AI Can Now Track the Soccer Players Your TV Camera Misses

When you watch a soccer match on TV, the camera follows the ball. That means roughly half the players are off-screen at any given moment. Analysts and betting companies use AI to figure out where everyone is and how much space each team controls — but if you can't see half the players, how do you fill in the blanks? This new paper asks a narrower, practical question: does adding player speed data actually make those guesses better?

The author tested four versions of the analysis, from ignoring speed entirely to pretending you know exactly how fast every player is moving, even the hidden ones. The surprising result: for filling in missing players, speed data barely helped at all. For mapping who controls which patch of grass, speed mattered modestly. For the final team-level verdict — who's winning the space battle — speed's contribution shrank to almost nothing. In other words, precision at the surface doesn't change the big conclusion.

There's one useful practical finding. The speed that matters most is what you can actually see on screen; guessing the speed of invisible players adds only 2–6% more accuracy. And the wider the camera angle, the more speed data helps — at 36 meters wide it removed about 7% of the error, but at 60 meters it removed 21%. So the advice to analysts is simple: on tight shots, focus on figuring out where hidden players are; on wide shots, focus on how fast they're moving.

The catch is scale. This is a small study — three matches in the main test and eleven short clips from a single game — and it's a research paper, not a shipping product. The author did release the code and data, which means other researchers (and eventually sports-tech companies) can build on it. Expect this to quietly improve broadcast graphics, coaching tools, and fantasy-sports stats before you ever notice it happening.