The English Premier League (EPL) will scrap its Video Assistant Referee (VAR) system from the 2024-2025 season and introduce a new system that will include dozens of iPhones to help with offside decisions.
Genius Sports designed the new semi-automated offside technology, internally dubbed “Dragon.” According to Wired, the previously used VAR system was notoriously controversial, leading to “significant delays and human process error” and “concerns about the accuracy of in-game calls.” Genius Sports has worked with NBA basketball for years, “working with optical tracking and data.”
Genius starts in EPL With 28 iPhones in each stadium, Genius said it may add more iPhones to certain stadiums later this year. Wired reports that these will mostly be iPhone 14 models, but The Verge reports that Genius CPO Matt Fleckensteinsaid in an interview that the iPhone 15 Pro will be used for the job.
The iPhones are housed in waterproof cases with cooling fans and power outlets. Two or three mounts that can hold up to four iPhones are set up around the stadium. The mounts are movable but typically stay in place during games to record video from specific angles.
According to a report in Wired, Dragon has “7,000 to 10,000 [data] “It constantly awards points to each player,” taking into account specific characteristics such as “muscle mass, skeletal differences, and even gait,” all of which contribute to the possibility of an offside. The Verge adds that other VAR systems only award players 30 or 40 points, according to Fleckenstein. According to Wired, Dragon can capture up to 200 frames per second, while today’s video technology maxes out at 50 or 60 FPS. The system will max out at 100 FPS to “balance latency, accuracy, and cost.”
It was impressive to see the new system also be able to predict offsides. During such events, the camera frame rate automatically increases to closely track the potential offside. Immediately after the event, the frame rate is reduced to save power. It also appears that Dragon can feed all the data it collects into a machine learning system running behind the scenes, internally called “Object Semantic Mesh.” This system uses intelligence to analyze the data and learn from it to get smarter.
