The Open Ledger of T20 Data: From Mirpur 2026 Back to an A-League Spreadsheet

I went looking for the A-League and ended up parked at a Mirpur spreadsheet

1 September 2026. It was two in the morning in Brisbane. Bangladesh and New Zealand were into the first T20I at the Sher-e-Bangla National Stadium, and open on my laptop was a much older file — the spreadsheet I had built in 2026 cataloguing the underlying numbers of every A-League club. Two windows on one screen: the live match, and a decade of phase-wise cricket data — powerplay, middle overs, death overs, dot-ball share, boundary frequency.

The match was lost. What I did that night was not journalism but bookkeeping. I laid the first innings side by side: every dot ball in the powerplay, the spinners' economy, the runs saved in the field, the restlessness hidden inside a strike rate. The numbers said Bangladesh's powerplay was "normal." My eyes said something else entirely. When the arithmetic says everything is fine and the eye says nothing is fine, that is where the real work begins.

Bangladesh then won three of the next four matches to take the series 3-2 — their first T20I series win over New Zealand. The country celebrated. I sat down to check how many of those winning matches the underlying numbers had actually favoured. Two of them were closer to the opposition on the model. The table and the model disagreed. The higher the numbers climbed, the louder the old eye test laughed.

Context: the language changed, the question did not

T20 cricket now describes itself in a different vocabulary. Every franchise league — IPL, Big Bash, CSA, SA20, ILT20, BPL — has analysts, matchup coordinators, strategy coaches. Commentary has absorbed "intent", "power splash", "matchup win", "phase-wise run rate". Some BPL teams now make bowling changes in front of a laptop, and to the spectator in the stands those changes often look like riddles.

Here is the problem. Models built for flat, hard, true-bouncing surfaces in England or Australia are being imported wholesale onto the slow, low, gripping Mirpur, Chattogram and Sylhet pitches. The assumptions inside the model — boundary rates, six-hitting probability, the value of a second spinner — behave differently in a different environment. Data tells the truth. But if the question is wrong, an accurate answer is worthless.

The September 2026 series needs its background. Most of New Zealand's frontline players were inside the IPL's bio-bubble in the UAE; New Zealand sent a second-string squad captained by Tom Latham. A spin-friendly Mirpur pitch, evening dew, empty stands — it was a laboratory in which Bangladesh's tools looked sharp and the opposition's looked blunt. The series win deserves full respect. Treating it as evidence about the future is a separate decision, and that is where I pushed back.

Core: what the spreadsheet said, and what it hid

First finding: many of Bangladesh's dot balls came on their own terms, played safely. That is not bad cricket. But what was each dot actually buying? Weightless dot balls keep wickets in hand and also keep the scoreboard jammed. Dot-ball counts only speak when paired with runs per over and wickets in hand. A raw dot-ball number tells you nothing about whether a side will win — much like possession percentage in football, the most deceptive statistic in the game.

Second: powerplay boundary rates are always suppressed at Mirpur because the surface is slow. Even inside a suppressed number there is a decision to make — which ball to leave, which to attack. In that 2026 series, Bangladesh's middle-overs strike rate ran slightly ahead of the opposition, and that was the hinge. But "slightly ahead" here rests on five matches. Judging a team's batting philosophy on five T20Is is like forecasting an entire summer from one hot day in June.

Third: spin economy. Mirpur spinners bowl slowly, turn it, and force batters to survive low bounce. The economy looks excellent. But praise and capability are not the same thing. The same bowler on a flat deck loses that low-bounce pressure. A friendly environment makes talent look bigger; it does not make talent bigger. This distinction is the most consistently buried idea in cricket talk.

Fourth, my strongest objection: the habit of downgrading what the eye plainly sees. A yorker and a quality length ball are hard to separate in a dataset, so models treat both as generic dots. In real cricket that one-foot difference decides matches. I have watched deliveries of identical pace and identical line split by a foot of length and produce three different outcomes on the scoreboard, while my laptop logged three identical dots.

My 2026 experience maps onto this directly. That year I pulled the whole A-League file together and wrote that Brisbane Roar's 42 points sat well above their underlying 36.8 expected points, and that Jamie Maclaren's 19 goals had come from only 14.7 xG. That gap was the nucleus of the piece. It drew 180,000 reads and 2,300 comments — numbers that made me build a full spreadsheet of every club's underlying data and start chasing statistical anomalies instead of routine match reports.

But which question did the Maclaren case actually answer? Two possibilities: he is a better finisher than consensus allows, or he is running hot and the model will correct itself next season. Both in medicine and in cricket, the second happens far more often. The same caution applies to a T20 batter's expected runs — the metric may be telling you something real, but we rarely print how thin the sample underneath it is.

Add a side-story from that Mirpur series. With empty stands, the loudest component of home advantage — crowd pressure on the umpire — was absent, yet decisions did not become perfectly neutral. I have been collecting crowd-noise and officiating data for years; my own sheet is still raw. But the accepted truth stands: home advantage is not only the pitch, it is also human judgement. When a model explains a result, that human component usually drops out.

Then came the test I was waiting for. Off the back of that series, Bangladesh entered the 2026 T20 World Cup dreaming of a Super 12 upset. They lost all five Super 12 matches. The Mirpur strike rates, the economies, the "we can fight back" — all of it evaporated in the dry UAE heat. The reason is not complicated: the ball turned in Mirpur and did not turn there; the boundaries were short for middle-overs spin at home and long away; and the opposition was at full strength.

My football experience returns here. In 2026 I wrote that Germany would not survive their group, using xG evidence that the 2026 title was an outlier and the 2026 Confederations Cup win a false positive. Germany lost 1-0 to Mexico, then 2-0 to South Korea and were out. I wanted Germany to prove me wrong. Their group-stage exit proved me right instead. Bangladesh's 2026 World Cup is the same photograph: an idea built in a small environment meets a large one, and every crack shows.

Now the claim that irritates me most — "matchup data." Left-arm spinner against right-hand batter always looks clean in numbers. The numbers do not say how much the ball is turning, whether the batter can use his feet, how strong the wind is. Mirpur turns in September evenings; a January morning behaves differently. A matchup is an average, and cricket is not played in averages — it is played ball by ball.

One more thing gets quietly removed from the ledger: fielding. Runs saved is a lovely metric, but a weak fielder simply mispositions himself and nothing appears on the scoreboard. The run not taken is never recorded. That is the darkest room in my own dataset.

Contrarian: where I might be wrong

An honest column needs to dig a hole in its own floor. My weakest point is sample size. Five matches and one World Cup are both small. If I build a theory that Mirpur success is illusory from those two samples, I am committing the sin I accuse others of. Separating form from randomness in cricket is hard; twenty to thirty matches are needed before judging anyone.

Second, my eye test is compromised. I watch from Brisbane, on a streaming feed, at odd hours. I cannot feel the Mirpur air, the pitch, the light, the singing. My data is right in some places and my eyes are wrong in others. A writer who leans only on the eye test commits the same error with a different name.

Third, there is a genuine counterexample. Teams that truly walked the data path — bowling changes by matchup, planned risk in the powerplay — have delivered results consistently. Their success is not theory, it is evidence. My quarrel is not with data. It is with the data user who accepts the output without asking how the input was built.

Takeaway: a test you can run tomorrow

If Bangladesh play five T20Is away from home and their powerplay dot-ball share stays above 45 per cent as it does at home, I have to concede that this team's identity is strategy, not surface. If the dot share falls into single digits while the strike rate climbs, it proves my spreadsheet was talking all along — and that I was listening to the roar of Mirpur instead.

I went looking for the A-League and stopped there. The league changed. The data did not. Neither did the question. Numbers never lie, but numbers alone never tell the truth. Watch the powerplay next series. Whether a ball is bowled a beat late may be worth more to you than the result.