Semi-Automated Offside and the Tolerance Threshold: When Football Law Is Measured in Millimetres

The silent whistle at 23:47 was a verdict. That night, in a World Cup group-stage match, the referee awarded a goal and then withdrew it thirty-seven seconds later. On the big screen beneath the stadium roof, a graphic was rendered: twelve dedicated cameras had scanned twenty-nine data points on each player's body, sampling fifty times per second, and concluded that the striker's toe had passed the last defender's knee by exactly three point two centimetres. Three point two centimetres. Smaller than the thickness of a yellow card. Smaller than the shoelace that same player was tying. And it was enough to erase a goal, to change a result, to leave tens of thousands in the stands silent as though a verdict had been read out with a ruler.

I stayed four hours after that match to rewatch the graphic, not to defend or attack the referee, but to answer a simpler question: what was the offside law written for, and how is technology reading it? Because between those two questions lies a gap, and that gap is widening with every season.

WHAT THE OFFSIDE LAW WAS BORN TO PREVENT

Before cameras and sensors, you have to go back to the source text. The offside law in the IFAB Laws of the Game was never written to reward or punish to the millimetre. It was written to stop one specific behaviour: a forward standing still in front of goal, waiting for the ball to drop, scoring without needing to run. That is why the law is always tied to three relative concepts: the position of the ball, the position of the second-last opponent, and the moment the ball is played. All three demand a judgement, not a measurement.

For more than a century, everyone who has ever held a linesman's flag has known one thing: they never precisely determine the moment the ball leaves the passer's foot. The human eye needs roughly one hundred milliseconds to identify an event and synchronise hand, flag and eye. In a counterattack where a forward covers three metres in two-thirds of a second, one hundred milliseconds means an error of nearly half a metre. Which means that throughout football history, offside was never determined precisely. It was only determined convincingly enough for the game to continue.

VAR arrived at the 2026 World Cup and did exactly one thing: it allowed the moment the ball left the foot to be re-determined. But it also did something rarely discussed. It turned a flexible judgement of the human eye into a rigid sequence of camera frames. There, the concept of a clear and obvious error began to expand, until it covered almost everything visible on a slowed-down frame.

HOW THE SEMI-AUTOMATED SYSTEM READS THE LAW

The semi-automated offside technology took the next step, at a level the human eye cannot follow by ordinary means. The architecture is almost anti-aesthetic in its neatness: twelve dedicated cameras mounted under the stadium roof, all aimed at one half of the pitch and calibrated in fixed positions. Each player is modelled with twenty-nine key data points, captured continuously thirty to fifty times per second. The ball carries an inertial sensor transmitting at five hundred times per second, enough to pinpoint the moment of contact down to a single frame.

When the system detects a player in a potentially offside position, it immediately alerts the VAR team. By then it has recorded the last touch and reconstructed the position of every data point on the attacker's and defender's body at that exact moment. The comparison is made at the valid point of the body permitted to score. Head, torso, leg are considered. The shoulder is considered. And because the law allows scoring with the shoulder, any body part capable of scoring a valid goal becomes a reference for position.

That is why recent disallowed goals often hinge on body parts that sound absurd in a team sport: a toe, a fingernail of the boot, a shirt shoulder, a knee. The system is not technically wrong. It is honestly reading what a legal text permits it to read. And that is precisely where the problem begins: the law is never wrong, only the reading of the law is wrong, but sometimes the correct reading pushes us toward a result that is meaningless as football.

Three point two centimetres gives the forward no real advantage. He still had to break away from the defender, still had to choose his position, still had to score with a header or a finish. The advantage the law tries to remove is the advantage of the one who waits for the ball, not the advantage of the one whose toe is three centimetres longer. When technology measures what humans never needed to measure, it does not merely improve refereeing. It rewrites the entire definition of a valid goal.

RAW DATA AND THREE ROUNDS OF VERIFICATION

Based on my experience watching matches from the press tribune and from historical video archives, I developed an unbreakable habit: before asserting anything about an offside decision, I must rewatch it at least three times, at three different speeds, and record the data before opening my mouth. Every time I skipped that step, I was wrong.

In 2026, while a journalism student in Busan, I watched a second-division Korean match from the press stand. I logged fourteen fouls, then noticed the referee repeatedly ignoring a defender's shirt-pulling inside the box, especially at two moments I recorded specifically. I stayed four hours, recounting the assistant referee's every footstep on slow-motion phone video, and found a pattern: whenever the striker ran diagonally from left to right, the assistant was always one beat late. That late beat, multiplied by the player's running speed, produced a predictable margin of error.

The two-thousand-word analysis I wrote that day, with hand-drawn data tables and a positional diagram, was shared by a local football site. But what I kept was not the sharing. It was the lesson that a predictable margin of error is more useful than a correct but baseless complaint. From then on, every piece I wrote started with raw data and a positional diagram instead of a general feeling.

That experience also taught me why semi-automated offside was welcomed with such relief. It removed exactly the margin of error I once spent four hours counting: the lag between eye, flag and hand. But at the same time, it removed something else I am not sure we should lose — deliberate tolerance, the thing that keeps a match a match rather than a measuring session.

THE AUTOMATION MECHANISM AND THE QUESTION OF ACCOUNTABILITY

There is a fundamental difference between the semi-automated system and what comes next. In the current version, the system only alerts, while the final decision still belongs to the referee in the VAR room, and the on-field referee still announces the outcome. That structure preserves a human link, meaning there is still someone to be questioned, held accountable, criticised.

But further development is heading elsewhere. Automated offside technology tested in some competitions no longer alerts humans; it decides on its own and delivers the outcome directly. When that becomes the standard, the question is no longer whether the system is accurate. The question is: when a decision is the output of an algorithm, who is responsible for it?

VAR does not fix referees' mistakes, it only exposes their fear. When offside becomes automated, VAR does not even expose that anymore. It simply issues a notification, and we all nod, because there is no one to blame. If there is no one to complain about, there is no one to explain. And when a game no longer needs explanation, it has crossed a line most viewers do not notice immediately.

What is striking is that most domestic leagues adopting this technology stress a very similar message: the system does not replace the referee, it only assists. Yet the operational reality tells a slightly different story. When error is reduced to near zero, humans have no room left for judgement, and therefore no reason to be trusted. People can always invoke the machine to explain a puzzling decision, and that very ability makes referees less interrogated, not because they are better, but because no one wants to interrogate a machine.

THE CONTRARIAN VIEW: ACCURACY IS NOT JUSTICE

This is where I part ways with most technology supporters. A decision can be correct to the millimetre and still unjust to the very meaning of the law. Accuracy is a property of measurement, not of fairness. People confuse the two because in football, refereeing error was once equated with injustice, so when error is eliminated, fairness is assumed to have arrived. Not so.

The purpose of the offside law is to protect a competitive structure, not to protect a number. When a forward is denied a goal because a toe crossed the line, the law was correctly applied but its purpose was not served. The competitive structure was not exploited. No one stood still in front of goal. No one gained an advantage from three centimetres of toe. What is punished here is a breach of the text, not a breach of the game.

On the pitch there are twenty-two players and one man who is not allowed to make a mistake. But when a system stands behind that man, we must ask who is really holding the whistle. A referee who holds the whistle but does whatever the machine says is not the decision-maker. And if he is not the decision-maker, blaming him in a controversy confuses the operator with the designer. The system is not wrong. The operator is wrong. But when there is no operator left, we must look elsewhere.

There is an overlooked blind spot: the companies supplying live data to bookmakers are among those who benefit most from the digitisation of sport. When every moment is digitised to the millimetre, an enormous volume of data is generated, and most of it does not serve referees. It serves markets the spectator in the stands never sees. This is the darkest consequence of digitising sport, and it rarely appears in press conferences unveiling new technology. I have spent several seasons reading the technical reports of these systems, and what struck me was not their accuracy but how much more vaguely the data-sharing clauses are written than the technical specifications.

THE LOCKDOWN SEASON AND A LESSON IN HUMILITY

In March 2026, when the entire league was suspended by the pandemic, I fell into a state of disorientation like many colleagues. Instead of writing pieces about empty, cold stadiums, I dived into the historical video archive. I tallied a large volume of penalties across top European and Korean leagues from 2026 to 2026 and found a pattern I had not expected: the miss rate in matches without spectators rose significantly, but almost only in stadiums with roofs. A bare statistic that explained nothing, yet it taught me that data only has value when placed in the right context.

I sent the long analysis to a veteran editor. He said I had found something many had overlooked. That pulled me out of the crisis through curiosity, not through a comforting cup of coffee. And it taught me something I now apply to every debate about technology: a figure accurate to the millimetre, if not placed in the context of the game, is merely a lonely number.

The humility of the analyst lies here. A decision can be legally correct yet wrong for the football culture in which it is delivered. Korean football, where I work, has a refereeing culture different from European football. The pressure from the stands, the way players react to a whistle, the way assistants are trained to decide in an instant — all differ. When a globally standardised system enters an environment with a different refereeing culture, small conflicts surface at points we do not anticipate, and we must read them through the eyes of both cultures, not a single lens.

From Vietnam to Korea, I have learned one thing about analysing refereeing decisions: the difficulty is not in knowing what the law says. The difficulty is in understanding why a person reading that law chose to act as they did, within a specific context a text can never fully describe.

WHAT COMES NEXT

The refereeing trend over the next few seasons is fairly predictable. Leagues will keep standardising automated offside technology, not because it makes football fairer, but because it makes decisions easier to accept, and that helps leagues reduce controversy pressure. Precisely for that reason, I believe lawmakers will be forced to do something long delayed: redefine the concept of advantage in the offside law, or introduce an official tolerance threshold written into the text.

A clearly stated tolerance threshold will not be perfect, but it is more honest than pretending every millimetre measurement carries meaning. If the law acknowledges that a small deviation creates no real advantage, it returns to its original purpose. That is not a retreat before technology. It is the maturity of a legal system confident enough to admit it knows its own limits.

It took me three months to believe I was right, and two years to understand that being right is never enough. With semi-automated offside, what has become clear is that we stand at a point where the question is no longer whether a decision is right or wrong. The question is whether we want a match that is measured, or a match that is played. And that question, no camera can answer for us. I may have missed a detail somewhere in the data chain, and if so, I am ready to watch it a fourth time.