The Invisible Number in the BPL Auction: Middle-Overs Strike Rate Against Spin

In the last BPL auction, one name went for more than four crore taka. In the same season, another batter scored at a strike rate of 142.6 against spin in the middle overs — overs seven through fifteen. Nobody bought him. I had placed both names side by side in the event dataset I had coded by hand. The difference, it became clear, was not talent; it was arithmetic. One had the bigger story, the other the bigger number. The auction rewarded the story. The talk around the Bangladesh Premier League auction is dominated by two kinds of names — overseas stars and power-hitting openers. The reason is understandable. In the first six overs of the powerplay the ball comes onto the bat best, and the sixes there glow brightest in the highlights. But a T20 result is actually decided elsewhere. In a twenty-over innings the powerplay is six overs, the death the last four or five; the remaining nine or ten are the middle. Yet in BPL auction strategy you almost never hear of a batter bought specifically for those nine or ten overs. In 2026 I hand-coded 1,200 events from 24 BPL matches. I watched every match twice — once with my eyes, once with my notes. Line and length, shot direction, part of the bat, type of assist — I tagged it all. The reason was simple: there was no ready API for this league, no standardised scoring database. No API, no shortcut — just ninety minutes of keystrokes and a little monk's patience. A number I have not verified myself is a number I will not speak from. This gap is not the BPL's alone; it belongs to the whole Asian T20 market. In the IPL, with large data teams, the middle-overs split is slowly earning a price; in the Lanka Premier League or ILT20 that practice is still a distant idea. The BPL stands between these two poles — where the cricket's emotion is Asian but the infrastructure is still handwritten. The comparison is uneven, yet the direction is clear: where infrastructure exists, information earns a price; where it does not, the story earns one. In that same dataset a pattern surfaces. Batters who can work the ball against spin in the middle overs and find the boundary often carry a strike rate higher than the openers'. But they are never given the "finisher" label, because they don't hit sixes — they hit fours. Fours don't sparkle in highlights, so they fetch a lower price. Consider two middle-order batters from the same season. The first, against spin in the middle overs, faces 180 balls at a strike rate of 138.9, one boundary every 9.4 balls. The second faces 140 balls at 121.3, but in the death overs his strike rate is 175. At auction the second goes for roughly double. One reason: the death-overs number is examined separately, while the middle-overs number is stored nowhere. The real work of batters like Mushfiqur Rahim or Towhid Hridoy happens in exactly this phase. Here lies the real gap. Death-overs figures occupy their own column on every live scorecard — first six, last four. The middle nine or ten overs sit in no one's column. So the batter who turns a match in that phase has his work captured in no frame. Count separately the forty-five to fifty balls of a middle phase and each small win — the single turned into two, the pressure on the spinner — and you see that the innings is founded right here. The problem, then, is not only cricket intelligence but infrastructure. In API-driven leagues the middle-overs split comes ready-made; here we must build it by hand. By hand means slow, but by hand also means accurate. I coded the BPL by hand before I trusted its numbers — because in this market provenance is the story, not the footnote. Correcting every faulty entry taught me that a dataset's reliability comes not from external validation but from internal consistency. I do not measure middle-overs balls in runs alone; I measure them in win probability. To every event, from the first ball of an innings to the last, I attach how much the team's chance of winning moved. It turns out a quiet middle-overs single — one that never reaches the highlights — can sometimes add more probability than a six. Because a six is often followed by a wicket; a single leaves the batter at the crease. Middle-overs batting is not only runs but pressure. A batter who rotates strike at two runs an over pulls the fielders in and forces the spinner to change his length. The result arrives later — in the death overs, ball after ball, the field set, the extras. In other words, middle-overs patience is really an investment for the death. The side that survives the middle cashes out at the end. Here is another bind. Franchises often push a young batter into the middle order whose body is not yet finished — young in years but rushed into senior rhythm. Early-maturing youngsters blaze for a few seasons, then fade through injury and loss of rhythm. The event data shows these batters' middle-overs strike rate falls sharply in the back half of a season. The problem is not talent; it is the schedule of use. This argument is rarely heard at the auction table. Selectors trust what they see with the eye — and the eye reads the middle overs as slow, the death as fast. A squad's arithmetic thus stalls on two questions: who hits in the powerplay, and who hits in the last over? The nine overs in between are left to duty. Yet it is in those nine overs that an innings is won, or lost. The price gap between an overseas name and a local middle-order batter tells the same story. In a given season an overseas batter's contract is often three or four times a domestic one's, while in venue-adjusted middle-overs strike rate against spin the difference between them is slight. The price is set by experience and publicity, not performance. A caution, though. Treating middle-overs strike rate as direct proof of quality would be a mistake. Correlation is not causation. That high strike rate may come on a flat pitch, on a day of weak attack, or on the back of one or two big innings. In a small sample of three or four innings a batter can look like a giant — which happens in the BPL almost every season. When I built my first model from 24 matches of data, I kept one warning in mind: when the ball count is low, strike rate is noise, not information. Another trap is home conditions. Across BPL venues the use of spin, the bounce and the amount of dew differ. At home a batter's middle-overs strike rate swells, then returns to normal away. Compare without venue adjustment and we are really measuring the pitch's strike rate, not the batter's. The model therefore learns to weight the context, not only the player. Data itself is a trap. A bigger number makes it feel as though the answer has been found; yet many middle-overs balls are "dead" — the outcome was already settled. Count those balls and the strike-rate story changes. So I always separate the "contested ball" — the one on which the match genuinely turns. Without that separation we are measuring the wind. Still, arithmetic has a limit. A model without a decision is a diary, not a weapon. A strike-rate filter that never reaches the auction table simply sits in a file. The real crisis of infrastructure, then, is not the lack of measurement — it is the lack of converting measured information into decisions. If I ran a franchise at the next auction, I would demand one column before all others: each batter's runs per ball against spin in the middle overs, adjusted for venue. Because the real question is not money; the question is — who takes the nine overs in the middle? The side that can answer it will not merely buy names; it will buy innings. And that question today belongs less to the batters than to us, the scorekeepers.