Football challenge


As I was giving a short lecture on Monte Carlo Markov Chain techniques, I ended up my course with a cool Bayesian hierarchical model for football scores. As I am fond of football, I was really excited about that and decided to play the game a bit further by creating this page where I would put my prediction for the Ligue 1 in due time. Check out if the model performs well ;-) ALLEZ L'OL !

Here are my forecasts for the 2016-2017 season of the Ligue 1. I will update (well my code will do it automagically) the results on a weekly basis---and this will also include an update of my model if I realize that the later is terrible!!!

Table: Forecasts for the next day of Ligue 1.



If you look at the above pronostic table, you will see that there are little "draw predictions" since was expected from the model and is clearly one of the major weaknesses of it!



Final results: For the whole 2016-2017 season, the model has an average performance of 53% for 1N2 bets and 15% for "exact score" bets.