I’m still playing around with the Glicko2 ratings system as a way to evaluate NFL teams. Right now, the model is pretty dumb1: it doesn’t take into account player injuries, movement, etc. I plan to incorporate that some day, but for now, here are the results for Week 1. Dots represent the rating and the bars are 95% confidence intervals:
At the top, this model probably overrates Tennessee and underrates Kansas City, but who knows? The NFL is weird.
- it essentially uses the last two seasons to inform 2019 Glicko2 ratings, allowing for significantly more volatility than in the traditional Glicko2 model, with some other minimal adjustments [return]