From gymnast-tracking to pole vault measurements mid-jump, the watch brand’s Swiss Timing division has a whole host of new timing tech for Paris 2024.
COURTESY OF OMEGA
On August 27, 1960, at the Olympics in Rome, one of the most controversial gold medals was awarded. At the 100-meter freestyle men’s swimming event, Australian swimmer John Devitt and American Lance Larson both recorded the same finish time of 55.2 seconds. Only Devitt walked away with the gold medal.
The way swimming was timed was by using three timers per lane, all with stopwatches, from which an average was taken. In the rare occurrence there was a tie, a head judge, in this case Hans Runströmer from Sweden, was on hand to adjudicate. Despite Larson being technically one-tenth of a second quicker, Runströmer decreed the times were the same and declared for Devitt.
It was this controversy that, by 1968, had led to Omega developing touch boards for the ends of swimming lanes so the athletes could stop timing themselves, removing any risk of human error.
Alain Zobrist, head of Omega’s Swiss Timing—the 400-employee branch of Omega that deals with anything that times, measures, or tracks near enough all sports—is full of stories like this.
How, for example, in 2024, the electronic starting pistol is now connected to a speaker behind each athlete because, in staggered-lane races such as the 400 meter, those athletes in the furthest lane previously heard the starting gun a fraction later than those closest to the gun, giving them a disadvantage.
Or how, when photo finishes were first used in the 1940s, it would take nearly two hours to come to a decision because you had to develop the footage first. Now Omega’s new Scan-o-Vision can capture up to 40,000 digital images per second, allowing judges to make a call in minutes.
To split hairs—or indeed seconds—Swiss Timing hasn’t really been in the business of simply timing a race for a very long time. Despite the Omega logo being on every timing device at every Olympics since 1932 (except for when Seiko got a look in in 1964 and 1992), what Swiss Timing does is much more than just start and finish times. “We tell the story of the race, not just the result,” Zobrist says. As for Paris 2024, that storytelling has got quite a lot more plot lines than before.
“2018 was pivotal for us,” says Zobrist. “That was when we started to introduce motion sensors on athletes’ clothing, which allowed us to understand the full performance—what happens between start and finish.”
These sensors didn’t show just distance, but also step counts, top speeds, acceleration, deceleration, and tailwinds. In swimming, they could monitor stroke count. Some 2,000 data points, in a monitor weighing 12 grams, transmitted within less than a tenth of a second. This data could then be passed on to coaches and the athletes themselves, to use as a training tool.
For Paris, Swiss Timing is, of course, using AI to analyze that data, to attempt to make sense of it in real time, so it can create a race narrative that can be used not just by those participating, but also by social media channels and broadcasters.
The biggest change is the ability to use historic data. “You can now input data from previous races, and use that to make predictions about how an athlete will perform,” says Zobrist. “AI can now analyze single performances and compare that to the others. We’re taking a picture of time, or rather what’s happening through time.” Zobrist has stated that Swiss Timing will be able to identify the favorites far more accurately, track and analyze their performance mid-race, and then compare this data with the other athletes, then map the reasons why an certain athlete won or lost at every point of that event.
But aside from whizzy new onscreen graphics and improved data analysis, perhaps the most impressive innovation is what Zobrist and his team can do with body-imaging cameras. Four years ago, when body imaging was introduced, the visuals were primitive, similar to line drawings done by a child. Now Swiss Timing has what it simply calls “computer vision”: single or multi-camera systems that feed images into AI models specifically trained for each sport to create fully realistic bodies (think the Silver Surfer from the Fantastic Four, but in Omega’s trademark red).
As an example, Zobrist shows videos of these AI-generated bodies spinning in a perfect dive, performing gymnastic twists and turns. In diving, the systems will track the athletes from the beginning to end of their dive, producing a 3D image using a specific mathematical algorithm, while at the same time generating image data and metrics such as speed of entry into the water. This data should also allow judges to see the distance between the diver and the board—a specific distance that was previously judged by eye, and that, if not maintained, is a reason for points deduction.
In tennis, in addition to tracking the players in real time, it will show the reaction to serve—how quickly it is returned, racket position, and accuracy. The data will be used to determine whether there is a correlation between a player’s success and their reaction to a serve, rather than just judging the outcome of a match on the power of serves.
For the pole vault, the AI system will measure the gap between the athlete and the bar to work out what effect that has on jump quality.
COURTESY OF OMEGA
In pole vault, for the first time it will measure the gap between the athlete and the bar to ascertain what effect that has on jump quality. For gymnastics, the multi-camera system will track the gymnasts’ bodies as they spin, tumble, and fly. The AI for this sport has even been specifically primed to track the human skeleton, so the angle of the athlete’s feet can be checked to see how it affects jump rotation.
“We use a different number of cameras for each sport, depending on what we’re wanting to measure. Therefore, with more or less cameras on-site, it changes the number of computers we need,” explains Zobrist. “As an example, swimming events have four cameras in play, while beach volleyball has up to 16. The unique aspect of our system, however, is not the computational power, but the software we’ve developed. We develop algorithms for each sport, and train AI models for the individual use cases, and that’s really the heart of our computer vision technology. The computers themselves are standard devices.”
Swiss Timing says its new software has been fed with data from several years of numerous other sporting events in order to detect the movement of the athletes correctly. It has collated data from local sporting events, national championships, and also bigger international events, such as the Diamond League in athletics, for which Omega is also the official timekeeper.
However, like all computer systems, AI is not infallible. At the Olympics, it’ll be used in a situation where something going wrong could mean the difference between being on the podium or on the plane home.
In tennis, the AI system will show the players’ reactions to serve, looking at how quickly it is returned, racket position, and accuracy.
COURTESY OF OMEGA
“Our software is developed to detect any unexpected deviations in the measurement process,” claims Zobrist. “The timing team can then correct it manually, or the system can do it automatically. We’re never losing any data. It’s always stored and managed for correct analysis.”
And despite Paris’ starter gun being just days away from being fired, preparations are already starting for the next two Olympic Games. But Zobrist is tight-lipped on what innovations will be seen. “We’re at the beginning of this evolution in sports timekeeping, and Omega has a clear roadmap in action for future Olympic Games, such as Milano Cortina 2026 and Los Angeles 2028. However, as we’re still in the development phase, we can’t reveal any of those details just yet.”
But the invention of the tap board in swimming removed the need for a jamboree of judges and timers around the pool. So, as technology becomes more sophisticated, does Zobrist think the human element in sports timing will soon be obsolete?
“There will be a shift, and we do want to remove the human reaction factor,” he says. “But you’ll always need humans to set up. Even if it’s just to control what the machines are doing.”
