I never learned how to read.
Apparently, you can (and are encouraged) to supplement data sources, but you are discouraged from sharing private data outside of teams.
So I scraped ESPN for all regular season data for the past five years, which can be found in data/espn/.
Source the load.R script.
- Generate the offensive and defensive rating for each team in each season. (
calculate_ratings.R) - Generate a linear model (currently via least squares, in
calculate_coefficients.R) as defined by:points ~ offensive rating + defensive rating + location - Simulate each game by selecting 10,000 random draws from the distributions of the model coefficients, and compare pairwise points as games.