The history of competition is by necessity also the history of cheating. This maxim’s truth is especially apparent when applied to cycling. For over a century, professional riders have sought illegal advantages through chemical, biological, and mechanical means. The effects of doping are more pronounced in a fitness contest like cycling than sports that hinge on delicate skill, so it shouldn’t be a surprise that amateurs in masters races regularly get popped. Since COVID-19 has trapped cyclists indoors, a spate of cheating scandals has surfaced among riders on the virtual racing game Zwift, and they’re wholly unlike their analog predecessors.
Zwift is an online game and social hub where cyclists hook their bikes up to smart trainers and race against each other. Riders must use trainers with power meters to measure their true energy output over the course of a race, and Zwift displays everyone’s avatars jostling next to each other like it’s a real race. If this sounds like a reskinned Mario Kart with an actual bike as a controller, well, even riders at the first-ever world championships could use temporary powerups. The first Zwift cheating scandal on my radar involved a potentially wonky path to obtaining a rare in-game item.
Zwift racing has been growing in popularity for years—Barry Bonds regularly logs Zwift rides on Strava—and the company raised $120 million in Series B funding in Dec. 2018. That was a few years before the pandemic, which has only accelerated its growth. Last December, UCI held the first-ever esports world championships on Zwift, complete with a first prize of €8,000. The second season of the Zwift racing league started last month, and with higher stakes comes a more pronounced incentive to cheat.
On Tuesday, Zwift announced six-month bans against two elite riders, both of whom finished in the top 10 of their national road race championships in 2020. Both riders were sanctioned for the same reason: The dual power meter data they submitted to Zwift did not match up. Antonina Reznikov, bronze medalist in the Israeli road race championships, was flagged by Zwift’s automated systems for submitting data that showed a 32-percent increase in power during a Zwift Racing League event (she finished fourth in the race). “Zwift discussed the issue with the rider, and although they initially denied editing the file, they eventually acknowledged that they had indeed changed their power data before submitting it to Zwift,” the Zwift Performance Verification Board wrote.
Selma Trommer was also banned for wonkiness with her power meter data. In her case, her data showed a nine-percent increase, though Trommer and her management claimed that the discrepancy arose from a software issue that forced them to upload the data through a non-standard platform. The Board did not buy this, and Trommer’s team eventually acceded and fessed up to editing the power meter values.
Both of these cases show how virtual racing and complicated remote verification measures open all sorts of doors for cheating. Zwift requires riders to submit precise height and weight measurements, as well as record their power data on two devices. A rider’s power-to-weight ratio is the single metric that most accurately captures effort and energy output, which is why a small tweak of either input will alter a rider’s Zwift performance. Say I weigh 175 pounds, and I average 250 watts for an hour. My W/Kg will be higher if I list myself as 10 pounds lighter. It will also be higher if I swapped in the power meter reading from a different ride where I put out 275 watts over the course of that hour. Theoretically, dual power meter verification—one on their trainer or bike computer, and one through Zwift—keeps riders honest, because the two measuring devices do their jobs in different ways.
At their core, Trommer and Reznikov’s violations are straightforward to understand. We don’t know exactly what they did to gin up some extra power in Zwift, but the system caught big enough discrepancies that funny business was obvious. However, two Zwift cheating sanctions from 2020 show how fine-tuned their detection process is, and how the labyrinthine verification process does not necessarily lead to clarity.
Last September, British triathlete Lizi Duncombe was banned from competition for six months after a race. Per Zwift, she initially failed to upload a second power meter file, because she said its battery died after 90 seconds. When she eventually did, regulators noticed a one percent discrepancy. They also discovered that a certain, telling file name could hint at data manipulation.
It contained a “Version ID” label of “562”. This number is given to FIT files produced by Zwift. FIT files produced by a Garmin Edge 820 use a value of “1250”, and examination of multiple other dual-recording files from the rider from different races show they all contain values of “1250”.Zwift
Duncombe has staunchly maintained her innocence, and wrote, “In hindsight, I should have just stopped after sending the original file which only contained a small section of recording and taken the single race DQ if that was going to be the outcome.”
When Zwift announced Duncombe’s ban, they also announced a six-month ban against Israeli rider Shanni Berger. In Berger’s case, she finished second at a Zwift event that required riders to record power data on a smart turbo trainer, not a power meter. The file that was eventually uploaded had a zero percent discrepancy, which is theoretically impossible, and Zwift’s decision got into very detailed breakdowns of data encoding. “The Board further consider that it is beyond reasonable doubt that the log.txt file provided by the rider as evidence of the pairing they used was edited after the event to change the primary power source,” the board wrote. Berger, like Duncombe, maintains her innocence.
I’m just devastated, I’m depressed, I feel like I’m wrongly accused. I guess you can understand how it feels, I just don’t know what to do. I don’t understand computers so I cannot defend myself. I mean, it’s me against Zwift. I don’t want to say it but it’s a little girl against the big company. I do not have the technological means nor the financial means to prove my innocence; Zwift is a rich company with both financial and technological resources. It’s not a fair game.Cycling Weekly
When a rider’s guilt or innocence hinges on lines of code in a log.txt file for a backup power meter, maybe the issue is that virtual racing itself is new and strange. Cycling has had plenty of problems with riders, out on a road, racing against each other already, and when the outcome isn’t determined by who crosses a finish line first, but who puts out the most optimal number of W/Kg while pedaling on a stationary bike in their garage, there’s a built-in degree of inscrutability. As Zwift racing grows and prize pools grow with it, more riders will certainly try to cheat. Given the detection methods, they’ll have to get even more creative.