Machine learning team outsmarts 40 Ortec Finance employees in Tour de France competition
Ortec Finance organizes a yearly Tour de France competition. Employees can guess which riders are going to score best during the Tour de France. And if they get it right, they take home a nice prize.
Coming up with the idea
Our colleague Oedsen has participated in the Ortec Finance Tour pool for 3 years now. The first two years he used all his cycling knowledge to come up with the best team possible. He finished a promising third the first year, but despite preparing even more, the second year didn’t go so well. He turned out to be only 24th. He was really disappointed and started thinking about alternative ways to predict the outcome of the Tour de France.
Pitching the idea
Oedsen quickly came up with the idea to use machine learning to win the Tour de France competition. He pitched his idea during the Tech Labs Hackathon, and it turned out that there were more Tour de France enthusiasts. So they formed a team to work on the idea. The team consisted of Daan, Boyd, Eric and Oedsen.
Collecting Data
First the team needed to collect data. They asked Roy, the organizer of the Ortec Finance Tour pool, for all the data of the competition of the past five years. They also gathered results of cycling races that took place before the Tour de France, such as the Giro d ’Italia. They looked at the performance of the cyclists in former cycling races, but also in former years of the Tour de France. They used this as input for the machine learning program and the data from the tour pool as desired output.
Testing phase
After gathering the data, the team ran a test. They tried the predict the results of 2017 and checked it with the actual results. It turned out that the program wasn’t that accurate yet. They would finish 20th if they would have entered in last year’s competition. Oedsen decided that this was not good enough. He made some changes to the input data during his holiday. This improved the result: the team generated by the machine learning program would have finished fourth. This improved the confidence of the team and they entered this year’s competition with their machine learning-generated team. They named their invention Team SkAI.
The results
During the first stages of the Tour de France it didn’t look so good for team SkAI, with a 22nd place in the intermediate standings after one week of racing. But when the mountain stages started their results became better and better and they steadily climbed the ranks, taking over the lead after stage 16. In the end they managed to win the Ortec Finance Tour pool!
Development
Oedsen is certainly planning to compete again next year. He and his team already have some good ideas to improve their program and get even better results next year.
