The Darkness of Dot Balls: How the BPL's Middle Overs Build the Table

Hook

Last season in the press box at Sylhet International Cricket Stadium I watched something uncomfortable. At the end of the 13th over the scoreboard said the chasing side needed 84 from 62 balls with seven wickets in hand and two set batters at the crease. The league baseline win probability for that position is 33 per cent. My ledger said 41. Then overs 14 to 17 produced 23 dot balls. The batting side did not lose a single wicket. It simply ate deliveries, five or six per over. By the start of the 19th over the equation was 34 from 18, and three overs later they had lost two wickets and the match by nine runs. The scoreboard wrote a thriller. My column wrote three numbers: 23 dots, zero boundaries in 17 balls, and a required rate that jumped from 6.8 to 9.2. Matches in the BPL are not lost in the powerplay. The ones that look close at the death were mostly lost silently, somewhere between overs 7 and 15.

Context: how the ledger was built, and where it stops

In 2026, sitting in Sylhet, I built the BPL's first expected-runs ledger, and the numbers rewrote the story of the game. It started with a ball-by-ball archive of 132 matches, 14,804 legal deliveries. Every delivery is broken into six variables: venue, innings number, phase (1-6, 7-15, 16-20), wickets in hand, bowler type, and a dew flag. Because there is no tracking camera in the BPL, my xR is derived from outcomes rather than trajectories. I publish the limitation every week: without trajectory data I work with shot outcomes, not shot quality, and without a stadium effect Sylhet's flat deck inflates the error band by five to seven per cent. A spreadsheet is a monastery, and I take my vows in rows and columns. I also trained two junior loggers to record shot coordinates and line-and-length type, so the ledger does not live inside one man's memory. I do not chase results; I audit the process until it confesses.

The central index in that audit is the Dot Pressure Index (DPI) — dot balls per over between the 7th and 15th, measured against the league baseline of 4.1. It works like PPDA in football: a measure of how much pressure you create in areas the camera ignores.

Core: what the data says

First finding, and it maps straight onto the table. Across the last five seasons, teams finishing in the top four averaged a DPI of 3.2 in overs 7 to 15. Teams in the bottom three averaged 5.0. Cutting one dot ball per over in that window is worth roughly two league points a season. That is arithmetic, not sentiment.

Second finding, more specific. Top-four sides played an average of 2.1 boundary-less overs before the 17th over; bottom-three sides played 4.6. Batters like Towhid Hridoy sit at the centre of this pattern because his rotation does more work than his hitting: last season, in overs 7 to 15, he scored 89 runs per 100 balls, and 61 per cent of those runs came in singles and twos, against a league average of 74 in the same phase. Mushfiqur Rahim reads the opposite way when he bats high: he survives those overs, but his boundary rate drops from 9.3 to 7.1. Two styles of rotation in one phase — one keeps the tempo, the other only protects a wicket.

Third finding, where decisions actually live. In the second innings, the league average required rate in overs 16 to 20 is 9.8. When the DPI across overs 7 to 15 stays below 3.4, chases succeed 63 per cent of the time. Once the DPI crosses 4.5, success drops to 28 per cent. Even with seven wickets in hand. Wickets in hand is a seductive number; boundary slots are a more honest predictor.

Fourth finding, from the bowling side. Middle-over spin choke is the cheapest weapon in the format. A leg-spinner like Rishad Hossain holding a 6.4 economy in overs 7 to 15 with roughly 0.9 wicket-expectancy per over breaks the opposition's xR chain fastest. By contrast, Mustafizur Rahman's death economy sits at 8.1 against a league average of 10.4, but that saving only matters if the 15th-over score is already controlled. Death magic cannot settle a middle-over debt.

Fifth finding, about individual overperformance. One side in one season outscored its xR by 61 runs, and roughly 44 per cent of that gap came from a set batter's finishing skill, which sits outside the process model. It deserves credit; it does not deserve a place in a forecast.

Contrarian: correlation is not causation

Here is my loudest warning. A low DPI does not mean a side is good. Three rival explanations sit behind any fall in middle-over dot balls: batting intent, the nature of the venue, and the quality of the bowling faced. Sylhet's flat surface suppresses dots naturally; Mirpur's slower turners inflate them. Dew changes the grip and shortens the spinner's line. When I run a multivariate model with venue fixed effects and innings number, the explanatory power of DPI falls from 29 per cent to 14 per cent. It does not fall to zero, and it stays above the league baseline.

The second trap sits in the auction. Prices rise around finishers because sixes are visible and memorable, while a middle-order rotator's marginal win contribution is often larger and invisible. I keep the process model and market-implied probability strictly separate; mixing them spreads the error in both directions. The market is not a perfect probability engine, but the price it sets has a structure, and ignoring that structure means you are only guessing.

Takeaway

In the next round I will watch one thing: who can push dot balls below three per over between the 7th and 15th of a chase, especially who still wants the single against spin. The day you see a side with seven wickets in hand play four or more boundary-less overs before the 17th, you will know where they lost — not in front of the scoreboard, but in the darkness of dot balls. The question left hanging is simple: is a league table written in runs, or in 42 quiet deliveries?