I really enjoy watching college softball during the spring, but one of my main frustrations is there are not many predictive ratings systems out there to gauge team strength (There is Massey’s Ratings but it contains a backwards looking component and is less interpretable). To try to help with that, I’ve created a ratings system for D1 softball. The goal is to turn the full season’s game results (games scores only, it’s all I can get access to easily) into opponent-adjusted offensive, defensive, and overall team ratings.
The model uses ridge regression, with each game contributing information about how a team’s run scoring and run prevention should be evaluated after accounting for opponent strength. More recent games receive more weight, so the ratings can move as teams improve, injuries matter, or schedules get tougher. I backtested a few different parameters and weighting schemes and am fairly happy with the results.
In the future, I anticipate adding preseason priors based on past season’s data to try to improve the ratings early in the year. I would also like to add pitching adjustments, but there is not a good data source for that I’m aware of.
The site updates daily from the latest model output during the season. You can view it here: Softball Ratings.