Cycling is on the verge of a data revolution
There is no doubt that we live in a data-driven world. Everyday, data helps advertisers determine what we like to buy, the fastest route to our destination, or even what medicines to take if we get sick.
“Big Data” has also had a tremendous impact on sports, from baseball to Formula One. The result: coaches know which players to use in key situations to provide the greatest odds of winning, and race crews can refine every aspect of their car to squeeze out every bit of speed.
Yet, the Cycling world has kept using what is essentially the same performance paradigms for decades, while the technology around us has exploded.
Interestingly, the type of data cyclists collect (wattage, heart rate, GPS, etc.) lends itself perfectly to “Big Data” techniques (Fig. 1). Why should Cycling be any different from these other sports? It’s true that “Big Data” has begun to leave it’s mark on Cycling. Although the impact thus far has been largely in the fringes, with the bulk of the expertise/knowledge concentrated among a precious few. This has left the majority of us to plod along as we always have. Therefore, we are asking the question, what would happen to the sport of Cycling if all riders had access this technology? The answer, we believe, is that Cycling is will undergo a data revolution, and here’s why.
Cycling produces “Big Data.”
In the general sense, “Big Data” are extremely large data sets analyzed by computational methods (e.g. machine learning and artificial intelligence) to discover patterns, relationships, or trends that would otherwise be undetectable. How does Cycling fit this definition? First, the devices used during training (power meters, heart rate monitors, GPS, etc.) capture extremely high-quality data at a second-by-second resolution. This means, during a single stage of the Tour de France a rider could generate over 100,000 data points! Most of us aren’t riding the Tour, but even an amatuer produces millions of data points during a season. Yet despite this mountain of information, most of us just review a few summary metrics (average watts, heart rate, or speed), and judge the ride based on gut intuition. Even for those who use a more rigorous/scientific approach, the current standards set by data analysis software options are not designed to leverage this flood of information. Thus, the bulk of your data goes unused; begging the question, what is hidden in all this unused data and how can we use this information to maximize our true potential?
Emphasizing the individual
Perhaps the most important benefit to using machine learning is that we can emphasize the individual. We can do this by mining each rider’s data to reconstruct their range of current abilities, as well as their best-ever abilities. This facilitates the creation of intuitive and fully personalized metrics so athletes know exactly how hard they’re training. In addition, these approaches allow us to directly measure aspects of physiology that traditional methods approximate. For example, it is well-known that a given ride or race can have highly variable ranges of wattage, which is the result of changes in terrain, race situations, or other ride-specific conditions. Weighted or Normalized Power are metrics which seek to account for the physiological impact of this variation in pace through mathematical transformation of the average wattage. Though this metric has utility, it is at best an estimate and not a measurement. However, through a bit of creativity and modern computational power, we can now directly measure an athlete’s maximum ability to tolerate variation of pace for any length of time, and report back any given ride as a percentage of their maximum ability (Fig. 2).
New Models of Human Performance
Machine learning can uncover the uniqueness of the individual athlete through algorithms that determine what factors drive your good versus bad performances. This is greatly facilitated by the creation of fully-personalized metrics. We believe this is critical, as the most popular methods for tracking performance make significant assumptions that athletes respond to training the same way. However, we can now move away from one-size-fits-all approaches, and the result is that athletes will be able to more accurately understand their own biology and boost their performance.
In short, we believe Cycling is on the verge of a major advancement. Incorporating “Big Data” methodologies, can teach us more about ourselves and the sport than ever before. Importantly, these advancements do not mean that we all have to become expert data scientists, make our lives more difficult, and disrupt the joy of riding our bikes. These approaches can be entirely automated, requiring no expertise whatsoever. If executed well, we believe that these concepts will make everyone’s life easier and help to inspire confidence that your training is efficient and specific to your goals. This is the reason we built Enfusion. From day one our goal has been to bring “Big Data” to the cycling masses in a simple package that will help each of us reach our unique potential. For Cyclists that want the ultimate in personalized performance monitoring, the future has arrived.
About the authors: Daniel Corum, PhD and Daniel Rotroff, PhD are scientists and former competitive cyclists. Their research interests include bioenergetics (mitochondrial metabolism), and bioinformatics with focus on precision medicine. They are also co-founders of the Enfusion Training software, which is available at www.enfusion.training. You can find them on Twitter here.