Beyond the Algorithm: When Accuracy is Not Enough

Earlier this year, in The Algorithmic World Cup, we looked at the growing role of AI and advanced technology across football, from officiating and performance analysis to broadcasting, fan engagement and tournament management. The 2026 World Cup showed how far that process has already gone. Sensors, cameras and increasingly sophisticated data systems gave officials access to information that would have been impossible to obtain only a few years ago.

But greater precision creates a different problem. Technology can establish facts with extraordinary accuracy. It cannot, by itself, decide what those facts should mean. As sport becomes capable of detecting events that no player, official or spectator could reasonably have perceived, the question is no longer simply whether the technology works but it is what sport chooses to do with what the technology reveals.

When technology provides new evidence

One of the most striking examples came when Croatia had a late equalizer against Portugal disallowed after the connected ball detected a touch that was barely perceptible to the human eye. Technically, this was impressive. The technology did not fail. Quite the opposite: it appears to have done exactly what it was designed to do. It detected information that human perception alone would almost certainly have missed.

But that does not mean the technology made the decision. The connected ball established the moment of contact. The Laws of the Game gave that contact significance. The officiating system then translated those two things into a decision: the goal did not stand. This distinction is fundamental. Many technology-assisted sporting decisions contain two different questions: 'What happened?' and 'What should the rules make of it?'. Technology is becoming exceptionally good at answering the first question. It can tell us that a ball touched a player, that a foot was several centimetres beyond a line or that contact occurred at a particular moment. But measurement alone cannot determine what kind of sport we want those facts to produce. That remains a human choice, embedded in rules, thresholds and interpretations.

The Croatia incident therefore raises a broader question.

Do rules developed around what humans could reasonably perceive remain appropriate when technology can identify something that nobody involved in the game could reasonably have seen?

There is no obvious answer. Nor is the argument that imperceptible events should simply be ignored. The point is that greater technological accuracy does not remove the need for regulatory judgment. In some cases, it makes that judgment more important. For decades, the case for technology in officiating was relatively straightforward: humans make mistakes and technology can help correct them. But once technology begins to reveal an increasingly granular version of reality, sport must decide which parts of that reality should matter. That is a question for rules and governance, not engineering.

When the system around technology fails

There is another problem, and it is considerably more mundane. Technology is infrastructure. And infrastructure can fail. Tennis has provided a useful reminder.

When Wimbledon removed its line judges in 2025 and moved to electronic line calling, the change was presented primarily as an improvement in accuracy. During that year's Championships, however, the system was inadvertently deactivated for part of a Centre Court match between Anastasia Pavlyuchenkova and Sonay Kartal. Three calls were missed before the problem was identified.

Importantly, Wimbledon did not blame an algorithm. The All England Club said explicitly that the line-calling technology was not an AI system and that the incident resulted from human error. It subsequently changed the system so operators could no longer manually deactivate ball tracking.

A year later, during Wimbledon qualifying in June 2026, electronic line calling again became unavailable across the courts following a power problem. Play was suspended for around an hour.

Neither incident demonstrates that automated line calling is less reliable than human line judges. There is no good evidence for such a conclusion. They demonstrate something more useful. Automation changes the human and organizational system around a decision: someone installs the cameras; someone calibrates them; someone supplies power; someone operates the system; someone determines when its output can be relied upon; someone establishes the protocol for what happens when it stops working.

Removing a human from one part of the decision-making chain can therefore create a considerably more complicated chain behind them.

This is why technology should not be governed only as a technical product. It is one component of a wider operational and decision-making system. The reliability of that system depends not only on whether the technology works as designed, but on whether the organization has considered what happens when it does not.

When responsibility becomes distributed

This leads to a more difficult question: responsibility. It is becoming a serious subject of academic research, but it is also a question we have returned to repeatedly at OrdoStrategica. As modern sport becomes more technologically sophisticated and institutionally interconnected, accountability can become increasingly difficult to locate.

Writing in Frontiers in Sports and Active Living in July 2026, Jun Woo Kwon of Seoul National University described the emerging problem in automated and assisted sports officiating as one of “error without accountability.” The point is not that nobody is responsible. It is that responsibility becomes distributed.

As we have argued elsewhere in our Ordo papers, including in Governing Modern Sport as a Living System and The Hidden Fragility in Modern Sport, modern sports governance increasingly distributes responsibility across interconnected systems rather than single decision-makers.

Consider a disputed VAR decision.

The referee remains formally responsible for the final call. But the decision may also depend on the VAR official, replay operator, available camera angles, tracking system, technical provider, review protocol and the rules established by the governing body.

If the resulting decision is wrong, where exactly did the error occur? That question was easier to answer when a referee simply failed to see a foul.

This matters because accountability is part of how sport maintains legitimacy. Players and supporters do not expect referees to be perfect. They do, however, expect consequential decisions to be understandable.

An imperfect human decision is relatively easy to explain: the referee did not see the incident, interpreted it differently or simply made a mistake. A technologically mediated decision can be considerably harder to unpack.

Was the underlying information wrong? Was the technology incorrectly configured? Did an operator make an error? Was the evidence correct but interpreted incorrectly? Did the rules themselves produce an outcome that appears unreasonable? Was the system unavailable and, if so, was the contingency adequate? These are different types of failure, involving different actors and different responsibilities. That is why technological assurance cannot stop with asking whether the product itself is accurate. The decision-making system has to be governable as a whole.

Beyond officiating

The same problem becomes even more important when technology moves beyond visible decisions on the field. AI is increasingly being used in athlete monitoring, injury prediction, tactical analysis, scouting, rehabilitation and performance management. Here, its outputs may influence consequential decisions without players, supporters or even others within an organization necessarily knowing that a model was involved.

Consider injury prediction.

A model identifies an athlete as having an elevated risk of injury. Medical and coaching staff decide not to play her. She remains healthy. Was the prediction correct? Perhaps. But perhaps she would never have been injured.

Reverse the situation. The model assesses the risk as low. The athlete plays and suffers a serious injury. Was the model wrong? Was important data missing? Did the medical team give its recommendation too much weight? Or was the prediction statistically reasonable even though the individual outcome was bad?

These questions are increasingly relevant.

A perspective published in npj Digital Medicine in August 2026 examined the growing use of AI in sports medicine and questioned the clinical readiness of current applications in areas including injury prediction, recovery and clinical decision-making. Among the concerns identified were black-box predictions, bias and the need for stronger data governance. A separate sports-science review published in September identified recurring weaknesses including limited datasets, poor validation, insufficient reporting of uncertainty, unclear translation from prediction into action and inadequate attention to athlete rights and post-deployment monitoring.

The important distinction is simple - prediction is not decision.

A system might be reasonably good at estimating the probability of an event. That does not mean it knows what a coach, doctor, referee or federation should do about that probability. Numerical precision can make this distinction surprisingly easy to forget.

If a scout says that a seventeen-year-old player probably will not make it at elite level, everyone understands that this is a judgment. If a sophisticated model assigns the same player a 23 per cent probability of reaching a first team, the assessment suddenly appears different. It has a number attached to it. But the number has a history. It depends on the population used to train the model, the variables selected, the quality of the underlying data, the definition of success, the assumptions built into the system and the environment in which it is subsequently used. Precision should not be confused with certainty. Again, technology can inform a decision, but it cannot relieve the organization of responsibility for making one.

Building the system around the technology

There are already signs that sports organizations recognize that technological accuracy alone will not generate trust.

In August, UEFA published a standard four-step process for pitch-side VAR reviews, recommending a common sequence for analyzing incidents. A week later it launched Clear Line, a repository of more than 150 refereeing scenarios explaining how the Laws of the Game should be interpreted and when VAR should and should not intervene.

For the first time in the 2026/27 season, the Premier League is publishing the findings of its independent Key Match Incidents Panel weekly, providing greater transparency around the evaluation of refereeing and VAR decisions.

While these may appear to be relatively modest changes, we argue that they are truly significant because they address something technology itself cannot solve: legitimacy. A technically accurate output does not automatically produce a defensible decision. For that, sports organizations need to build an institutional system around the technology.

At a minimum, that system requires six things.

• Clear rules. Organizations need to establish what technological evidence can be used for, what significance it carries and when intervention is appropriate. Those rules also need to evolve as technology becomes capable of identifying things that were previously impossible to observe.
• Defined authority. It must remain clear who owns the final decision. Technology may inform, recommend or identify, but authority cannot disappear into the system.
• Assurance. Organizations need confidence not only in the technology itself but in the data, configuration, operating processes and people on which its performance depends.
• Review and challenge. Where decisions are consequential, there should be clarity about whether an output or resulting decision can be questioned, by whom and through what process.
• Contingency. The organization must know what happens when the technology is degraded or unavailable. Failure arrangements cannot be invented after the system goes down.
• Accountability and explanation. Someone must remain responsible for explaining the outcome, identifying what went wrong when errors occur and ensuring that lessons are incorporated back into the system.

None of these are principally technological requirements. They are governance requirements. And they matter precisely because the technology will usually work. The more accurate, sophisticated and embedded these systems become, the easier it is for organizations to treat their outputs as decisions rather than inputs into decisions. That is the mistake sport needs to avoid.

For most of sporting history, human error was an accepted part of the game. A referee missed a foul. A scout misjudged a player. A doctor underestimated a risk. The decisions could be frustrating and sometimes consequential, but their origin was usually visible. Technology allows many of these decisions to be better informed. It can see things humans cannot see, process information at a scale humans cannot process and identify patterns humans might never detect.

That is precisely why its governance matters.

The objective should not be to create technological systems that never fail. That is an unrealistic standard for humans and machines alike. Nor should sport assume that a system which produces an accurate measurement has therefore produced the right decision. Technology should not be governed as a product in isolation. It should be governed as part of the decision system in which it operates.

That system must determine what technological evidence means, who has authority to act on it, how its reliability is assured, how decisions can be challenged, what happens when the technology is unavailable and who ultimately remains accountable for the outcome. The more capable technology becomes, the less useful it is to think of it as a substitute for human judgment. Its real value lies within a system that determines what its outputs mean, who has authority to act on them and who remains accountable for the result.

Ultimately, technology may change the evidence on which sport decides. It does not remove the obligation to govern the decision.

‍