with noise and leave with discipline; enter with discipline alone and the doors close behind you. Takeaway Next series, I will track three things, none of them on the scoreboard. One, runs per over and the dot ratio in the first ten — below 1.3 and the middle-overs maths is already dead. Two, the spin matchup between overs 11 and 16; how many overs of spin the opposition bowls and which batter is on strike. Three, the timing of the third and fourth wickets before over 30, because chases die late but they die first in the two unlogged overs.