The archive
38 articles on rating systems, win probability, expected points and the methods behind them.
Why Coaching Changes Are Hard to Judge with Data
EPAFootball Rating Systems: Markov Chain and Massey Methods
RatingsHow to Ask for Help With Analytics Code, and Get It
NotesBuilding an NFL Analytics Department: Roles and Workflow
NotesK-Means Clustering for Fantasy Football Tiers
MethodsCounterintuitive Findings and Publication Bias in Sports
MethodsCross-Validation for Sports Prediction Models
MethodsSituational Thinking in Football Analytics: Where Data Helps
NotesHow to Get a Job in Sports Analytics: Real Paths In
NotesCrowdsourced Football Rankings: How Polls Aggregate Votes
RatingsElo Rating Formula for Football: K-Factor and Home Field
RatingsEigenvector Power Ratings for Football Teams
RatingsHow Consistent Are Football Rating Systems With Each Other?
RatingsWin Percentage Formula: Why It Falls Short as a Ranking
RatingsMassey Least Squares Ratings: The Simple Version
RatingsRanking Algorithms and the NFL: Offensive Rankings Explained
RatingsOutliers and Cook's Distance in Football Models
MethodsFeature Selection vs. Sample Selection in Football Models
MethodsHow to Grade Football Forecasts After the Season
EPAForecasting QB Performance With Multilevel Regression
EPAReproducible Research in Sports Analytics
MethodsWhat Social Media Data Reveals About Football Fandom
EPADo Down and Distance Matter to Win Probability?
Win probStatistical Decision-Making vs. Prediction in Football
MethodsWin Probability: Feature Engineering & Evaluation (Part 4)
Win probA Good Win Probability Model: Brier Score & Log Loss (Part 3)
Win probConfusion Matrix, Precision, and Recall for Football Models
MethodsLogistic Regression vs Random Forest: Win Probability (Part 2)
Win probProbability vs. Outcome in a Win Probability Model
Win probFootball Win Probability: Building the NFL Model (Part 1)
Win probWhere Football Analytics Is Discussed and Published
NotesWin Probability, Uncertainty and Overfitting in NFL Models
Win probThe archive lists every article in the order it was first published. The rating-system series starts with the ranking algorithms primer and runs through win percentage, power ratings, Massey least squares and Elo; the win probability series runs in four parts, from a first logistic model to feature engineering and evaluation. Methods articles stand on their own and are the place to start for cross-validation, outliers, confusion matrices and clustering. For the same material grouped by topic, the models page is the better index; for the numbers themselves, the ratings board and the calculators.
Frequently asked questions
What does the archive cover?
Every published article on the site: rating systems such as Elo, Massey and Markov chains, win probability models, expected points, the statistical methods behind them, and notes on how football analytics is practised. Weekly ratings reports are listed separately on the ratings page.
Why do some articles carry dates from 2013 to 2015?
Those articles were first published then. Each one has been rewritten since, with current data and sources, and shows the date of its latest review under the headline.
Where does the data in these articles come from?
Game results and play-by-play come from the open nflverse data releases under a CC-BY 4.0 license. Studies and methods are cited in a sources list at the end of every article.
Is there a way to follow new articles?
Yes. The Atom feed at /feeds/all.atom.xml lists every article with its latest review date, and the ratings page links each weekly report as it is published.