MyTwin builds a first injury-risk prototype for tennis, padel and pickleball

The prototype works on the player’s data in their MyTwin twin, not on a one-off test: training load, recovery, physiological indicators and history. It returns an estimated risk for the next 14 days, meant to be read by the player and their staff alongside everything else they know.

Joshua’s story shows the doubt it is meant to address. An amateur competitive tennis player, drained before an important tournament, he had no clear way to decide whether to push through or ease off.

An estimated risk is not a forecast of what will happen. Our article on what a risk score really means explains the difference. The prototype is a decision aid for training and recovery, not a diagnosis: an injury, or the pain that announces one, remains a matter for a doctor or a physiotherapist.

We tested the prototype on historical professional tennis data: about 180,000 matches involving some 6,800 players. The injuries it had to anticipate were those reported in the press. On that data, it flagged more than 80% of them within the following 14 days.

Press reports are a practical reference at that scale, and an imperfect one: they capture the injuries of professional players that become public, and miss those that don’t. The study will have to work from more complete injury records.

In statistical terms, that is sensitivity: the share of real injuries the model caught. On its own, it is not enough. A model that raised an alert for every player, every fortnight, would catch every injury and be useless. What matters just as much is precision: how many alerts turn out to be real injuries. We have not measured it yet.

Published research explains the caution. A study on professional football players found a model that detected around 80% of injuries with about 50% precision: one alert in two was a false alarm. A systematic review of 204 injury-prediction models in sport found that none had been externally validated, and could recommend none for use in practice. And screening tests have long struggled to predict injuries, because at-risk and not-at-risk players overlap so much.

In tennis, most injuries affect the lower limbs, according to a review in the British Journal of Sports Medicine. Acute injuries tend to hit the legs, while overuse injuries more often affect the upper limbs and trunk: the kind of damage that builds up over weeks, which is where an early signal can matter.

Padel is growing fast, with more than 35 million players according to the International Padel Federation. Research on its injuries is still thin: a 2023 systematic review found the elbow to be the most common injury site, based on limited literature.

Pickleball counted 24.3 million players in the United States in 2025, up from 4.2 million in 2020. A ten-year study of US emergency departments found that injured players had a mean age of 64, that most injuries came from falls, and that the wrist was the most injured body part.

Three sports, three very different populations: professional players, fast-growing recreational communities and older adults. So far, the benchmark covers professional tennis only.

Physical load is only part of an athlete’s picture. A meta-analysis in elite sport found symptoms of anxiety or depression in 34% of current elite athletes, and the International Olympic Committee now recommends routine screening and monitoring of mental health with validated instruments.

That is why the same twin can also carry voice analysis from our partner Virtuosis AI, in pilot in MyTwin’s private beta: signals associated with stress, anxiety and low mood, read alongside physical data. Like the injury-risk signal, it is information to act on with a professional, not a diagnosis.

The scientific study will take the prototype where the first benchmark could not:

• measure false alarms as carefully as missed injuries;
• rely on more complete injury records than press reports;
• test the model on data it has never seen;
• have the method and results reviewed by the scientific community.

Padel and pickleball, whose players and injuries differ from professional tennis, will need their own data. Each of these steps will be told here, as its own news.