Opinion: How AI will make pro racing even faster

A classic scene sets up the storyline in many movies: a parent, or friend, reads the private diary of a teenager that’s been tucked away in a drawer, approaches them about what they’ve read or gossips to others about something salacious in it or, even worse, uses it against them. Like a personal diary, an athlete will write about their feelings, health issues, and emotions in their training diary, to document mental, physical, and personal fluctuations and how they affect performance. An honest diary can provide the deepest insights into human condition, which is why they are often protected with lock and key.

The evolution of training diaries

No longer tucked in a drawer or backpack, training diaries that live on the cloud have become the norm for most athletes over the last twenty years. And as data analysis has proliferated in everything from sleep quality to body weight, to heart rate to on-bike power, metrics are now drawn from algorithms to predict performance. Of course, many athletes also type in their daily emotions and health issues, from which trends are also seen, and interpreted.

Michael Barry on performance and the power meter

Through data accumulation and modeling, Ai tools will further our understanding of that data, improving performance and making races even faster. The coach will always have their place because personal contact is important in athlete development. But the tools will allow both the athlete and coach to discover aspects of performance that were not before obvious. Beyond the analysis of biometric data, through an accumulation of film, AI modeling will analyze bike position, cornering effectiveness, pedal stroke, and even race tactics. As is already happening in medical science, AI, will make correlations through the analysis of large data sets that humans have yet to discover: perhaps, training techniques used in football are beneficial to cycling etc.

Stats, stats, stats

Twenty years ago, Michael Lewis wrote Moneyball, the story of Billy Beane , the manager of Major League Baseball’s Oakland A’s, and his use of player statistics to win games. Statistics told a different story of how to win games than what most scouts, fans or coaches believed: getting on base was more important than any other aspect of the game.

A player who got on base consistently, had more value to the team than the superstar who hit dramatic home runs but also struck out. In a game where team budget often determines who wins, Beane turned the A’s into a winning team on just a third of the budget of competing teams. Now, every baseball team, and most sports’ teams, look at every statistic available to determine which athletes will be of most value.
This means the athlete’s greatest asset is their biometric data, which can be used in their favour or against them. In baseball, teams look at a player’s power, strength and velocity and those aspects are measured and monitored. Cycling is no different: power and efficiency are what teams look at first.

The power of data

Baseball players now protect their data closely, to the point of paranoia. Players hire independent companies, such as Driveline, to analyze their biometrics, and to maintain privacy from the teams. In their collective bargaining agreements with Major League Baseball players negotiated how teams can use their in game data. Legally, within Canada, athlete’s have a greater say in how their data is used as sports’ organizations are not allowed to collect biometric data without the athlete’s consent.

In most realms, and especially online, privacy is our greatest asset yet we give it up too easily to feed our egos or make life easier, often it is ultimately to our detriment as that data is taken and used to manipulate us to profit corporations.

At the moment, race performance and training files are the two key metrics teams use to gauge a rider’s ability and potential. As data analysis proliferates through every level of the sport and teams, coaches, agents, and scouts scour the globe to find the next superstar, they are increasingly asking young riders, who, even more problematically, are often minors, to share their training diary access so they can analyze and rate ability.

Using training data to determine future perfomances

Teams then develop databases of training files, watch over potential riders who they might sign through the season, and then pick who they think may be the best. Within those files, a rider may have typed something that will be construed as a weakness and without a scout knowing the context, the rider will be overlooked. Or perhaps, inaccuracies arise if the rider hasn’t inputted enough data to draw clear conclusions, or they don’t input accurately, or their powermeters are inaccurate. Wearables can track sleep data which can provide a view into the cyclist’s social life while not at races, which can also be negatively interpreted without context, leading to judgment or reprimand from the team. Careers can be determined by data which may be false, or misinterpreted.

Often, riders are asked to give up data with no knowledge of how it will be used, or who it will be shared with. If the riders have managers, or agents, the agent should be protecting the rider’s confidentiality, and at least forcing the team to sign confidentiality agreements to who can see the data, and how it will be used. As sports betting proliferates internationally the data has more value and the demand for it will only increase. Already, cycling websites and podcasts that are closely followed by people placing bets on riders use their training data and race data to make predictions.

Training data isn’t always indicative of race performance, and most of the data available through an online file is likely not even indicative of the long term potential of a young rider. This is compounded by the fact that interpretation of data can be varied and not always accurate. A developing rider should not only be judged on physical characteristics but also on race performance, race craft, ability to integrate into a team and on his or her character.

Data isn’t everything in sport

In a sport with countless variables, both mental, physical and environmental, analysts have yet to understand how to interpret the data to paint the entire picture. As a result, there is now a pervasive overconfidence on power data as it is the one way, aside from race results, in which we can quantify a cyclist’s performance and ability. Yet it is often in the countless other variables that are overlooked, or perhaps not understood, or yet quantifiable, that make a great cyclist.

As has already happened in baseball, Artificial Intelligence will simplify the data while also providing depth to our understanding of what parameters matter most and how athletes and teams can use them to their advantage to manipulate performance.

The riders are the team’s asset, and the individual should understand that a team will use data to help a rider become better but will also use the same data to determine their future value and employment which is why the two should be separate and riders should limit access to protect their privacy and value.

Most professional teams mandate that riders use coaches hired by the team to ensure they are training properly. This creates an innate bias where the coach is paid to help the athlete perform while also knowing their strengths and weaknesses which ultimately determine their value. A good coach not only gives training programs and analyses data, but also supports the athlete through mental and physical challenges. Within a team there is little confidentiality between rider, management and coach, as the coach will report to the team management on a rider’s performance.

Few people would be comfortable and open with a psychologist who is paid by their employer. Ethically, and even on a subconscious level, the coach will prioritize the team, which pays their salary, over any athlete. With access to training diaries, that relationship becomes even more one sided, as the athlete has little privacy from their team. Of course, with the knowledge of this imbalance, the athlete may be less honest and therefore the relationship less productive. And, if the athlete leaves the team for another, their new competition now knows every strength and weakness.

Cyclists need to be using training diary platforms on which they can segregate data so that only that which can be properly interpreted, and they are comfortable sharing, is accessible while also understanding their value is in their biometric data, and that AI is coming for that data. With the development of future AI modelling, maintaining control of biometric data will put the riders in control of their future, not their teams.

Michael Barry is a former WorldTour rider, author and co-owner of Mariposa Bicycles