UPDATE (Sept. 7, 2016; 12 p.m.): Our 2016 NFL Elo predictions follow the same methodology as our predictions from last year. Check out the article below for more details.
A good deal of FiveThirtyEight’s NFL coverage last season used Elo ratings, a simple system that estimates each team’s skill level using only the final scores and locations of each game. For 2015, we’re not only bringing Elo back (with a few small tweaks — more on those in a moment), but we’ve also built a continually updating Elo NFL predictions page that allows you to see the latest rankings, plus win probabilities and point spreads for the current week of NFL games.
How do our Elo ratings work? FiveThirtyEight editor-in-chief Nate Silver wrote a detailed FAQ about the formula before the 2014 season, and almost all of it still applies. The only changes we made mirror the methodology we used when applying Elo to the entire history of the NBA back in May and involve what to do when new (expansion) teams are added to the closed circuit of a league.
Originally, our Elo formula started each franchise (at its founding) with a rating of 1500, which also represented the rating of an average team. This worked in general, especially since it had been a long time since the league had expanded. But it’s not such a good assumption for handling expansion teams and analogous situations, such as mergers between different leagues. Eventually, we determined that new franchises should be given a rating of 1300,1 and in conjunction with this change, we also regress teams toward a mean of 1505 (instead of 1500) after every season.2 This helps balance against the low ratings assigned to expansion teams, though it does mean the average team no longer carries a 1500 Elo rating.3
Aside from those slight adjustments, Elo still works exactly the same way it did last season: Teams gain and lose ground based on the final score of each game and how unexpected the result was in the eyes of the pregame ratings. Under Elo, teams pick up where they left off: The initial team rankings for 2015 are by definition the same as last season’s end-of-year rankings,4 only more compressed because of the regression toward the mean.
Going into Week 1, that means the Seattle Seahawks and New England Patriots are once again the NFL’s highest-rated teams, albeit with lower Elo ratings than when they faced off last season in one of the strongest championship matchups in NFL history. Why? Like other well-designed predictive rating systems, including ESPN’s new Football Power Index, Elo is appropriately cautious early in the season; a team needs to prove itself to warrant a very high or very low rating.5 Combine that with the luck inherent in the NFL — the best teams don’t always win — and even Elo’s top-rated teams, the Seahawks and the Patriots, have just a 15 percent and 14 percent chance of winning the Super Bowl, respectively.
Just like last season, we’ll be writing a weekly column using Elo as a jumping-off point to discuss the week’s games. And in between, you can find ratings and predictions on our interactive page.
Here’s to another great NFL season!