machine learning

The cycling predictions by Mortirolo are purely based on a statistical model, not on (expert based) opinions. The method used is a combination of Bayesian modelling and machine learning. The statistical model uses the results of the same (type) of races during the last ten years to determine which individual and team characteristics are important predictors for the classification of that race. Subsequently, the statistical model is used to run 10000 simulations of a race and produces probabilities for each rider to win, end up on the podium and in the top 10 and a predicted classification.