MIT's AI Helps Olympic Skaters Perfect Their Jumps

Figure skaters make quadruple jumps look easy, but landing on a blade just 5 millimeters wide after spinning through the air takes incredible precision. Now, MIT researchers have created an AI system that could help athletes finally master the sport's most difficult moves.

Jerry Lu, a former MIT Sports Lab researcher, built OOFSkate, an optical tracking system that watches video of a skater's jump and provides specific recommendations for improvement. The mobile app tracks physical metrics like jump height and rotation speed, then compares them to data from Olympic champions and elite athletes.

Team USA figure skaters are already using the technology to sharpen their technical performance. The system catches tiny flaws that might look fine to the human eye but could mean the difference between landing a jump cleanly or falling.

"Skaters can always keep pushing, higher, faster, stronger," Lu explains. The app shows athletes exactly how an Olympic champion performed the same element, giving them a clear target to aim for.

The technology works remarkably well because figure skating movements happen to avoid AI's biggest weakness. Most AI pose estimators struggle with depth perception when analyzing video from a single camera angle, but skating jumps only require understanding height, rotations, and landing quality.

Lu will bring his AI expertise to NBC Sports during the 2026 Winter Olympics in February. He'll help commentators and viewers understand the complex scoring system in figure skating, snowboarding, and skiing, using AI to explain why judges award specific scores.

Why This Inspires

This technology democratizes elite coaching in a sport where access to top-level training often determines success. A skater in a small town can now compare their jumps to Olympic data and get feedback that previously only came from expensive coaching sessions.

Professor Anette "Peko" Hosoi, who co-founded the MIT Sports Lab with Lu, is now exploring whether AI systems could eventually evaluate the artistic side of figure skating. Her team wants to understand if AI platforms assess beauty the same way humans do, or if they're just mimicking what they've heard people say.

Figure skating provides the perfect testing ground because unlike paintings in a museum, every performance gets a numerical score. The research could reveal fundamental insights about how both humans and machines understand aesthetics.

As for whether we'll ever see a skater land a quintuple jump with five full rotations, the MIT team's data might soon provide an answer.