Broadcast football runs at fifty frames a second. One picture every twenty milliseconds is quicker than any eye can resolve, and for almost everything that happens on a pitch it is more resolution than anyone needs.
Then you ask it to judge offside, and twenty milliseconds stops being a duration and becomes a distance. So why did the 2026 World Cup need a match ball that reports its own motion five hundred times a second?
The short answer: the ball is a clock, not a ruler. World Cup offside technology already knew where every player stood. What it could not pin down, to better than about a hundredth of a second, was the instant the pass left the foot.
Video runs out before the law does
The offside law is written around an instant, and video only supplies intervals. Law 11 asks where the attacker was at the moment the ball is played or touched by a teammate. That is a single point in time with no tolerance attached to it. A camera cannot hand you a moment; it hands you the nearest frame it happened to expose.
That gap has been the quiet problem underneath every virtual offside line since the technology appeared. FIFA’s own offside documentation is candid about the error sources inherited from broadcast-camera systems: field topography, camera-angle distortion, player occlusion, body-part tracking. Every one of those is an error in where. None of them is the error in when, and the error in when is the one that scales with how fast the players are running.
What twenty milliseconds costs
Put a speed on it and the frame rate turns into centimeters. We ran the arithmetic from the published sampling rates rather than borrowing anyone’s summary of it.
A frame gap at sprint speed
Take an attacker breaking at 9 meters per second, roughly 32 km/h, near the top of what tournament tracking records for a sustained sprint. Now take the second-last defender stepping up at 4 m/s in the opposite direction, which is what defensive lines do at exactly the moment the pass is played. The two are closing on each other at 13 m/s.
Table 1 — player displacement inside one sampling interval, calculated from published sampling rates (FIFA and adidas figures, July 2026)
| Sampling source | Interval | Attacker alone | Attacker and defender closing |
|---|---|---|---|
| Optical tracking, 50 Hz | 20 ms | 18 cm | 26 cm |
| Ball sensor, 500 Hz | 2 ms | 1.8 cm | 2.6 cm |
Twenty-six centimeters. The offside calls that fill three days of argument are toe-and-shoulder calls, a few centimeters either way, so the timing term on its own was close to an order of magnitude larger than the decision it was supposed to settle.
That is a worst case, and worth saying so plainly. Most offside geometry is not a head-on closing at full speed, and a real system does better than the raw frame gap. But the worst case is the one that ends up on television.
Why interpolation doesn’t fix it
The obvious objection is that nobody snaps to the nearest frame. Tracking platforms interpolate player positions between adjacent samples, and over twenty milliseconds that interpolation is very good, because no athlete can change velocity much in a fiftieth of a second. Straight-line motion is a safe assumption over a window that short.
Interpolation answers where the player was between two frames. It cannot tell you which moment between them you should be interpolating to.
If the touch instant is derived from the same 50 Hz video, you can be ten milliseconds wrong about the target time, and then you interpolate very precisely to the wrong one. The output looks clean. It is confidently placed at a moment that is not the moment the pass was played. An inertial sensor sampling at 500 Hz drops that timestamp error to around a millisecond, and it does it with a physical signal — the acceleration spike and the change in spin when a boot meets the ball — that owes nothing to the camera clock. That independence is the point of the whole exercise.
What offside technology changed in 2026
The 2026 system kept the 500 Hz sensor and rebuilt the workflow around it. The ball technology itself is not new: adidas put a 500 Hz inertial measurement unit into Al Rihla for Qatar 2022, developed with FIFA and the sensor firm KINEXON, whose Global Sports Lead Maximilian Schmidt described the goal at the time as improving the experience “without changing the game of football.” The Trionda carries the same class of sensor inside the four-panel shell we followed from waterlogged 1930 leather in our history of the World Cup match ball. That piece treated the ball as an object in a wind tunnel. Since then it has become an instrument, and the instrument is what moved.
The alert route changed
In 2022 the semi-automated result went to the video assistant referee, who then talked to the pitch. In 2026 a clear positional offside goes straight to the officials on the field, so that, as FIFA’s Director of Innovation Johannes Holzmüller put it, “the assistant referees can flag for positional offsides” without waiting for a review. Holzmüller was equally direct about the boundary: the system handles positional offside and does not determine interference with play. Geometry travels down the fast path. Judgment does not.
That distinction is the honest version of “semi-automated”, and it survived the tournament intact.
The avatars are a modeling change
FIFA and Lenovo scanned all 1,248 participating players before the tournament and fed the resulting avatars into the offside system itself, not only into the broadcast replay. A skeleton gives the system sparse joint positions. A scanned body tells it how far the boot actually extends past the ankle joint, and where the shoulder ends and the arm begins — which matters, because arms do not count under Law 11.
FIFA also reported sixteen optical tracking cameras in each of the sixteen stadiums, producing over 150 million tracking data points per match. It has not published the rate those cameras sample at. The 2022 system tracked up to 29 body points per player, fifty times a second, so we did the only check available:
29 points × 22 players = 638 points per sample 638 × 50 samples per second = 31,900 points per second 31,900 × 5,400 s (90 minutes) = 172,260,000 points per match
172 million against FIFA’s “over 150 million”. Run it at 100 Hz and you get 345 million, which is a peculiar way to describe a number as “over 150 million”. So the 2026 cameras are probably still sampling near 50 Hz, and the sixteen of them are buying coverage and occlusion resistance rather than speed. We cannot confirm that from any public document, and we would drop the inference the moment FIFA published the real figure.
The number FIFA has never published
Here is what falls out of the arithmetic once the timing term is fixed. Drive timestamp error down to a millisecond and its contribution to the offside line drops to roughly a centimeter. Whatever is left is calibration and pose estimation — the error in where, which the ball never touched.
Nobody publishes that number.
Not FIFA, not its technology partners. There is no stated uncertainty on an offside line, no confidence interval, no per-stadium calibration residual. The 3D animation on the stadium screen is a rendering of an estimate, and it is drawn with the same crisp edge whether the margin was forty centimeters or four.
So we will say the thing the coverage has mostly stepped around. The 500 Hz figure is the least interesting true statement anyone made about this system, and quoting it as though it were an accuracy specification has let a sampling rate stand in for a measurement uncertainty nobody has been asked to disclose. A sensor’s sampling rate tells you how often it speaks. It tells you nothing about whether it is right.
We are leaving the heat science out of this piece on purpose: the hydration breaks, the wet bulb globe temperature studies, the host-city risk assessments. That is a different measurement problem with a different literature behind it, and folding it into a paragraph here would do it less justice than giving it an article of its own.
What we would want to see next
One number, published per stadium: the positional uncertainty of the tracked body points after calibration. Then put it on the animation. When the margin sits inside the uncertainty, say so on screen and let the referee’s call stand rather than dressing a coin flip in geometry.
None of that needs new hardware. It needs a willingness to show an error bar in front of eighty thousand people, which is a harder thing to build than a sensor.
The instinct worth keeping from all of this is small and portable. When a system shows you a number, ask what the number measures, then ask what it is quietly standing in for. We wrote this for you at FreeAstroScience.com, where we would rather hand you the arithmetic than the conclusion. Run it yourself. Disagree with us if it comes out differently.
Gerd Dani
President, FreeAstroScience
References and sources
- FIFA (2026). Faster offside decisions, more stable referee body cams and more analysis opportunities for teams. FIFA Innovation. Published 3 June 2026. FIFA innovation media round table
- FIFA (2026). New innovations developed with Technology Partner Lenovo shine at the FIFA World Cup 2026. FIFA Innovation. Published 30 June 2026. FIFA report on player avatars and Referee View
- adidas (2022). adidas reveals the first FIFA World Cup official match ball featuring connected ball technology. adidas News, Herzogenaurach. Published 1 July 2022. adidas release on the 500 Hz connected ball
- The IFAB. Law 11 — Offside. Laws of the Game. Accessed 27 July 2026. IFAB Law 11, offside
- Mallick, S. (2026). World Cup 2026 Offside Technology: AI, Computer Vision, and the Connected Ball. LearnOpenCV, Big Vision LLC. Published 2 July 2026. LearnOpenCV analysis of the offside pipeline
