How Machine Learning Helps Team Jumbo-Visma “Fuel Up”
Those who follow us on social media know that we’re passionate about sports, especially when it comes to Team Jumbo-Visma. We are the proud title sponsors of this professional speed skating and cycling team. Of course, our sponsorship would not be complete if we didn’t add Visma Connect skills to the mix. In this article, you’ll learn how our Data Science team applies machine learning to help the cycling team optimise performance.
Fueling up to the finish line
According to a Business Insider study, Tour de France cyclists burn an average of 6,071 calories per day. As you can imagine, good nutrition matters to ensure that athletes can gain back the energy they need to go on with the race. This is especially important during a grand tour, which takes up to three weeks. For such an event, Team Jumbo-Visma travels with a cook and a special bus, where they stock all the ingredients they will need for meals in advance. The main title Jumbo, a popular supermarket chain in the Netherlands, supplies the goods. To know exactly what kind of ingredients they need and how many, coaches need to estimate the calories each cyclist will burn. During big races, cyclists typically have breakfast, two recovery meals, dinner, a snack in the evening and sometimes a pre-race meal as well.
In the past, coaches would start preparing their calorie estimates more than three weeks in advance to make sure they had adequate estimates per cyclist, per stage. Using information like the stage’s profile, cyclists’ body-mass-index, elevation, and total distance, they would manually arrive at their calculations. But out on the track, unforeseen factors almost always impact how much energy gets used up. The weather, for instance, can cause cyclists to exert themselves more, or perhaps the team’s tactics need to change due to other circumstances. This means coaches would often have to review their estimates before each stage of the race. Needless to say, this was a very time-consuming exercise. Working together with the team, we embarked on a journey to improve this with automation.
Machine learning to the rescue
The first step was to collect the necessary data. A Garmin device on the bike gives us the actuals (like total distance and climbing meters) for each race. A crank based power meter also gives us a precise calculation of the calories burned. We have a TCX file with GPS coordinates which gives us metrics on the profile of the race (distance, climbing metres, difficulty, etc). We also have information about riders’ individual weight, height, and role (whether they are a sprinter or a climber, for instance). Finally, we take weather conditions into account. Combining the weather forecast with GPS information and the location of the rider allows us to calculate the wind effect - whether it’s a tail wind or head wind.
Data gathering and visualisation is powered by Smartbase, a data management and analytics platform for athletes. Coaches use this environment to enter the actuals. We then use this data to prepare the forecasts. Part of the work also entails pre-processing this data and removing outliers. For example, a cyclist might forget to turn off their Garmin device, so we need to exclude the “noise” from the data set. We also made some variables relative, like power, energy and elevation, to make it easier to compare stages and races.
Using training examples, we apply supervised learning to teach our algorithm what the outcome of the calorie predictions should be. This is a regression problem, and we selected random forest as the best machine learning algorithm to solve it. 
The results
So, how accurate is the resulting model? We have used different metrics to evaluate its performance. For example, we looked at R-squared of our model, also in comparison to the manual predictions done by coaches. R-squared measures the strength of the relationship between the model and the dependent variable (calories) on a convenient 0-100% scale. The machine learning model got a score of 82% while the manual predictions got a score of 52%. Coaches now get the results in a split second, so they’ve gained time savings as well as accuracy.
Smart Food Coach App
With the calorie estimates powered by machine learning, coaches need to enter riders’ calories into Jumbo’s smart Food Coach app, which provides sample meals with optimised proportions for each meal in return. That’s how our team keeps winning!
Would you be interested in improving your nutrition? Check out the Food Coach App on the Apple App Store and Google Play. Feel free to reach out if you’re also focused on improving your business with automation.
