Two Minutes, One Clock, and the Three Numbers That Failed the Scorebook

Arun Jaitley Stadium, Delhi, November 6, 2026, the 25th over of Sri Lanka's innings. Angelo Mathews has just walked to the crease, his helmet strap snaps, he signals towards the dressing room for a replacement. Marais Erasmus raises his finger. Timed out. The first such dismissal in the 146-year written record of international cricket. I watched it in a café in Sylhet on a buffering feed, and the first question in my head was not about morality. It was about a clock. Which clock? Whose clock? Who started it, and who stopped it?

Forty-eight hours later a Stage-1 packet landed on my desk: forty-one claims, each with a source, each with a confidence rating. I began with 1,058 incidents, and the anomaly was hiding in plain sight — three of those claims do not survive cross-check against the public record. One number, one timestamp, one action. This is the ledger of those three failures, and the larger question nobody asked: what does cricket's official record actually record, and what does it refuse to record?

A Stage-1 packet is the raw post-match data bundle — scorecard, commentary log, broadcast timecode, vendor datasets, brief notes from the match officials. Stage-2 is the scalpel laid against it. But before the scalpel, you have to ask what authority the packet rests on.

Cricket's hierarchy of authority is written in the language of Law, and it stops at Law 3 — the scorers. The MCC Laws of Cricket, 2026 Code, state that two scorers keep separate records and must agree at the end of each innings and the match; only agreement produces a record. That arrangement has survived almost unchanged since the Articles of Agreement of 1744.

Here is the point worth making explicit. Cricket's scorebook is the world's oldest append-only ledger, and it is the least discussed technological achievement in the sport. Old entries cannot be deleted, only appended; two independent validators must reconcile; errors require a written correction rather than a quiet edit. Immutability, distributed validation, audit trail — the three supposed virtues of a blockchain have existed in cricket since 1744 under a different name.

The problem is that this ledger records outcomes, not processes. There is a cell for 'out — timed out'. There is no cell for when the batter reached the crease, whether an appeal was made, whether Time was called. Law 2 recognises the umpire as the sole judge of fact, but no Law provides a compartment for preserving the evidence of what his watch showed. The resulting silence is filled by the broadcaster, and the broadcast timecode acquires the status of record without anyone announcing it. The Stage-1 packet was using that timecode.

There is another layer nobody audits. The match referee's report — over rates, conduct, pitch comments — is not published in full; a summary appears, the document does not. So when a data packet cites an 'official source', the reader has no route to verification. Transparency stays a slogan and never becomes a process.

A larger example exists that nobody frames as a data-integrity question. The DRS 'umpire's call' — how much of the ball must be striking the stump — is fixed by a convention, and the number inside that convention is never shown on air. The viewer sees 'hitting leg stump'; the record says 'umpire's call'; the gap between them lives in no scorecard. Same architecture, same silence.

Failure one: versioning. The packet stated the timed-out limit as three minutes. The number is not wrong; it is stale. Law 40.1.1 of the 2026 Code set three minutes; the 2026 Code reduced it to two, and the 2026 World Cup playing conditions applied two. The packet cited the 2026 Code and presented the claim in the present tense. Every line of software carries version control; cricket's law-derived datasets carry none. The difference between two minutes and three — a difference capable of ending an international career — disappears quietly inside a packet.

Failure two: the timestamp. The packet claimed the batter was 2 minutes 5 seconds late. The pose of precision is attractive; the foundation is weak. The figure came from broadcast timecode — when the camera captured the frame. Law 40 says 'within two minutes' and says nothing about which instrument measures it. The fall of the wicket, the replay, and the pressing of a stopwatch are three separate events.

— Root: The Silence Dataset.

Building the behind-closed-doors dataset in 2026 taught me this: what the record does not hold gets filled by inference, and inference dressed as a number never faces a question again. '2 minutes 5 seconds' is unverifiable because the instrument of verification is absent from the record. The error here is not a lack of precision but an excess of it — finer than the evidence.

Failure three: the action. The packet said the umpire gave the batter out on his own initiative. Law 31.1 makes that sentence impossible: neither umpire may give a batter out unless appealed to by a fielder, even where the batter is out under the Laws. The appeal is the spark without which the dismissal machinery does not turn. The public record contains evidence of an appeal; the scorebook has no cell for one. The packet's author saw a void and filled it, and the filling was wrong.

I have seen two vendors produce two different over rates for the same match — one using the DLS-reduced overs, one using the scheduled overs. Both are right; both answer different questions. Nobody documents the divergence, because vendor contracts have no field for version numbers.

None of the three failures is a lie. Each has a defensible explanation: an older Code, a broadcast clock, information the record never held. Dangerous data is not false; dangerous data is plausible. Falsehood gets caught. Plausible error survives for decades.

My problem is that I stop when I see an anomaly. Stopping reveals that the real anomaly is not the first failure but the second, because the first and third are fixable and the second is not. A version number repairs the first; a cell for the appeal repairs the third. But where the record holds no clock, a claim about a clock can never become proof.

This is where my own profession tells me to distrust itself. Almost the whole argument after Mathews was about morality — the spirit of cricket, sportsmanship, who said what first. Law against emotion. The rulebook gave me a verdict; the freeze-frame gave me a question, and the question was not about emotion. It was about measurement.

The blind spot is this: Law 40 states a limit on time and says nothing about proof of time. Where a Law does not specify its own method of verification, each side can pick its preferred clock, and both sides can be correct under the Law. The argument was never about the characters of cricketers; it was about the authority to measure, and nobody admitted it. That is also where institutional suspicion belongs: where the record holds no clock, the owner of the clock owns the record. The broadcaster supplies the feed, the feed supplies the timecode, the timecode becomes 'evidence', and nobody answers for it. This is not conspiracy; it is incentive. Nobody is hiding anything. The compartment was simply never built.

The way forward does not need a grand reform. It needs one small annex to the scorecard: time of arrival at the crease, record of the appeal, whether Time was called, and which clock was used. Four cells. Fill those four cells and the next timed-out argument descends from who was politer to what the clock said.

The referee's eye is less a gift than a burden of proof. And the first condition of proof is knowing where the proof is kept. Law 3 calls the scorers the guardians of the record; but if the record has no compartment for the clock, who keeps the clock's account?