Lionel Messi of Argentina shoots the ball past Rayan Aït-Nouri (Algeria) and goalkeeper Luca Zidane (Algeria) to score, in front of a packed crowd in blue and white stripes. — Lionel Messi's 4th-minute goal against Algeria in a 2026 World Cup group stage match was disallowed due to a close offside call © Tom Weller/dpa

Football by numbers: data and tech at the 2026 World Cup

June 2026

AI and data analytics are among the technologies changing the face of football. Leonie Mercedes reports on the new developments we’ll see in the 2026 FIFA World Cup and what it means for the future of the sport.

Did you know?

Each match in the Premier League generates 3.2 million data points – with plenty of rich insights to be extracted
Every one of the 1,248 players participating in this year’s FIFA Men’s World Cup will be scanned to generate a 3D avatar, which will be integrated into semi-automated offside technology to make it more precise
FIFA and Lenovo have developed an AI chatbot, based on a ‘football language model’ developed with Arsene Wenger

It’s a feeling football fans know all too well. The euphoria of a goal quickly draining away as the referee signals for a video assistant referee (VAR) review. Screens around the stadium read: “Checking goal. Possible offside.”

What follows could be three, four, even five minutes of anguish while the VAR team reviews the match footage, drawing the offside line on the frame at the kick point (where the ball was kicked – see box out), before making a decision. All the fans can do, and indeed the players, is wait.
In just a few years, VAR has changed the game. Now, further advances in technology and data are poised to continue this evolution – not just in officiation, but also in game strategy and talent scouting.

As the biggest ever FIFA World Cup kicks off in the US, Mexico and Canada, what innovations can fans expect to see play out, and how will they affect the future of football?

Video assistant referee (VAR) team monitoring multiple screens during a football match review in a control room. — Fans can face an agonising wait during potential offside calls, as video assistant referee officials can take several minutes to review match footage © Shutterstock

The offside rule

The offside rule states that if the attacker is ahead of the last defender (apart from the goalkeeper) at the time the ball is played, they’re offside. If a goal is scored by the attacker while they’re in an offside position, the ball will be disallowed.

The referee needs two pieces of information to determine whether a goal is offside or not:

Kick point: the point at which the ball was kicked
Offside line: an imaginary line that runs parallel to the goal line.

According to FIFA, the offside line is drawn at the last defending player(excluding the goalkeeper)’s “closest relevant body part”, that is, any body part that can be legitimately used to score a goal.

Offside decisions can be the most contentious of the game, and with the introduction of technologies including SAOT, the rulings have become increasingly marginal, sometimes to the centimetre.

Arsene Wenger has proposed a “daylight” rule, where if any part of the attacker’s body is on the same line as the defender, they’re not offside. This approach will, in theory, avoid those “millimetre” rulings that have become so contentious.

That said, as BBC Sports reported, there will always be a point where a player goes from being onside to offside. The systems are there to pick out that line.

Contentious offside rulings may influence game outcomes, though not always in the way you might think. In a Carabao Cup semi-final between Newcastle and Manchester City in January, the latter’s then manager Pep Guardiola suggested that an offside decision that hadn’t gone the team’s way had actually led City to victory, as it had made his players “angry”.

Assistant referee holding a fluorescent offside flag during a football match, with blurred stadium crowd behind. — © Shutterstock

Fair play

VAR was introduced in the Premier League’s 2019–20 season to avoid errors that could affect match outcomes. However, fans have complained that long reviews take the momentum out of the game and dull celebrations, as every goal comes with a temporary asterisk.

To keep things moving, football leagues including the Premier League have introduced self-automated offside technology (SAOT), which uses AI to automatically find the kick point and offside line for the VAR. The VAR will check those two points, as well as the positions of the players, before feeding back to the referee, who makes the final call.

As the “semi-” in semi-automated suggests, the technology doesn’t take the decision entirely out of match officials’ hands. The final call will always be made by the referee. But SAOT is hoped to make things more transparent to fans, as they’ll get to see an 3D animation of the play generated from the same data the VAR team used to come to their decision. The technology, provided by Hawk-Eye Innovations, relies on cameras around the pitch tracking the ball and players’ positions and movement on the pitch.

For this year’s World Cup, SAOT has been enhanced in a few ways. First, each of the 1,248 participating players will have their own 3D avatar – scanned to order by technology company Lenovo – rendered into the Hawk-Eye tracking systems. (Before, SAOT tracked 29 points on the players’ bodies to follow their movement in space.)

Filming setup with cameras and tracking markers capturing a person in a calibration booth for AI motion capture testing. — Lenovo’s 3D scanning technology will create an avatar for every player participating in the tournament. The avatars will be rendered into the Hawk-Eye tracking systems to improve its precision © Lenovo

A spokesperson for Lenovo told Ingenia that the scans are a more precise representation of the individual players, and that they’ll give match officials “a fuller picture from which to make their decisions”, but that “ultimately, the decision remains with them”.

FIFA’s director of innovation Johannes Holzmüller has said that the innovation helps with officiating, but also makes things more exciting for the fans. In the broadcasted 3D replays, he says, “the players really look like the players and it’s immediately obvious which players are involved in the offside position”.

The tournament ball will also be rigged with its own tech. Adidas’ Trionda ball contains an inertial measurement unit (IMU) that records its position in space with even greater resolution than the cameras – 500 times a second. Knowing the ball’s position at every instant will help in tight offside decisions.

Adidas Trionda football with removable panel, showing internal electronic mechanism against a bright studio background. — A specially created panel in Adidas’ Trionda football contains an inertial measurement unit motion sensor chip, which sends ball data to the VAR system in real time. It will even help video match officials identify individual touches to the ball, such as possible handballs © Adidas

The second enhancement in SAOT could speed offside decisions up further. Assistant referees, (also known as linesmen), will receive real-time audio alerts if a player is found to be more than 10 centimetres offside. In an earlier version of the technology, this audio alert would only play if the player was more than 50 centimetres offside. The assistant referee can then choose whether to raise their flag. The idea is that the tech will enable assistant referees to raise their flags sooner in offside scenarios instead of than waiting for a move to play out, which could prevent avoidable player injury.

Twenty-two players, a ref and a ball, each tracing out unique trajectories on 8,510 square yards of playing surface over the course of 90 minutes. That’s a lot of data. According to Genius Sports, which make SAOT tech for the Premier League, each match in that league generates about 3.2 million data points. How are clubs using it, and can it give teams an edge at the World Cup?

Translating movement into data

Action on the pitch can be translated into 1s and 0s after its capture in a few ways, including from those optical cameras around the pitch, or from broadcast-quality match footage. Clubs can use this data in scouting, to find the next Lionel Messi, for example, or to improve their team’s performance.

Game-Intuit, a spinout from Loughborough University, has developed a group of AI models that aim to help clubs identify and evaluate footballing talent.

While data is playing a greater role in football, Game-Intuit’s co-founder Aaron Gu notes that different clubs have different levels of “data maturity”, with some relying more on human judgement alone to make transfer decisions than others.

All 48 competing teams at the World Cup this year will have a chance to sample data-led match analysis. Football AI Pro, developed by FIFA and Lenovo, is a chatbot through which team coaches can access pre- and post-match analysis.

Sebastian Runge, FIFA’s head of football technology and data, told the New York Times that the tool applies language developed by that sage of soccer, Arsene Wenger (himself FIFA’s chief of football development).

Close up of Arsene Wenger wearing a suit and glasses, speaking at an event, in front of a blue background. — Arsene Wenger, FIFA’s Chief of Global Football Development, has been developing a unified football ‘language’ to keep the organisation’s Football AI Pro chatbot consistent © Shutterstock

Language is a key element for keeping these tools’ outputs comprehensible. You want to be sure that when you’re querying the chatbot about “line breaks”, you’re on the same page as to what that means. As clubs may have different terms for different actions or events, FIFA has been working with Wenger, technical experts, performance analysts and data scientists, to ensure the language is unified and help with data analysis.

Coaches can ask the chatbot queries such as “What was the pass completion rate for Harry Kane?”, or “Which player most often makes line breaks through the midfield line?” (A line break happens when an attacking player moves the ball past the defending players’ “line” of defence. It is one of the most important events in football, says Wenger, as it reduces the number of opponents between the ball and the goal.)

Meanwhile, Game-Intuit is building a platform for all clubs, including those that may not have the resources for data science teams, aiming to make it insightful as well as intuitive to use. Coaches can ask questions such as “Which striker in the transfer market fits best in my tactical system?”, and it will show the user a selection of players and its reasoning for those recommendations, Gu explains.

The platform takes in data from thousands of games from different leagues. Game-Intuit is developing a language for the model, so that it can “understand” terms like “creative”, for example.

Can AI’s decisions be explained?

From time to time, referees’ decisions can be baffling – that was clearly a foul! – and will remain incomprehensible. Such decisions can anger both fans and players. For AI to truly be accepted in football, how can we ensure the decisions it’s making can be explained, or understood by us?

Before considering the explainability of AI’s decisions, it’s helpful to look at how machine learning models are trained to “understand” states of play, such as whether a football player is offside.

“Models don’t really ‘understand’, says Christina Halmich, a data scientist from the Human Motion Analytics group at Salzburg Research, who studies athletes’ locomotion across different sports. “They learn the statistical associations between the inputs and outputs.”

To train an AI model, you would present thousands of different states of play that are labelled, for example, either onside or offside, she explains. “Then the model will learn the decision boundary.” In this example, the decision boundary would be the line, or more properly, a surface, that separates all the “onside” states from the “offside” states. “It is not always a straight line,” Halmich clarifies. “It could also be a curved or multidimensional surface.”

In simpler situations such as a single offside event, it is possible to use a model whose reasoning is easy to follow, such as a decision tree, which lets you trace the exact path taken to a decision. “Decision trees are often used if transparency really matters,” Halmich says, such as in medicine or law.

When the task becomes more complex, such as recognising different playing styles, we often need a more powerful model, like a deep neural network, Halmich continues. These models can can capture patterns that are too subtle or complicated for a simple decision tree, she adds. “The trade-off is that they are less transparent; instead of following one clear decision path, the model makes its prediction through many internal layers, which makes the reasoning harder to interpret.”

This is where explainable AI comes in. “Explainable AI helps us understand what the algorithm ‘thought’,” Halmich explains. “It’s a family of techniques that are used to understand the reasoning behind the model’s predictions.” One of these is known as SHAP, short for SHapley Additive exPlanations. It works by determining how much each feature contributed to a certain prediction, or output.

The more complex a task, the harder it becomes to explain, says Gu. “If you want to do a very simple task, asking something like, ‘find the best passer in the Premier League’, you might be able to do it with an explainable model,” he says. However, if you ask about something more complex such as “Would this Premier League player be the best passer in La Liga?”, which encompasses much more data, how the model came to its decision is more challenging to parse.

Gu’s team is working on making their outputs explainable, so that their clients can understand why their model made a certain decision. He emphasises, however, that Game-Intuit exists merely to help the decision-maker, not to replace them.

AI and football: the perfect match?

With AI playing a greater role in the beautiful game – not just in officiating and fan engagement, but also recruitment and training – it’s worth asking: what’s this all for?

Fundamentally, sport is a form of entertainment. While accurate interpretation of the rules, as enabled by technology, is important, as lengthy VAR decisions have shown, as much as possible it should not at the expense of entertainment.

Transparency and a sense of fairness will have a big impact on a fan’s enjoyment of the game. At time of publication, SAOT has already had at least one controversy because of a technical hitch in generating the offside animation during an incident in Switzerland and Qatar’s group stage match. FIFA said the VAR’s decision was not affected, but didn’t release the animation until hours later, leading fans and pundits to ask for greater transparency. Assuming no further issues going forward, spectators will hope the SAOT animations that accompany offside decisions will be made public immediately, to bring this transparency. Referee View, which at the World Cup will let spectators see what the referee is seeing via a chest-mounted camera, should help too.

Beyond the World Cup, enhanced data capabilities for leagues large and small may surface talent that might otherwise have been missed, providing further gains for the game and fans. Advances in tracking may bring fans as close to the game as it’s possible to go without actually being there.

Genius Sports’ executive vice president of sports and technology, Michael D’Auria, has suggested that 3D recreations of gameplay may one day be able to put fans into the position of the player, so they can feel what it’s like to be on the pitch, and perhaps better understand a decision they made in a crucial moment.

The technologies we’ve met here are about to be tested on the largest possible stage – the World Cup 2026 will be one of the biggest sports competitions ever seen – and how they are received will make yet clearer what is most important.

If anything, with an animation accompanying every close offside call, the most confusing rule in football will become much clearer.

Contributors

Dr Aaron (Chaoyi) Gu is CEO and Co-Founder of Game-Intuit, a football AI company building context-aware intelligence for clubs. He holds a PhD from Loughborough University, specialising in machine learning and sports analytics. His expertise spans football data modelling, player performance prediction, and data-driven decision support for clubs.

Christina Halmich is a Data Scientist at Salzburg Research, where she specialises in machine learning, human motion analytics, and data augmentation for biomechanical time series data. Her work focuses on developing data driven solutions to real world challenges, with the goal of improving health, performance, and quality of life. She holds an MSc in bioinformatics from Johannes Kepler University Linz and is currently pursuing a PhD in artificial intelligence at the Paris Lodron University of Salzburg.

Leonie Mercedes is a freelance writer based in London.