What the Sylhet Ledger Says: Bangladesh's Middle-Overs Expected Runs and the Result Gap

Last week, from the press box at the Sylhet International Cricket Stadium, the number on my laptop at the end of the fourteenth over read 38.4 percent — the batting side's win probability, as estimated by my expected-runs model. Dew had not yet arrived, but dot-ball pressure had: eleven consecutive deliveries produced only two boundaries, and strike rotation was almost nil. The side went on to win with six balls to spare. The scoreboard will record a clean victory. My ledger records something else: most of that win was manufactured in the final four overs by two batters taking personal risk, not by any structural advantage.

I built the first xG ledger in Sylhet, and the numbers rewrote the game. In 2026, while sitting at PitchMetrics Asia and assembling a model from 132 Bangladesh Premier League matches and 14,800 shots, I learned that Abahani Limited Dhaka's overperformance of 14.2 goals was clinical finishing rather than fortune. In cricket I run the same architecture under the name expected runs: before each delivery, the model weighs wicket, over, field setting and pitch state to estimate what that ball should yield.

A spreadsheet is a monastery, and I take vows in columns and rows. The basis of this piece is my own ledger, not the official scorecard: ball-by-ball logs from more than 2,400 innings across nine domestic venues, a pitch-pace index, a dew index and average outfield speed. The limits are just as explicit. Not every venue has ball-tracking, so some data is logged by hand; the model carries an error band of plus or minus 4 to 6 percent; and selection bias exists, because where cameras are fewer, data is thinner.

Through the current domestic season one signal keeps returning. ICC records show Shakib Al Hasan's 246 Test wickets are the most by any Bangladesh bowler — a record that is the product of one generation's bowling culture, and that culture is now raising fresh questions about the middle overs of the batting innings. I do not chase results; I audit the process until it confesses. Overs seven to fifteen are the quietest zone of Bangladesh's batting innings this season: on average 2.8 dot balls per six deliveries, against roughly 2.1 in overseas series. The gap looks small, but 27 extra dot balls in a single innings means more than four and a half overs of wasted deliveries.

Dot-ball pressure is an intermediate variable, because dots accumulating over after over force the batter into lower-probability shots at the death. My log shows that in innings where the dot-ball rate between overs seven and fifteen stayed below 45 percent, the six-hitting rate in the last five overs rose by about 38 percent. Risk does not disappear; it is deferred, and it returns more expensive.

The boundary-dependency index points the same way. In the middle overs, 58 percent of a side's total boundaries come off the bats of just two players. The structural picture is simple: when one of them falls, the expected-runs curve drops steeply through those overs and the load shifts onto the finishers. On domestic pitches where spinners change pace every over, the rotation skill of the number three and number four is what actually separates sides.

The Sylhet-versus-Mirpur comparison is decisive here. By my reckoning Sylhet's outfield is quick, so ones become twos: 0.38 singles per ball, against 0.31 on slower outfields. Yet in Sylhet evening matches, dew lifts expected runs by 0.18 per ball after the sixteenth over. Read those two numbers together and a clear tactical decision emerges: winning the toss and batting first in Sylhet means starting two steps behind.

The bowling ledger shows a similar fracture. When Taskin Ahmed and Nahid Rana hold a hard length with the new ball, the opponent's false-shot rate climbs; Nahid's pace reaches close to 150 kilometres per hour, and that extra speed takes away the batter's reaction time. Trouble begins after the sixteenth over: the fifth bowler's economy this season is 9.6, and that is precisely when the opponent's expected runs rise fastest.

Physical condition sits inside this picture too. Day matches in Sylhet regularly push the heat index past 36 degrees, and between the thirtieth and fortieth overs the speed of twos drops by roughly 7 percent. What looks like a lapse in concentration is often arithmetic about heat and fluid management, even in a competitive domestic setup.

The transfer market is not a bazaar; it is a probability engine with agents. A domestic auction prices a middle-overs batter on strike rate, while his contribution to win probability comes mostly from how rarely he absorbs a dot ball. In my model, players in the top quartile of the rotation index frequently out-contribute bigger names who spend deliveries in the middle overs.

That skill cannot be taught at the last minute. From more than thirty years of watching from the stands, the two most underfunded areas are age-group rotation training and coach education. An academy that teaches six-hitting shows visible results quickly; an academy that teaches the two-run sprint and the decision to leave the ball shows results six years later, in a ledger. The gap between market price and team need therefore persists.

Now I throw the counter-question at myself. My model may weight dew too heavily, and the pitch-index sample is small. A relationship between dot balls and defeats is not causation. The reverse is plausible: a side falling behind bats defensively, and that defensiveness manufactures the dots. If the scoreboard is truer than the process, I have to concede that too — in this format, deferring risk to the death may be the most rational choice available.

Empty stadiums taught me that silence has its own expected goals. Over the next three matches I will watch two things: the dot-ball rate between overs seven and fifteen, and the rotation speed of the number four. If the dot-ball rate does not fall below 43 percent, the market will price this side down next series however many wins arrive. The question is not victory or defeat; the question is which confesses first — the scoreboard or the ledger?