Ghent University has developed an AI model to predict the outcome of cycling races. ‘We use a complex system that works with data from 2017 until today,’ says researcher Bram Janssens. ‘Wout van Aert has the best cards for Sunday.’
‘Wout van Aert to win Paris-Roubaix’: Gent University uses AI to predict cycling victories
Colin Clapson
Published:
‘The cycling predictor is a bit of an out-of-control hobby,’ admits Janssens. ‘We were curious to see if an AI application would work on predicting cycling results as well.’ Janssens has been working on AI applications at UGent for some time.
He started AI testing at U18 races years ago. "We were trying to track talent. The model suggested Tom Pidcock, who has since become a top cyclist. The model also selected four other cycling champions, who are now competing at the highest level.’
Course, specialism and field of participants
‘We use an algorithm from the gaming world and add other parameters. Not only the results play a role, but also the course, specialisms and the field of participants. For example, the Tour of Flanders is not the same as the E3 race.’
‘The road surface and the width of roads also play a role,’ Janssens stresses. The model works on the basis of all results since 2017 and is being updated with new data. UGent is now making a first prediction for this season.
"Wout van Aert is the best rider on the cobbles"
‘The model sees 3 top favourites for Sunday's Paris-Roubaix race. They are Wout van Aert, Mathieu van der Poel and Tour winner Tadej Pogačar. That feels logical, but the system only works on data. It is extraordinary that the AI prediction is so close to our expectations,’ states Janssens.
‘Wout van Aert gets the highest score. He is considered to be the best rider on cobbles. Mathieu van der Poel often scores just a bit better in very big races, but that's not in the model yet.’
Janssens wants to put more parameters into the model soon. ‘We are also trying to integrate information on what happens during the race. That can also play a role to determine riders’ fitness.’
Tool for betting agencies and cycling teams
The AI model could have all kinds of applications. ‘It could be used by betting agencies to estimate the odds before people start betting,’ he says.
‘It could also be a tool for teams that can make probability calculations during the race.’ The cycling world has been using AI tools for some time, including in training or to optimise riders' nutrition'.
The AI model is not infallible. ‘It remains a tool, but one that provides information that was not always available beforehand. It can help make decisions, although it will not replace team leaders," Janssens concludes.