OnSport AI has raised $2 million for software it says flags athlete movement associated with injury risk. The reviewed sources do not establish whether its alerts prevent injuries.

The Miami startup announced the pre-seed financing on November 4. It intends to spend the money on product development, engineering and biomechanics staff, sports partnerships and military applications, according to the announcement.

The big change

  • The company describes live and retrospective alerts.
  • The company describes comparing an athlete with their own movement history, using tracking data or video.
  • A warning is not a diagnosis or proof of prevention.

A signal for a decision

Refresh Miami reports that OnSport co-founder and CEO Brett Burman said decision-makers need something actionable rather than raw numbers alone. The outlet reports that the system compares movement over time with a player's usual patterns, using league tracking data where available and video otherwise.

The intended question is whether a change warrants attention from people already managing the athlete. Its website also presents movement information as an input to return-to-play decisions alongside staff judgment. The sources do not disclose tested error rates or show that an intervention worked.

Refresh Miami reports that OnSport says it has worked with an NBA team on a pilot for more than a year. The team was unnamed and the outlet did not supply results of an independent evaluation. The company also reports agreements across basketball and professional soccer.

Why evaluation matters

The uncertainty is broader than one product. A 2022 systematic review examined 30 studies of sports musculoskeletal injury prediction. The authors found no external validation in those studies and reported that most models had high or unclear risk of bias. Testing a model on an independent population matters because an apparently useful signal in the data used to build it may behave differently elsewhere.

That review did not assess OnSport and cannot establish this product's accuracy. For a team considering an alert, useful unanswered questions include how often it misses injuries, how often it flags athletes who do not get hurt, and whether decisions based on it improve outcomes. The published descriptions do not answer those questions.