MLSEloForecast
MLSEloForecast is a group of R scripts used to analyze and forecast Major League Soccer matches.
Installation
Simply download this repo into a location most convenient for you. It is also a good idea to install the SoccerEloForecast package, as some scripts will require it.
Scripts
MLSElo.R: Script that produces Elo ratings for all teams using MLSMatchesRefined.csv, which are adapted off the methodology used in the World Football Elo Ratings. The starting/mean Elo rating is 1500, and team ratings regress to it at the end of season (3/4 old Elo + 1/4 mean Elo).
CalibrationTest.R: Tests the zero-sum calibration of the produced Elo ratings. Outputs csv and can produce 45 degree line calibration plot. Note: uses environment variables, so it is best run after MLSElo.R.
TestingGames.R: Requires SoccerEloForecast. Produces the chances of a match result between to teams or simulates a user defined number of them.
ScoreMatrix.R: Produces likely score matrix between two teams. Note: uses environment variables, so it is best run after TestingGames.R, which should be set to return a simulated match scores (15000 recommended).
Data
MLS Names.csv: A csv file that contains the full names and abbreviations used for the scripts. Also contains the 3 main colors for all active teams.
MLSMatchesRefined.csv: Csv file adapted from the Major League Soccer Dataset, which was created by Joseph Mohr. It has been adapted by removing unplayed games, sorting chronologically, and abbreviating the type of match played.
MLS Names.csv: Used in MLSElo.R to format the team Elo ratings.
MLS Elo Histories: Folder of csv files containing the match-by-match Elo rating of all active and defunct MLS teams. These can be changed by augmenting MLSElo.R.
Future Updates
Full season + Playoff forecast script coming soon.