The 2022 World Cup will be remembered for several reasons. It’s likely the last hurrah for superstars Lionel Messi and Cristiano Ronaldo and it’s the first time an Arab country has hosted the tournament. But another first might be the most consequential for soccer itself. Artificial intelligence is teaming up with flesh-and-blood referees to make the offside calls.
Soccer’s offside rule is notoriously complex. In the broadest of strokes, it penalizes an attacking player who receives the ball while being closer to the goal than the opposition’s second-last defender, but there are endless nuances. The accuracy of a decision also depends on unimpeded sightlines and split-second spatial recognition by the refs. Correctly made or not, a call can be the most controversial moment of an entire match.
“One of the worst feelings in the world is when your team gets knocked out because of a bad call,” says Ellen Hyslop, co-founder and head of content at The GIST, a women-owned and operated sports media company. “I think this technology enhances the integrity of the game.”
Semi-automated offside technology creates a 3D image of the action using 12 grandstand cameras that track 29 points on the bodies of every player on the pitch and a ball that contains a sensor sending out location data 500 times per second. The AI software generates an alert when a player commits a possible offside infraction, but the call is still ultimately confirmed by human referees. The 3D picture goes on the stadium’s big screen for all to see.
Just 158 seconds into the opening game of the tournament, the technology was used to rule out a goal by Ecuador against hosts Qatar, which if it had stood would have been the fastest-ever goal in a World Cup. Old-time fans may grouse about the “robot refs.” Millennial audiences might not even blink. But one thing’s for sure. There’s no turning back.
The debut of AI on soccer’s biggest stage is the most redolent example yet that futuristic tech is becoming an integral part of sports. Tennis uses the Hawk-Eye computer vision system to track if balls are in or out. Major League Baseball expects to adopt an automated strike zone system in 2024, though it’s not yet decided how robot umpires calling balls and strikes will be integrated with play.
But some of the biggest innovations have been happening behind the scenes. Teams, which first used data primarily for training and injury management, now develop their own proprietary algorithms for pre-game preparation, in-game adjustments and postgame analysis. And owners are eager to adopt cutting edge AI tools to manage stadium operations, ticketing and merchandising. One research firm predicts the global market for AI in Sports will soar from $1.4 billion in 2020 to $19.2 billion in 2030. Tech is now so important that major franchise owners like Maple Leaf Sports and Entertainment even have their own incubators supporting entrepreneurs with promising new ideas.
Data usage has been associated with sports since the early 2000s Moneyball era, when baseball’s payroll-challenged Oakland A’s began to apply computer driven analytics to player evaluation. They were able to identify underappreciated — and less costly — skillsets that allowed the club to compete with the Major League’s biggest spenders.
Former Toronto FC captain Steven Caldwell, who is in Qatar as TSN’s soccer analyst, was first introduced to sports data when playing for Wigan Athletic in the English Football League in 2010. “We’d come in each morning and get a booklet of stats. You’d see the distance you covered last match, the passes you made,” he says. “As a player, I could see that any small gain might give you the edge you need.”
“Of course, that’s all very clunky compared to what they have now.”
Today, when every metric imaginable can be measured, pro and college teams are increasingly using AI and machine learning to improve player performance. In October, the Toronto Raptors showed off their massive new multimedia analytics board at the OVO Athletic Centre, the team’s practice facility. The three-metre high, 37-metre-long series of screens uses computer vision and proprietary algorithms designed by U.S.-based Noah Basketball to track a player’s shooting motion. The arc and trajectory of every single shot is recorded and analyzed to generate more swishes than misses.
Not everyone can afford a multimillion dollar, 448-screen monster to help them work on their game. “These technologies are very resource intensive,” says entrepreneur Marianne Bell. “The capital and skillset required to build them can be prohibitive.” Bell is co-founder of Spensor Tech, a Waterloo-based startup that pitches its sensor-and-data-platform product as being suitable for everyone, from elite athletes to casual players.
The company’s sensor, weighing only one gram and no bigger than a fingernail, is durable enough to capture high-fidelity motion data from every swing, hit, kick and shot. Last year, its first commercial application came to market. BOWdometer (pictured above), an archery practice tool which collects 5,000 data points from every draw and release, sells for $140. It was quickly adopted by members of Canada’s national archery team, including four-time Olympian Crispin Duenas.
Spensor Tech is currently working with several sporting goods manufacturers to put its technology into golf clubs. “This is a bit of a different process because BOWdometer was a retrofit. The sensor was installed on the bow,” Bell says. “Here, the clubs will be built with the sensor inside. So, there are mechanical questions to answer. How many sensors do we need? Where’s the optimal part of the club for data capture? Is it the head? The shaft? The grip?”
Bell, who expects to pilot the smart clubs in 2023, believes they will lead to better scores than a swing simulator because the data reflects performance in actual gameplay. “This is not something that simply recreates a course environment,” she says. “Here, you can go out and play round of golf just as you always do, but your clubs can collect thousands of data points every second that might help take three strokes off your game.”
Beyond simply improving skills, Caldwell believes high-tech tools are the future of athlete recruitment. His other gig is CEO of Best Athletes, another Waterloo-based start-up that develops sports analytics and video platforms for youth soccer. “I think the way we use data can help democratize sports,” he says. “Not every player has the same opportunity to get noticed. Everyone should have a fair crack at a Division 1 college.”
Best Athletes’ software measures 45 metrics from four categories — physical, cognitive, technical and tactical — and allows players to benchmark their skills, and then show them off.
Coaches from North American colleges, organizations and clubs looking for a certain type of player can scout any of the 3,400 who have signed on with Best Athletes. Who knows? The next Messi might come from Regina, Saskatchewan rather than Rosario, Argentina.