Skip to main content
Menu
40 Percent Of The NFL Is Real

It’s Week 8 of the 2015 season, which makes it a weird time on the NFL calendar. Almost half the schedule has been played, so it feels like we should have a decent grasp on which teams are good and bad. But this is the NFL, where we don’t know anything and we never will.

For instance, the New York Jets, who were not supposed to be very good according to our in-house rating system Elo, have played like one of the NFL’s best teams1 so far. And the Indianapolis Colts, who were supposed to be one of the league’s best teams, have played like one of its worst.

ELO RATING
TEAM PRESEASON HAS PLAYED LIKE … DIFF
New York Jets 1440 1734 +294
New England Patriots 1664 1907 +243
Arizona Cardinals 1541 1777 +236
Cincinnati Bengals 1540 1739 +199
Atlanta Falcons 1462 1612 +150
Pittsburgh Steelers 1563 1662 +99
Carolina Panthers 1536 1634 +99
Philadelphia Eagles 1545 1642 +97
Green Bay Packers 1601 1692 +91
Washington Redskins 1371 1459 +88
St. Louis Rams 1477 1552 +75
Oakland Raiders 1389 1449 +61
New York Giants 1480 1527 +47
Minnesota Vikings 1479 1492 +13
Denver Broncos 1595 1597 +1
Tennessee Titans 1339 1337 -3
Buffalo Bills 1534 1522 -12
New Orleans Saints 1486 1457 -30
Cleveland Browns 1422 1389 -33
Miami Dolphins 1484 1439 -44
Seattle Seahawks 1678 1604 -73
Kansas City Chiefs 1540 1464 -75
Jacksonville Jaguars 1357 1269 -87
Baltimore Ravens 1577 1484 -93
San Diego Chargers 1523 1409 -113
Chicago Bears 1419 1302 -117
Tampa Bay Buccaneers 1377 1244 -132
Dallas Cowboys 1590 1457 -133
Indianapolis Colts 1581 1399 -182
San Francisco 49ers 1529 1324 -204
Houston Texans 1505 1282 -223
Detroit Lions 1538 1302 -237

(For those new to FiveThirtyEight, Elo is a system for ranking NFL teams that assigns them strength ratings based on how likely they are to win in the future. Each team’s rating is calculated using the scores, locations and opponents from its games, plus a prior based on how good Elo thought the team was the last time its rating was calculated.)

We know some of what we’ve seen thus far is real. But some is an illusion, the product of injuries, personnel changes and the random variance that can occur in the small sample of a half-season. So what’s a football fan to think? How much stock should be placed in what we think we’ve learned over the past two months, compared with what we thought we knew before any games were played?

We can get a sense for how much we’ve really learned by looking to past seasons2 and searching for the mixture of preseason Elo and season-to-date Simple Rating System3 that does the best job of predicting games from Week 8 on. Since 1978, 40 percent of what we think we’ve learned through seven weeks is real. That is, after the first seven weeks of the NFL season, roughly 40 percent of a team’s deviation from its initial outlook should be preserved going forward. The rest tends to fall away in due time, as teams see corrections in health, coaches make adjustments (and counter-adjustments), and teams see aberrant feats regress their way to the mean.

That fraction should be enough to change our opinions about the extreme outliers: The Colts are probably closer to average than to a top-five team,4 while the Jets are likely an above-average squad on the verge of being in the league’s top third. But at the same time, it can’t really help us with the nuances of teams that look like they belong in a tier adjacent to where they started. For instance, the Cincinnati Bengals and Carolina Panthers are both 6-0 despite being in the league’s middle third before the season. We can think of them as top 10 teams, but it’s difficult to know where in the top 10 they fall.

Despite our proximity to the season’s halfway marker, there’s still a lot left to be learned about the NFL this year. And that means plenty more surprising outcomes, both real and illusory.

Footnotes

  1. According to Pro-Football-Reference.com’s Simple Rating System (SRS). I took each team’s SRS, multiplied it by 25 (as commanded to by Nate Silver), and added it to 1505, which is the average team’s Elo in the NFL.

  2. I used data from NFL seasons since 1978 that featured a 16-game schedule (in other words, excluding the strike-shortened 1982 and 1987 seasons).

  3. Converted to Elo’s scale.

  4. Yes, they were missing Andrew Luck for two starts. But they also went 2-0 under Matt Hasselbeck in those starts and are 1-4 under Luck.

Neil Paine is a senior sportswriter for FiveThirtyEight.

Comments