Offside looks like the perfect use case for automation. Freeze the frame, draw the line, announce the decision immediately. Goal-line technology does this with goals. So why is football still using humans in a video room?

The remaining delays cluster around five things, and most of them trace back to IFAB Law 11. The key line in Law 11 is not the offside line itself but the principle that being in an offside position is not an offence. A player is penalised only if they are in an offside position when the ball is played or touched by a teammate and then become involved in active play — defined through concepts such as interfering with play, interfering with an opponent and gaining an advantage. That is the border between automation and refereeing, and it shows up in each of the five delays below.

Validation. Every system asks VAR to confirm kick point, players, body part and outcome — the safety layer on top of the automation. This is why Pierluigi Collina, the chairman of the FIFA Referees Committee, has pushed back on the idea of “robot offside”. In FIFA’s explanation of semi-automated offside, Collina says the technology supports officials, but referees and assistant referees remain responsible for the decision.

Occlusion. Players block cameras and the ball disappears in crowded areas. FIFA’s offside standards page lists field camber, camera-angle distortion, player occlusion and body-part tracking as common sources of virtual line error.

Body-part definition. Law 11 does not count hands or arms, and IFAB defines the upper boundary of the arm as in line with the bottom of the armpit. The system is not asking for the furthest forward point on the player but the furthest legally relevant body part.

Active involvement. A player can be in an offside position without committing an offence, and can equally commit one without touching the ball — by obstructing the goalkeeper’s view, challenging an opponent or making an obvious action affecting an opponent. Tracking can show where players were; it can’t always decide what their presence did.

Deliberate play. Law 11 says an offside player is not penalised if the ball comes from an opponent who deliberately played it, unless it was a deliberate save. IFAB’s deliberate-play guidance asks whether the defender had control, distance, speed, clear view and whether the action was instinctive or controlled, none of which a camera can settle on its own.

The 2025/26 law change on goalkeeper throws makes the point sharply. IFAB says the first point of contact is normally used to judge offside position, but when the ball is thrown by the goalkeeper, the last point of contact is used instead. A fully automated system therefore cannot merely detect “ball contact”; it has to know what kind of contact happened, which means even the timestamp is physics interpreted through Law 11.

The technology is already building a live 3D model of the match

Semi-automated offside technology is much more advanced than a person dragging lines across a TV picture.

The system calibrates the stadium so cameras understand the pitch geometry, field markings, lens distortion and the relationship between the 2D image and real-world 3D coordinates. It tracks all 22 players across frames while they sprint, overlap, jump and disappear behind bodies, maintaining identity throughout. It estimates body position, usually through tracking limb points or, in newer systems like the Premier League’s Genius Sports rollout, through mesh tracking that reconstructs a dense surface model of each player. It tracks the ball, either with cameras alone or with a connected ball containing a sensor. And it identifies the kick point — the exact moment the ball is played by a teammate.

Once that timestamp is selected, the system calculates the position of the second-last opponent, the ball and the relevant attacking body part, generates the virtual offside line and alerts the VAR team, which still validates the kick point, the players, the body parts and any subjective Law 11 issue before the decision is given.

The timestamp matters more than the line

If the moment of the pass is marked wrong, the line can be perfectly drawn and still produce the wrong answer.

At 50 frames per second, a camera gives you one frame every 20 milliseconds; at 100 frames per second, one every 10. A connected ball transmitting at 500 Hz sends motion data every 2 milliseconds. That difference matters on marginal offsides, because if the attacker and defender are moving in opposite directions, one frame can change whether a shoulder, knee or toe appears level or offside.

A connected ball also helps when cameras are blocked. In a crowded penalty area, the ball can disappear behind a cluster of players, and a camera-only system has to infer the kick point from what it can see. A connected ball can register motion and impact even when the optical view is messy.

Two different bets: FIFA’s connected ball vs the Premier League’s cameras

At the 2022 World Cup, FIFA’s semi-automated offside system used 12 dedicated tracking cameras tracking the ball and up to 29 data points on every player, 50 times per second. The Adidas Al Rihla match ball also contained an inertial measurement unit sensor sending data to the video operation room 500 times per second. FIFA’s later Women’s World Cup FAQ said the technology cut average offside checks from around 70 seconds to around 25. UEFA has used a similar approach with connected balls in Champions League and Euros matches.

The Premier League went the other way. From 12 April 2025, the league’s system uses up to 30 newly installed stadium cameras, some capturing at 100 frames per second, tracking the ball and up to 10,000 surface mesh data points per player. Crucially, it does not use a chip in the ball — the league’s bet is that more cameras, higher frame rate and dense mesh tracking can replace a connected-ball timestamp.

That bet has a clear weakness. In January 2026, The Guardian reported that an English semi-automated offside delay in a Carabao Cup semi-final was blamed on an edge case caused by player congestion around the ball, where the system could not detect the ball position cleanly and officials fell back toward manual review.

Other leagues sit on a spectrum between these poles. LaLiga uses skeleton tracking and, according to Football España, wants chip-in-ball “automatic offside” from next season. The Bundesliga uses 11 roof-mounted cameras and tracks 21 offside-relevant body parts. Serie A adopted early in 2023, and Brazil’s CBF moves to mesh-based semi-automated offside in 2026. Semi-automated offside is now a family of systems rather than a single machine.

FIFA is already testing the next version

For the 2025 Club World Cup, FIFA said an advanced semi-automated offside system would provide automated real-time alerts for clear offsides, with challenging scenarios still validated by VAR. At the 2025 Intercontinental Cup, FIFA also tested “Real-time 3D Recreation” to give VAR virtual feeds replicating both goalkeepers’ views, directly relevant to line-of-sight and obstruction questions.

For the 2026 World Cup, FIFA’s TRIONDA ball carries a 500 Hz motion sensor chip mounted in a side panel of the ball, with counter-balances for flight stability. FIFA and Lenovo announced that players at the tournament would be digitally scanned to create AI-enabled 3D avatars capturing accurate body-part dimensions, helping track players through fast or obstructed movements.

Even with all that, FIFA keeps the same distinction: clear positional offsides can be accelerated, but challenging scenarios still need VAR validation.

What full automation would require

A fully automated offside system is technically imaginable. It would need a connected ball in every match, with a certified sensor package, common timing standards, low latency, charging and backup protocols, and clear procedures for sensor failure.

It would need high-frame-rate optical tracking in every stadium, reconstructing all 22 players, the ball and the pitch in 3D. The strongest version would combine FIFA-style connected-ball timing with Premier League-style dense player mesh tracking.

It would need player-specific modelling, and FIFA’s 2026 World Cup avatar project points in this direction. A generic skeleton is useful, but a player-specific model is better because players have different body dimensions, running shapes and limb proportions. If the decision is about a shoulder, knee, toe or head, the model should represent the actual player as closely as possible.

It would need sensor fusion — optical tracking, ball sensor data, pitch calibration, player identity and the rule engine combined into one synchronized model. On marginal offsides, a few milliseconds matter, and if the kick timestamp and body-position timestamp are not aligned, the output can look precise without being trustworthy.

It would need a deterministic rule engine, which is the requirement people most often skip. Machines can make fast, trusted decisions when the law is written like code. Goal-line technology works because the rule is binary: did the whole ball cross the whole line? Tennis line calling works because the rule is binary: did the ball touch the line? Current Law 11 is not binary in that way.

And it would need confidence thresholds. If the ball sensor fails, the cameras lose a player, two players touch the ball almost at the same time, the model loses track of identity or the body-part confidence drops below a threshold, the system needs a fallback — VAR review, the on-field decision, or manual crosshairs. Even “fully automated” needs a failure protocol.

The technology stack is not science fiction. Changing the law is the harder problem.

The most automatable law would be position-and-next-touch

If football wanted offside to be almost fully automated, the cleanest law would be a position-and-next-touch rule: a player would be offside if they were beyond the second-last opponent or the ball at the moment a teammate played it, using clearly defined scoring body parts, and then became the next teammate to touch the ball.

That would be easy for a rule engine. The connected ball timestamps the pass, the tracking system freezes the player positions, the software identifies the second-last opponent and the attacking player’s relevant body part, and the system waits for the next attacking touch. If the next teammate to touch the ball had been in an offside position at the pass, the decision is generated. Clean through-balls would become close to instant, and many crosses, cutbacks and direct receptions would also be much faster.

But it would be a different offside law. An attacker could stand in an offside position in front of a goalkeeper, never touch the ball, and still affect the goalkeeper’s view, and current Law 11 can punish that. A touch-only automated law may not. The technology becomes cleaner, the law becomes cruder.

A feet-only law would make the technology easier

Another possible simplification would be feet-only offside, attractive technologically because feet are easier to define than shoulders, knees, torsos, armpit boundaries and leaning upper bodies. If only feet counted, the game could combine connected-ball timing with optical tracking and possibly approved lower-limb wearables.

Football already has a framework for this kind of technology. FIFA’s Electronic Performance and Tracking Systems programme covers optical and wearable devices, and FIFA says lower-limb wearable systems can be approved under its quality programme after additional validation. A future feet-only system could use a connected ball, boot or shin-pad sensors, optical tracking, stadium positioning anchors and a synchronized rule engine.

It would also change offside. Current Law 11 counts the head, body and feet, while excluding the hands and arms, so a striker leaning beyond the defender with a shoulder can currently be offside because the shoulder is a playable part of the body. A feet-only law would treat that differently — once again trading a simpler law for easier automation and ending up with a different sport.

Where tennis is useful, and where it misleads

Tennis shows what full automation looks like when the rule is narrow enough. A line call asks whether one ball touched one fixed line, and the ATP announced that Electronic Line Calling Live would be adopted across the tour from 2025, with Wimbledon following the same year. Football already has its own tennis-like technology in goal-line technology, which asks whether the whole ball crossed the whole line.

Offside only looks similar from far away. The offside line is not painted on the pitch but generated for a single instant from the position of the second-last opponent or the ball. The relevant object is not one tennis ball but the ball plus 22 moving players, and the decision depends on timing, playable body parts and active involvement.

Tennis also proves automation does not remove human-system risk. Reuters reported a line-calling error during Wimbledon 2025 that was attributed to operator error, and the French Open kept human line judges for 2026 while the other Slams used electronic systems. Automation works best when the rule is simple, the line fixed and the output binary — and even then the system still needs operators and fallbacks.

Bottom line

VAR cannot simply be fully automated because VAR is not one technology and offside is not one measurement.

The measurable part is already being automated, and a more automated future is technically plausible, combining connected balls, high-frame-rate optical tracking, player mesh models, AI-enabled avatars, sensor fusion, confidence thresholds and a deterministic rule engine. But that only works cleanly if Law 11 is simplified. Under the current law, offside still includes active involvement, interference with an opponent, deliberate play, saves, rebounds and phase of play, and those are football judgments rather than measurements.

So the answer is not that football cannot automate offside.

It can automate offside position.

It cannot fully automate the offside offence unless it changes what offside means.