This is a prediction model for the NHL that uses Monte Carlo methods to predict the winning probability of the home team for each game in the season. It does this by sampling poisson distributions of each teams former scoring and comparing the goals scored by the home and away teams and determing who has more goals.
The scripts in here run a 10,000 iteration simulation for each game and store it to an SQL table in order to track the predictions and calculate accuracy and log loss metrics. It also graphs the distributions of the probabilites and then posts the results of the simulation and graph to Twitter.
If you have any questions please contact me at barloweanalytics@gmail.com or @matt_barlowe on Twitter.