The Umpire's Call Ledger: In Asia, the Machine Testifies but Never Judges

On the ball-tracking screen a pale purple half-ellipse travels towards the stumps and stops. Frame by frame, the ball will graze the inside edge of leg stump but strike less than half of it. Umpire's call. The on-field verdict stands, not the projection. Four words appear on the big screen: On-field decision stands. The check ran two minutes and fifty-two seconds. In that time the crowd learned that the technology knows the truth, but has no right to say it.

I was not watching the scoreboard. I was counting. How many reviews on Asian soil this season had ended in exactly that place, where ball tracking confirmed where the ball was going but could not confirm whether it was certain enough to decide.

The commentary box filled with the language of grievance rather than the language of technology. Somebody said shame. Somebody said if the technology exists, let it judge. I have heard that sentence since 2026, when I watched all sixty-four matches in Russia twice and built a ledger of 455 checks. Eight years on, the question is the same. The answer has moved further away.

Context: here the machine is a witness, not a judge

DRS began in July 2026 in Colombo, in the India–Sri Lanka series. The promise was simple: fewer human errors. Fifteen years later the arithmetic is far messier. The system now stands on three layers.

First, the ball-tracking camera network, which draws a trajectory from a handful of frames. Second, the audio sensors—UltraEdge, Real Time Snicko—which report whether the ball touched the edge. Third, the projection model, which announces the ball's future. The first two testify. The third passes judgement, but does not write a verdict. It writes a probability.

The rule is this: on impact with the stumps, if the projection shows more than half the ball hitting, the decision is overturned. Less than half, and the on-field call survives. That half-ball threshold is what we call umpire's call. In May 2026 the ICC changed one thing: the soft signal was removed, so the on-field umpire's opinion no longer carries into low catches. It took effect from the World Test Championship final.

What has not changed: two reviews per innings, retained on success, and the equipment itself. Ball-tracking versions differ by tournament, because suppliers differ. ICC events run Hawk-Eye; some bilateral series run a host broadcaster's alternative. No press conference mentions which build is live. A technology that decides a semi-final gives the viewer no right to know what changed inside it.

Core analysis: where the error is born

After years of watching and modelling, the gist is this: ball tracking does not measure the point at which the ball strikes the stumps. It is a network that watches a few frames from roughly two metres away and imagines the ball's future. Error is born in three places—bounce-point detection, the estimate of deviation after pitching, and the estimate of stump height.

In Asian conditions the first of those is the most dangerous, because the more sharply the ball turns after pitching, the more a tiny error at the bounce point returns as a large error at the stump line. When a spinner lands on the same length, the ball travels less than two metres after pitching before reaching the stumps. Two millimetres of doubt in pitch detection becomes twenty millimetres at the stumps. At Perth or the WACA the ball flies three metres after pitching; there the error is washed out, because the model has far more room to correct.

Same batter, same line, two different geographies of certainty. Umpire's call arrives more often in Asia not simply because of low bounce, but because of that error amplification. The same projection model that is seventy per cent confident in Europe drops into the fifties on a subcontinental turner.

The second problem is visual. What the viewer sees is a line—precise, firm, confident, a purple streak that thuds into the stumps. What actually exists is a cloud of probability with frayed edges. Broadcasters do not draw the confidence interval, because clouds do not look good on television. Show the cloud and people will realise the decision is an estimate, not a verdict. When the margin decides who stays, the frame rate is a moral position.

The third problem is structural. The pattern that keeps returning in my ledger on Asian soil is not about ball tracking but about review strategy. Big teams burn reviews on hunches; smaller teams save them and lose the atmosphere. With Bangladesh I have seen the same sample repeatedly: nearly intact through the first two sessions of a Test, two reviews gone back-to-back in the third, and nothing left in hand for the over that matters.

One more pattern, which I offer as pattern rather than proof: by my own grading in the 455-check ledger, where the stadium is loudest—Mumbai, Dhaka, Karachi, Kolkata—on-field calls tend to flip from out to not out more often. The mechanism is not mysterious. Crowd pressure and media pressure push umpires towards uncertainty. Favour towards a bigger side is not a conspiracy theory; it is the ordinary result of atmosphere. A caution belongs here: my sample is limited and I do not hold the internal audio of every controversial review. What I have is a pattern, and a pattern demands inquiry, not a verdict. I do not count checks to find blame. I count them to find patterns.

Small ledger entries still decide who stays. Over-rate fines, points deductions, net run rate—these decimals push teams down the table. Transfer windows and relegation tables both hide their verdicts in decimals. It was the same in 2026: unweighted points per game sent Tranmere Rovers down, by 0.09 points per game. The table is a verdict, not a story. The out-or-not-out boundary in cricket is the same thing—a decimal with a career resting on it.

One thing data never leaks but the eye does: time. Two minutes and fifty-two seconds is not only a crowd waiting; it is a bowler's rhythm breaking. When a spinner stands over two full overs waiting for ball tracking, what he loses in the next over never appears on any scoreboard. In Tests the loss accumulates quietly; in T20 it is instant, because one over is the match.

Contrarian angle: abolishing umpire's call would not make the system honest

The loudest demand is to scrap umpire's call: let the line that hits more of the stumps win. I reject it outright, and the reason is mathematics, not conservatism.

Umpire's call is the only part of the system where the machine admits it is not sure. The half-ball threshold is an honest confession—it says the error band in my model is wide enough that I will not sign a boundary-line decision in my own name. Remove the threshold and the error band does not disappear; it simply loses its owner. Deciding at fifty point one per cent is not accuracy. It is the theatre of precision.

The real gap is not in the verdict but in the disclosure. Broadcasters do not show three things: the confidence interval of the projection, the detection error at the bounce point, and the software build used in the match. Without them, the argument about reviews is not a dispute about evidence. It is a war of belief.

One experience of mine, from another sport, belongs here. On 1 December 2026, on the data floor in Qatar, I watched 1.88 millimetres decide the fate of a season—inside the line or outside it. That day I understood that placing technology beside a referee is not about silencing one of them; nor is it about seating the machine in the judge's chair.

Takeaway

My proposal is simple: the ICC should publish an open review ledger. Beside every review, the supplier's name, the software version, the confidence interval of the projection, the detection error, and the duration of the check. Add those four columns and the argument about umpire's call will shrink, because people will no longer dispute belief. They will dispute numbers.

I began with a 455-check ledger and ended with a question about the watchers. When the technology testifies, who delivers the verdict? The answer is not tactical. It is ethical—whoever delivers it must also publish the size of their own error.

— A referee — Root: Archetype Referee