Thirty-Six Millimetres: The Tolerance Band Where Cricket's Truth Hides

Thirty-Six Millimetres: The Tolerance Band Where Cricket's Truth Hides

July 14, 2026. Lord's, a few minutes short of seven in the evening. England need 15 from Trent Boult's final over, and Ben Stokes is fighting alone. The fourth ball is dragged towards deep midwicket. Martin Guptill picks it up and throws at the stumps, and in that instant the ball strikes the diving Stokes' bat and deflects away to the third-man boundary. The umpire's arms go up, a circle is drawn in the air, six runs go on the board. The match slides into a Super Over. The trophy goes to England.

That six has been argued about ever since. But the argument is not about the number of runs. It is about protocol. I have watched that passage from three angles — the broadcast main camera, the third umpire's feed, and a low angle set parallel to the pitch. All three angles are crisp. None of them could answer the one simple question that would have settled it: had Stokes and Adil Rashid crossed for the second run before Guptill released the throw? Because the run count on an overthrow was not on the list of reviewable matters. The law book had the truth written in it, but the door to verifying that truth with video evidence was bolted shut. Across fifty years I have watched this gap between the law on the field and the explanation off it, and at Lord's in 2026 the gap decided a World Cup.

Within a week of the incident the Marylebone Cricket Club issued a clarification on Law 19.8, covering overthrows: runs completed by the batters plus the boundary allowance of four are scored, with the second run counted only if the batters had crossed at the moment of the throw. The decision therefore rested on a physical position held for less than a second — something no umpire is expected to see with the naked eye, and something the review list never covered. The MCC clarified the law; the result did not change. The law tells you what is true; the protocol decides whether you are allowed to reach the truth.

This gap is not new, and it is the least discussed question in the modern game. The Decision Review System began in July 2026, in a Test between Sri Lanka and India at Colombo. Ball-tracking became a regular element from 2026. In 2026 the ICC amended the LBW protocol, fixing the ball-tracking tolerance at half the width of the ball — and that is where the 'umpire's call' mystery took root in international cricket. On June 7, 2026, at the World Test Championship final, the ICC abolished the soft signal for catches, so that the on-field umpire's prior instinct no longer interferes with the third umpire's decision. Read those three dates together and one pattern appears: cricket is drifting away from judgement on the field and towards estimation on a screen, while the tolerance band inside that estimation has never been explained to the audience.

When I began my weekly replay-analysis channel in March 2026 from a 2.6m by 2.1m box room above my garage in Didsbury, the first thing I understood was that viewers do not want to see the decision; they want to see the number behind the decision. The first twelve videos averaged 380 views. By the end of the year the channel had 11,400 subscribers and 1.9 million total views. I never changed the format, because the format was right: one decision, one law, one verdict. That habit is why I still look for the law number before I look at the picture. And this is where the third umpire's room and the living-room box room meet: both stare at a screen, but only one of them knows the structure of the law.

Understand the architecture. A batting side gets 15 seconds to challenge the standing decision. In those 15 seconds, a decision is assembled between the captain at one end and the senior batter at the other, usually from incomplete information. Then it goes to the third umpire, who sees ball-tracking, UltraEdge, real-time Snicko and multiple frame-by-frame cameras. The verdict appears on the stadium's giant screen, and that screen is cricket's real courtroom — because that is where the crowd first reads 'WICKETS' or 'UMPIRE'S CALL', and that is where the booing, the cheering and the social-media army are born.

From June 2026, when matches moved behind closed doors, I began hearing this process differently. Across those 92 matches I logged 1,104 whistle events and 63 face-to-face verbal confrontations, and out of that came a habit: before writing a report, I list what I heard, then what I saw. Applying the method to cricket, I recorded 2,683 stump-microphone events from behind-closed-doors internationals between July 2026 and February 2026. Where a crowd's roar used to mask umpiring error, silence exposed it — the batter screaming 'I hit it' before the ball reached the keeper, reaching the third umpire's ear faster than an UltraEdge spike.

Angle one — what the screen does not say. Under the laws a cricket ball measures 22.4 to 22.9 centimetres in circumference, which is a diameter of roughly 71 to 73 millimetres. The LBW ball-tracking tolerance is half that width, about 36 millimetres. If the centre of the ball passes outside that band relative to the outer edge of the stump, the outcome is final; inside it, the on-field call survives. The television graphic never shows that band. It shows a thin line and a word. The viewer believes he is watching a measurement; in reality he is watching a decision rule. The graphic on screen is not a measurement but a political decision about tolerances — and the politics is never shown to the viewer. Frame rate, grass on the pitch, camera calibration, how much the seam has moved: all of it feeds the estimate, yet the screen presents an immaculate whole. My notebooks say ball-tracking is most reliable in obvious cases and weakest precisely at the edges — the ball passing over the line of the stumps — which is exactly where ninety per cent of cricket's arguments are born.

Angle two — how trustworthy is the sound. UltraEdge and real-time Snicko detect the noise of bat, pad and ball contact. But a camera captures a finite number of frames per second, and if bat and pad are struck within the same frame, the spike is singular; the two sources cannot be separated. That is why the ICC protocol states that, absent conclusive evidence, the on-field decision stands. In my view that clause is the most honest and the least popular provision in the game. Before 2026, the noise of forty thousand people fused false edge and pad thud into one; in domestic silence we heard for the first time that there had been two separate sounds in the same second. A large part of my log shows umpires leaning towards suspicion rather than proof, because proof creates argument while suspicion lets the match continue.

Angle three — whom does the process reward. Here is the sharpest contradiction. If a side challenges a decision in which ball-tracking shows 'umpire's call', it burns a review and the decision does not change. A correct challenge is punished. As a result captains become strategists, using reviews only where they are certain, rather than where justice demands. Removing the soft signal in June 2026 has made the arithmetic more complex still: the third umpire now starts from nothing and must decide from zero, and that decision falls back into the same tolerance band.

I have seen this match from three angles, and the truth was still in the replay — it was simply never read with the law book in hand. The crowd will shout that the umpire is blind; the real problem is elsewhere. In 2026 I passed the refereeing exam; in 2026 a torn hamstring ended my running; in August 2026 Piccadilly Radio in Manchester hired me as its referee analyst for £180 a broadcast. Since then I have written every explanation in two layers: what the player felt, and what the law actually said. That dual frame taught me that a picture on a screen is a witness to the truth, not its judge. Between 2026 and 2026, review controversies in world cricket increased rather than fell. Why, as technology grew, did controversy grow with it? Because controversy moved off the field and into the tolerance band. The question used to be 'what did the umpire see'; now it is 'what did the graphic forget'.

Which brings me to my least welcome opinion: the crisis is not umpiring skill, it is protocol design — and blaming the umpire is the cheapest way to bury it. A home crowd will boo a board's decision, but the trouble begins with the box-room effect that took hold after 2026, when clipped replay debate shifted to the second screen and every disputed call began to be staged like a civil trial — prosecution, defence and verdict, all three in a studio. The screen in front of us now claims more authority than the pitch, while knowing nothing of the 36-millimetre tolerance or of the frame rate that can merge two contacts into one. Data models are strong in the middle and weak at the edges, and every argument in sport lives at the edge. The analyst who treats a model's output as final truth is merely a spokesperson for a model's limits.

So the answer is not more cameras or more slow motion. I propose three things. Publish the tolerance on the big screen beside the graphic, so the viewer reads a number — 'tolerance ±36mm' — and knows the limits of the estimate. Broaden the review list so the third umpire can examine overthrow run counts, because Lord's 2026 proves that gap can decide a World Cup. And extend the 15-second window to 20, so a captain can gather information from the far end before deciding; a quarter of all reviews are currently wasted purely on the speed and solemnity of the process.

What does the game actually want — justice, or a picture of justice delivered inside two minutes? Those six runs at Lord's are still on the scoreboard, and nobody will erase them. But if a knockout match at the next World Cup is decided through the same gap, there will be one more reason to curse an umpire — while the real fault sits in the section of the protocol that remains unamended seven years after a legal clarification. On every disputed call I still say, 'Check the replay, the call stands.' Yet every time I watch a disputed replay, I ask: who built the replay, and who signed off on its limits?