Scientists have developed an AI model that allows cycling teams to fine-tune the diet of their riders. The model takes into account, among other things, the course, the weather conditions and the energy consumption of each rider.
Artificial intelligence has now also entered the cycling peloton. With the help of AI, scientists from Maastricht University and software company Visma have developed a statistical model with which top cyclists can accurately estimate how many calories they need for a stage in the Tour de France, for example. The researchers worked together with the Dutch cycling team Team Jumbo-Visma. Their results are described in a prepublication.
During the Tour de France, riders burn about 6,000 calories a day. To achieve success in such a competition, it is very important to take in the right amount of food. That is why most cycling teams employ various chefs and nutritionists.
READ ALSO
We don’t need miracles to save the world
If we focus on revolutionary solutions such as carbon capture and geoengineering …
‘The way they predicted the energy intake of the riders was… well, not too efficient,’ says first author Kristian van Kuijk, master student of data science at Maastricht University, also works at Visma. ‘They calculated that purely on the basis of experience. There was no real reasoning behind it. That while you have to get it just right – every small improvement naturally helps to achieve profit.’
Calorie requirement
The researchers used data from previous competitions. These include information about the stature and energy consumption of each rider, as well as race information such as the route, the altitude profile, the weather and the wind direction. Of machine learningtechniques (a form of AI), the researchers analyzed this data. For example, they developed a statistical model that can be used to estimate the calorie needs of every rider on every possible course.
The researchers then tested the model in an experiment. They asked trainers to estimate the calorie needs of riders for some stages of the 2019 Tour de France and Giro d’Italia. They compared those estimates with estimates from the model.
The researchers tested both sets of estimates against the actual calorie needs of the riders in those rides. This always resulted in a score between zero and one. The coaches scored an average of 0.55, while the model scored 0.82.
Vingegaard
Team Jumbo-Visma was so impressed by these results that the team started using the model for its diet planning. The team did this during last year’s Tour de France, in which Jumbo rider Jonas Vingegaard rode defending champion Tadej Pogačar on a climb and then won the entire race. According to Van Kuijk, it was due to a nutritional error that Pogačar had no energy left.
“We know that other cycling teams have since picked up the use of AI,” says Van Kuijk. ‘Now it’s the big hype in the cycling world.’
Read also: ‘The wind tunnel was the puzzle piece that was missing for Tour winner Jonas Vingegaard’
He believes the AI model will be of little use for diet planning for non-athletes, because most people, unlike professional cyclists, do not generate a wealth of data on a daily basis. ‘Of course you can simplify the model, but that comes at the cost of accuracy,’ he says.