Minute 81 in Saint Petersburg: The Limits of Refereeing Judgment When the Evidence Falls Short

On July 10, 2026, at the Krestovsky Stadium in Saint Petersburg, the 81st minute of the World Cup semi-final between France and Belgium. France led 1-0 through Samuel Umtiti's header in the 51st minute, from a left-sided corner. Belgium were forced to push their lines forward. A cross from the right travelled into the French penalty area. Umtiti, the scorer himself, dropped back to mark Marouane Fellaini. His left hand came to rest on the upper back and right shoulder of Fellaini. Fellaini lost his balance and went down.

Referee Andrés Cunha of Uruguay stood roughly twelve metres from the point of contact, slightly to the left of the main broadcast camera axis. He saw the shoulders of both players. He did not see the hand clearly. No whistle. The VAR team in Moscow reviewed the incident in silence and did not recommend an on-field review. Four minutes later Belgium had a final chance. Seven minutes later the referee blew for full time. France went to the final and lifted the trophy three days later.

Afterwards I wrote a three-thousand-word analysis reconstructing the incident from fourteen camera angles. It drew around 1.2 million reads and was used as reference material by a group of European refereeing instructors. But what I remember most is not the conclusion. It is that after watching all fourteen angles, I still could not say with certainty how much force Umtiti's hand applied, over how many milliseconds, or whether Fellaini fell because of that force or because he had already lost his footing.

The whole problem of the refereeing profession sits there.

Law, intervention thresholds, and the cost of a fraction of a second

The 2026 World Cup was the first to use video assistant referees across every match. IFAB wrote VAR into the Laws of the Game after a trial period beginning in 2026, and in Russia the protocol was standardised: only four categories of incident are reviewable, namely goal or no goal, penalty or no penalty, direct red card, and mistaken identity. For an 81st-minute incident, only the first two could apply.

The intervention threshold was fixed as well: VAR recommends an on-field review only when it identifies a clear and obvious error, or a serious missed incident. That phrase sounds like a technical standard. In practice it is a qualitative threshold set by people, and it shifts with the competition, the round, the minute of the match, and which team the referee has already favoured.

Article 12 of the Laws defines direct free kick offences, including holding, pulling, pushing and impeding. Article 12 does not define how much force constitutes holding, how many seconds constitute pulling, or how many centimetres of displacement constitute pushing. Nor does it say that a hand on an opponent's back is an offence in every case. The Law speaks of a player using a hand or arm to impede an opponent unfairly. Umtiti's hand becomes an offence only if it is established that it impeded Fellaini's movement, rather than merely making contact with him.

Contact does not equal foul. In a sport where two players inside the penalty area routinely stand half a metre apart, distinguishing those two concepts is almost the entire job.

I spent most of my time on this incident answering a much narrower question than the one asked on social media. Fans asked: was it a penalty? I had to ask three others. First, if it was a penalty, where is the error. Second, does that error cross the clear and obvious threshold. Third, if it crosses the threshold, why did the VAR team not recommend a review?

Three questions, three layers of evidence, and all three can be answered independently. The interesting result is this: the first two layers lean toward Belgium, while the third leans toward Cunha.

Layer one: the geometry of positioning

In modern referee training, positioning is taught as applied geometry. Referees do not chase the ball. They run the diagonal, always keeping an angled view of the point of contact. That means every position has a visible zone and a blind zone, and the blind zone is not personal error but the mathematical consequence of having two eyes and one observation point.

In the 81st minute Cunha stood in the textbook position. He was outside the play, slightly left, far enough to hold Umtiti, Fellaini and the ball in one frame. But precisely because he stood left, Umtiti's left hand was on the far side of Umtiti's body relative to his line of sight. Cunha saw the shoulder. He saw the loss of balance. He did not see the force.

The far-side assistant had the better angle on the hand, but the protocol allocates penalty-area challenges to the referee's priority remit. Assistants intervene only when something is clear and obvious. A hand on a back inside the penalty area, at twenty metres, in a phase involving at least four moving players, is not clear and obvious.

This is the point television analysis usually skips. When a broadcaster reconstructs an incident, it picks the best angle. A referee has no choice of angles. He has one, fixed within an effective field of roughly sixty to seventy degrees, decided before the ball arrives.

Layer two: the mechanics of contact

The Russian broadcast was recorded at fifty frames per second. Each frame is twenty milliseconds. When I broke the incident into frames, the interval from Umtiti's hand touching Fellaini's back to Fellaini hitting the grass was under six frames, meaning under one hundred and twenty milliseconds.

Within that window I measured three things. First, Umtiti's hand moved laterally by a very small distance, less than the width of a hand. Second, Fellaini's right shoulder rotated before his feet left the ground. Third, his centre of mass was already outside his support base before contact.

The third point is decisive. On my reconstruction, Fellaini entered the challenge in an unbalanced posture, head forward and hips rotating. Umtiti's hand may have amplified that imbalance, or it may simply have touched a player already going down. Those two possibilities produce two entirely different legal conclusions. If the hand caused the fall, it is holding and a penalty. If the hand merely touched a player already falling, it is a legitimate challenge.

At fifty frames per second, the distance between those two possibilities fits inside a single frame. A referee does not have frames. He has a moment.

Layer three: the clear and obvious threshold

This is the least discussed and most decisive layer. The VAR protocol does not require the referee to be right. It requires that the on-field decision stands unless there is clear evidence to overturn it. The burden of proof lies with whoever wants to change the outcome, not with the referee.

Measured against evidentiary standards in adjudication, this sits considerably higher than a balance of probabilities. It approaches proof beyond reasonable doubt. And it was designed that way on purpose, because the drafters understood that every reversal of an on-field decision carries a cost: time, match rhythm, and most importantly the referee's authority in the eyes of twenty-two players.

In the 81st-minute incident the VAR team had fourteen angles. Of those, three showed the hand clearly. All three were angled from behind, and none showed both the hand and the point of contact on the grass simultaneously. A four-person VAR team, each watching a different monitor, can split evenly on an incident like that. And when the VAR team splits, the protocol defaults back to the on-field decision.

That is why no monitor was offered to Cunha. Not because nobody saw the hand. Because nobody saw enough.

A risk index: quantifying an invisible decision

Since 2026 I have kept a scoring table for every controversial decision, built on five variables and a hundred-point scale. It is not used to judge referees. It measures the risk level of the situation a referee is forced to handle.

The first variable is field position, weighted thirty. An incident inside the penalty area scores maximum. The second is match timing, weighted twenty. The 81st minute of a semi-final scores near maximum. The third is score state, weighted twenty. A one-goal margin in the 81st minute scores maximum. The fourth is the number of players within a five-metre radius, weighted fifteen; this incident had four. The fifth is available image quality, weighted fifteen; fourteen angles but only three showing the decisive detail scores roughly six.

The total risk score for the 81st-minute incident is about eighty-nine out of one hundred. In my scale, anything above eighty-five falls into what I call the zone of no perfect ruling. Not the zone of no ruling. The zone where any decision will leave a group of people convinced something was taken from them.

An incident in the twelfth minute, midfield, at 0-0, with two players and clear images, scores roughly fifteen. The same act, the same hand, at two different moments, produces risk levels differing by a factor of six. The Law does not change. The context does.

Two-pass verification: handling an incomplete file

The mispronounced name in 2026 taught me that credibility is built through correction. That year I mispronounced the name of Syria's number 9 three times in a World Cup qualifier, and the broadcaster had to run a correction caption at half-time. I spent the following month reviewing every Syria match tape and building a pronunciation list for more than two hundred Asian players.

Since then, every analysis I publish passes two verification rounds. Round one checks facts: which Law is cited, which minute, what score, who stood where. Round two checks identities and sources: player names, referee names, assistant numbers, and retrievable data sources.

But the 81st-minute analysis forced me to add a third round, and that is the most important one. Round three covers what cannot be verified. When a detail cannot be verified, I state in the article that it cannot be verified, rather than picking the version that suits my argument. I watched two hundred incidents from the 2026 World Cup looking for an error nobody saw. What I learned was not how to find more errors. It was how to sort errors that can be proven from errors that can only be suspected.

The edge ball and the same unquantifiable threshold

I follow professional table tennis for the Guangzhou market, and the problem there has an identical structure. A table tennis edge ball contacts the table edge for under a millisecond, on a surface roughly two centimetres thick. Since 2026 some international events have introduced video review for edge balls, hidden serves and illegal toss heights.

The result mirrors football. Video review does not eliminate the edge-ball argument. It shifts the argument from whether the ball touched the edge to which frame is the decisive frame. Neither question has an absolute answer, because both depend on selecting a point on a continuum.

This holds for the edge ball, for the hand on the back, and for the offside line. In every sport that uses technology to adjudicate, there is a point where the technology reaches its own resolution limit, and behind that point lies human judgement. Building a good sporting legal system is not about erasing that point. It is about deciding where to place it and stating clearly who is accountable for it.

Audiences demand certainty; the Law permits uncertainty

Here is the counter-intuitive angle. The VAR protocol was designed with a tolerance for uncertainty, and audiences read that tolerance as a compromise.

Fans approach an incident like the 81st minute assuming a single correct answer exists somewhere among fourteen angles. That assumption is reasonable in physics and unreasonable in law. The Law does not ask whether the hand made contact. It asks whether the hand impeded. And the second question cannot be answered by video, because video records motion, not causation.

The clear and obvious threshold exists not to shield referees from criticism but to protect the rhythm of the game. Anyone who genuinely wants maximum accuracy must accept an entirely different system: every penalty-area incident reviewed automatically, stoppage time doubled, and the on-field referee reduced to enforcing decisions made in a video room. No major competition has chosen that, including the wealthy ones.

The blind spot is this: most refereeing controversies are actually arguments about which tolerance to choose, not about whether the referee was right or wrong. Because nobody states that openly, every non-intervention is read as system failure, when the system is in fact running exactly as designed.

VAR does not reduce controversy; it relocates it

Before VAR, people argued about the referee. After VAR, they argue about the intervention threshold. The same volume of emotion, a new target. But there is one important structural difference: before VAR, a referee's authority was accepted even when his decision was deemed wrong. After VAR, that authority became conditional, dependent on whether the video room agrees.

In the 81st minute Cunha decided within roughly two hundred milliseconds, with one angle, under the pressure of a World Cup semi-final. The VAR team had four people, fourteen angles, and time. Both sides chose not to intervene. Only one of them is criticised in the media, and it is the one without a monitor.

I do not claim Cunha was right. I claim there is not enough data to assert he was wrong, and in a legal system those two statements differ in nature. That difference is precisely what social media argument erases.

Publishing VAR audio and the right to say: I don't know

The progressive direction I consider viable over the next three years has three steps, and all three are within the reach of federations without new technology.

Step one is publishing the audio exchanged between the referee and the VAR team after each match, as some competitions have begun trialling. The public does not need the conclusion. It needs to hear the process.

Step two is recording the degree of uncertainty in the match report. A field for incidents where the VAR team split evenly, published after the tournament. Publicly admitting that some incidents cannot be resolved will reduce controversy more than any intervention ever has.

Step three is extending semi-automated technology to penalty-area contact incidents, but with a different purpose: not to rule in place of the referee, but to produce a standard record of force and direction of contact as training reference data.

The Laws of football are like the whistle: small, but they decide everything. Stopping the ball is an art; stopping the words is a responsibility. If an adjudication system has no room for the sentence I don't know, it is not an adjudication system but a system for finding someone to blame.

And if you had to explain a decision in a fraction of a second, before fourteen camera angles, before seventy thousand spectators and a billion television viewers, where would you choose to stand?