About the Ratings
The team ratings hosted here are based on methodologies that are best known for ranking college football teams here in the United States. The aim is to examine how many goals are scored and conceded by a team, while accounting for the strength of opposition. They are designed to be predictive—not tools to describe past games, but rather to accurately gauge a team’s chances of victory in an upcoming match. They are also, of course, objective, as goals are the only statistic considered.
An important feature of these ratings is that they are easy to interpret: a team with a rating of 1 would be expected to defeat an average opponent from their domestic league by one goal playing in a neutral venue. Similarly, a team with a rating of -1 we would expect to lose by a goal to an average rival from their league.
The ratings use time weighted goal data from every league game since the start of the 2012-13 season.
The match data was retrieved from the most helpful site: www.football-data.co.uk
About the Predictive Model
The team ratings are then used as inputs to a multinomial logistic regression model to assign a probability to each result (home win, away win, or draw). I have also applied the Kelly criterion to look for value against the odds posted on Bet365.
I’m a hobbyist with a background in Quantitative Analysis, a passionate fan of both American & European football, and a supporter of Arsenal FC and the New York Giants. You can follow me on Twitter here.