American figure skater Andrew Torgashev recently experienced firsthand the precision of new AI technology at an invitation-only camp. While attempting a quad toe loop, a four-revolution jump, a camera-based app instantly revealed he landed a quarter-revolution short, an error invisible to the naked eye but critical in a sport where minute differences in points determine success. This innovative app, called OOFSkate, aims to revolutionize figure skating training and potentially influence future judging.

Developed by Jerry Lu and Jacob Blindenbach, two computer specialists with no prior figure skating background, OOFSkate utilizes AI-powered computer vision via a tablet or mobile phone. It analyzes jump height, rotation speed, airtime, and landing quality without requiring sensors or wearable technology. The vision behind OOFSkate is to automate the technical calling in figure skating, allowing human judges to focus solely on the artistic components of the sport. Lu emphasized that the system functions as a semi-automated technical assistant, measuring specific elements of a jump, and also serves as a valuable coaching tool for instructors nationwide.

The system’s operation is deceptively simple: it captures a skater in motion using a standard camera, then overlays key points of a jump or spin, comparing them against an idealized version of the element. It instantaneously records metrics traditionally used by technical panels, providing immediate feedback. Coaches and judges can determine if a skater completed the correct number of turns for a triple lutz or landed on the appropriate blade edge for a salchow. Beyond technical accuracy, it measures jump height and spinning speed, both crucial judging criteria. Skaters themselves benefit significantly, able to compare their current practice jumps against past performances or even benchmark their technique against elite athletes like Mikhail Shaidorov. U.S. Figure Skating plans to integrate a team library, allowing athletes to compare their skills against extensive benchmark data collected over years.

Lu and Blindenbach, who met as swimmers at the University of Virginia, initially explored technology to aid aquatic athletes. Their paths diverged, with Lu joining the MIT Sports Lab and Blindenbach specializing in artificial intelligence at Columbia, but they remained in contact, driven by a shared desire to apply emerging technology to Olympic sports. NBC, the U.S. broadcaster for the Olympics, spurred their entry into figure skating, seeking technology to enhance live commentary for analysts Tara Lipinski and Johnny Weir. U.S. Figure Skating quickly recognized the project's potential for athlete training, with Olympic skaters Jason Brown and Alysa Liu, alongside her coach Massimo Scali, providing valuable feedback. The system undergoes regular testing at the Skating Club of Boston, suggesting a potentially revolutionary shift in a traditionally conservative sport.

The name OOFSkate initially stemmed from a common skater exclamation upon seeing disappointing feedback: “Oof, that wasn’t very good!” However, U.S. Figure Skating later ascribed a second meaning: “Obsessed Over Form.” The app’s ability to remove subjectivity from judging form is a compelling aspect, akin to how Hawk-Eye technology standardized line calls in tennis. Blindenbach highlighted that under-rotations should always be called consistently, eliminating controversy.

While acknowledging the potential for AI to enhance fairness by precisely identifying elements like blade edges, Lu and Blindenbach remain cautious about immediate widespread adoption for official judging. The slow pace of technological integration in sports, exemplified by Wimbledon’s decades-long shift to fully automated line calling, and the presence of established data providers like Omega for the Olympics, suggest a gradual implementation. For now, their focus is on refining OOFSkate as an assistive tool for coaches, athletes, and commentators, particularly as preparations for the Milan Cortina Olympics in February intensify. They emphasize an assistive rather than a replacement role for AI, focusing on quantifiable metrics like jump height or rotation, while leaving artistic evaluation to human expertise.