Semi-Automated Offside Explained: Why VAR Calls Are 30 Seconds Faster

A striker peels off the last defender, latches onto a through-ball, slides it home. The bench celebrates. Then a graphic appears on the broadcast: coloured lines bisecting an armpit, a trailing knee, a frozen frame. Goal disallowed. Crowd deflates. Pundits argue.

That graphic is the visible tip of Semi-Automated Offside Technology — SAOT for short. In 2026 it's running in the Premier League, the Bundesliga, La Liga, Serie A, and is the offside engine of the FIFA World Cup. It cuts the average offside VAR check by roughly 30 seconds and removes most of the line-drawing subjectivity that used to make these calls feel like art.

Here's exactly how it works, why it's "semi-automated" rather than fully automated, and what it can and can't do.

What SAOT actually is

SAOT is a multi-camera, multi-sensor optical tracking system, not a magic algorithm. The Premier League's current build uses:

• Up to 30 dedicated tracking cameras mounted around the stadium roofline
• Several of them capturing at 100 frames per second (most TV cameras are 50–60fps)
• A computer-vision pipeline that tracks up to 10,000 surface "mesh" points per player in real time
• A separate sensor inside the match ball that detects the moment of contact (kick, header, deflection) and stamps it with sub-millisecond accuracy

Combine these and you get something no human assistant referee can ever match: the exact frame the ball was kicked, plus the exact 3D position of every player's relevant body part at that frame.

What it does, step by step

Here's the full chain when a forward runs onto a through-ball:

1. Kick detection. The ball sensor flags the moment the passer's foot connects.
2. Camera sync. SAOT pulls the frame from every tracking camera that matches that exact timestamp.
3. Mesh reconstruction. The system rebuilds each relevant player as a 3D skeleton/mesh in space.
4. Defender + attacker selection. It auto-identifies the second-rear-most defender (which sets the offside line) and the attacker being assessed.
5. Line drawing. Offside lines are drawn from the relevant body parts of both players.
6. Verdict. Onside or offside, with sub-centimetre precision.
7. Audio alert. For positional offside, an automated alert goes directly to the assistant referee — the flag goes up in seconds.
8. VAR confirmation. A human VAR validates the result before it's communicated to the on-field referee.

Why it's "semi" — what humans still decide

SAOT removes subjectivity from position. It does not remove subjectivity from two things that are still very much human judgement calls:

1. Whether the attacker is actively involved in play.
Law 11 says a player in an offside position is only penalised if they're involved in active play — interfering with an opponent, interfering with play, or gaining an advantage. A player standing offside but uninvolved in the move is fine. SAOT can locate them; it can't decide whether they were involved.

2. Whether the ball was deliberately played by a defender.
A defender's deliberate clearance "resets" the offside trigger — the attacker isn't offside if the defender played it to them. A deflection doesn't reset. The judgement of "deliberate vs deflection" is still made by VAR + on-field officials, not the system.

These two carve-outs are why "semi-automated" is the honest label. The system handles where; humans still handle whether it matters.

What changed for 2026

Three concrete upgrades made it into the FIFA 2026 build (and most of them are now live in the Premier League as well):

• Skeletal mesh tracking replaces the older limb-tracking model — finer-grained, less prone to occlusion errors when players overlap.
• Direct-to-AR audio alerts for positional offsides — the assistant referee gets told before the VAR booth completes its review, so flags go up faster.
• Sub-20-second average review time for clear-cut offsides, down from the 50+ seconds typical in 2022/23.

Marginal calls — the "armpit offside" graphics you've seen — still take 30–60 seconds because a human still has to confirm the involvement and contact judgements.

What SAOT can't fix

It's worth being honest about the system's limitations:

• Toenail offsides still happen. SAOT doesn't change Law 11; it just measures it more precisely. If the law says any body part that can score is decisive, then a toe sticking 2 cm past the defender's heel is still offside, and the graphic will show it.
• Occlusion at corner kicks and crowded boxes remains hard. When 14 players overlap, even a 30-camera setup can have low-confidence frames.
• Camera calibration drift is a real failure mode. Each match starts with a calibration step; if a camera nudges, the line can shift.
• It's still a recommendation. Every SAOT verdict is reviewed by a human VAR. The system is faster and more consistent than humans, but it's a tool, not a final authority.

The OURVAR.AI angle

We don't run a 30-camera stadium rig. We run frame-by-frame reasoning over the footage that already exists — broadcast feeds, fan clips, archive footage — using the same IFAB Law 11 framework SAOT operates under.

What that means for you: when you can't see the SAOT graphic (lower-league matches, archive incidents, social media clips), you can still get a structured offside verdict. The AI identifies the kick frame, identifies the relevant defender and attacker, cites the specific Law 11 clause, and tells you its confidence. For incidents where SAOT graphics do exist, you can cross-check.

Try a clip from this season's most-debated offside call and see how the AI reasons about it — first 20 credits are free, no card required.

OURVAR.AI is an independent AI Video Assistant Referee. Frame-by-frame reasoning, confidence scoring, and IFAB Laws cited on every decision.