Cricket Analytics' Verification Crisis: Three Widely Quoted Numbers That Fail the Public Record

In a Champions Trophy match the big screen flashed a number: Bangladesh's powerplay run rate, 6.8. Sitting in the third row of the press box, I opened my own log of the same 32 matches. Mine read 7.4. A gap of 0.6, not the kind of thing that stops a conversation. But that 0.6 carried almost the entire argument behind a decision—whether the top order would be broken up, or the same batters handed another series.

A week later, an analytics deck had to be sent back. Its first page carried three numbers, and none of the three matched the public record. One had the wrong source, one the wrong denominator, one the wrong sample. Get one number wrong and one decision goes wrong. Get three wrong and the whole story goes wrong—and in cricket, the story is now the biggest product on the shelf.

That Bangladesh bats slowly is one of the most protected beliefs in the sport's analysis industry. The belief is not evidence-free. Bangladesh won two of nine matches at the 2026 ODI World Cup. At the 2026 T20 World Cup they lost all three Super Eight games. Year after year the conversation returns to how hard runs are to score on the spin-friendly Mirpur surface. Anyone holding a scorecard would say the problem is real, and large.

I am not challenging those numbers. I am challenging the method by which numbers become decisions. Television graphics, fantasy-league dashboards, the one-page summary placed in front of a selection panel—the same kind of figure circulates in all three places, and none of them carries an audit trail. My test is simple: a number earns the right to drive a decision only after it passes three questions—where is the source, what is the denominator, how large is the sample. Call it the data-integrity gate. Three numbers from my own submitted deck have just failed it.

The log I have kept from the ground over the years—now past 1,400 matches, ball by ball, tagged by hand—is my only receipt. In 2026 I opened Excel to test a hunch, and a religion called possession died. The lesson came from football, but the method transfers: a number without a denominator and a stated falsification condition is not proof, it is decoration.

Number one, the blended strike rate. The trouble starts in the denominator. Splitting Bangladesh's last 40 T20Is in my log gives a strike rate of 118 in innings batting first and 126 in innings chasing. The practical difference is not enormous, but the reason matters: a chasing side knows the target, so the risk calculation is clean. A side batting first plays toward a par score, and on a slow wicket that par is itself low. Broadcast blends the two into one figure, then reads that figure as a character flaw. A moral verdict cannot be extracted from a blended number; only a wrong verdict can.

The second problem is the opposition mix. Inside a single strike rate sit innings against associate bowling and innings against top-five attacks. Change the mix and the number changes, while the verdict stays exactly where it was—because the verdict never came from the number. It came from habit.

Number two, the omnipotence of the powerplay score. This is my favourite false belief, because the supporting statistics are easy to find, and that is precisely the danger. In my log, the side with the higher powerplay score wins roughly 57 percent of T20s. Fifty-seven sounds excellent until you ask how much information the figure adds once venue par and wickets in hand at the six-over mark are removed. The answer is very little. Powerplay runs and match results both flow from a third thing—the character of the pitch and the quality of the bowling attack. A number that is the shadow of another number should not be given a job in selection; that is employment for a shadow.

Afghanistan at the 2026 T20 World Cup is the cleanest illustration. In the match they beat Australia in Kingstown, the foundation was pressure with the ball and a match-up plan, not an elite powerplay score. Anyone who watched that tournament remembers Afghan batting seizing up more than once; the side still reached the semi-final. Bowling plans decided the results, not powerplay graphics.

Number three, the magic of the all-rounder and the captain. Shakib Al Hasan's record is the most frequently deployed proof here, and the figure is genuinely unprecedented: more than 700 international wickets and more than 14,000 international runs, a combination nobody else owns. But a pipeline cannot be run on one exception. The question is not about Shakib; it is about where the second, third and fourth all-rounders of that standard are—a question about the selection record. The answer sits inside structure: the domestic calendar, match density, and the supply of bowling overs a young player needs to become an all-rounder are all limited. When a Mirpur Test calls for four spinners, nobody gets the chance to prove himself as the fifth bowler.

Captaincy falls into the same trap. A captain's win percentage is really a map of the fixture list. On home wickets built for spin, almost anyone looks like a good captain; abroad, against equal opposition, the same captain's numbers collapse. A captain's win percentage is a schedule's win percentage—a quality of the calendar, not of the player. Mushfiqur Rahim's 200 at Galle in March 2026, Bangladesh's first Test double century, is a story in our memory; in the ledger of numbers it is a sample of one. So is the first Test win, against Zimbabwe in Chittagong in 2026. We explain two decades with those two events, and both rest on a sample of one.

The sample problem bites harder in Bangladesh because the total number of Tests is small. Cut the data any way you like and the cells hold eight or ten matches. A decision built on eight matches is not strategy, it is a coin toss. Here is my second discomfort: the analytics industry wants fast answers, and to make an answer fast, a small sample has to be dressed up as a large one.

Then there is the cult of intent. In Bangladeshi debate, intent has become an unfalsifiable theological term. A batter gets out and we say the intent was missing; he plays well and we say the intent has returned. A concept that cannot fail is not analysis, it is a blessing. A test nobody can fail is not a test; it is a compliment.

There is one more layer that never appears on a graphic: the schedule. Bilateral T20I series are now revenue projects, not development projects. The statistics we argue over are generated by a calendar written under broadcast contracts. The number was a commercial sample before it was ever read as a player's character. Where structure fixes the inputs, moral lectures about the outputs are dishonest.

I have no complaint against graphics. When a magic show refuses to publish the knots, that is not fraud—that is the format. My complaint belongs where the same graphic is carried into a selection meeting and given weight it should never have had. Television's job is entertainment; a panel's job is decision. Two different professions, one shared number—and that is exactly where the integrity gate gives way.

I could be wrong, and in places I probably am. My 1,400-match database is a convenience sample: one laptop, one pair of eyes, and whatever matches I could find the hours to watch. Too many Bangladesh games, too many home games, too few night games. My tagging carries my own bias. It is entirely possible that the broadcast's 6.8 is right and my 7.4 is wrong—a missed no-ball, a different definition of the powerplay, a wide counted on the wrong side. So I am registering the condition in advance: if a third party reproduces 6.8 from the same raw ball-by-ball data, my claim is dead, and it will carry an F grade in the public receipts file.

And here is an uncomfortable possibility: the mainstream verdict may be right even though its numbers are broken. Bangladesh really does bat slowly—that conclusion can survive, even as the argument supporting it falls apart. In cricket we routinely pardon a method because the result pleased us, and the next error is born precisely there.

Before the next home Test series begins, I will publish the full receipts file—every number, date, denominator and source. Whatever fails verification gets struck through in red ink. The question is no longer whether Bangladesh bats slowly. The question is who audits the analysis that is doing the deciding.