Skip to main content
Menu
This NFL Season Is Super Weird

When we were prepping for this week’s episode of Hot Takedown, FiveThirtyEight’s sports podcast (subscribe today!), our producer Jody Avirgan pointed out an apparent oddity in the NFL standings: Despite a record number of undefeated teams through six weeks, only 10 clubs have winning records, with four more at .500 — meaning 18 of the league’s 32 teams have lost more games than they’ve won. Channeling our best Bernie Sanders: That’s an awfully top-heavy distribution, with a handful of zero- and one-loss teams gobbling up what seems like a disproportionate share of the victories.

One measure of the inequality of any distribution — wealth, wins, etc. — is the Gini coefficient, which runs from zero (perfectly equal) to one (perfectly unequal). We’ve applied it to sports a few times before; in this case, we looked at how unequally dispersed NFL wins have been in 2015 compared with other seasons1 through the first six weeks of the schedule. And the NFL is definitely seeing more inequality this year than the average season. Out of the past 36 full-length seasons, the 2015 standings have the 12th-highest Gini coefficient through six weeks.

But more striking than the strict inequality of the way wins have been distributed this season is simply the strangeness of the distribution in general. For every full-length season since 1978, we also calculated the frequency at which teams had each possible record (i.e., 6-0, 5-1, etc.), which can be used to describe the typical distribution of victories in an NFL season. Big deviations2 from that distribution, then, represent how abnormal a season has been in terms of divvying up wins and losses. And by that standard, 2015 has been very abnormal indeed, ranking as the fourth-strangest of the past 36 full-length seasons.

In other words, the number of teams with any given record this year is pretty far out of line with historical norms through six weeks. Some of what we found intuitively follows from the inequality hypothesis above — for instance, there are more undefeated teams than we’d expect (usually we’d expect about 1.5 undefeated squads through six weeks, not five) and, correspondingly, more one- and two-win teams than in the typical season. But some of the results are just, well, weird: There are fewer zero-win teams than we’d expect, and there are far fewer teams with four and five wins — neither of which squares neatly with the high-inequality argument.

So although this season has been marked by an unequal distribution of wins, it’s been even more defined by a flat-out strange distribution of wins. And the great irony is that, in terms of the distribution of talent in the league according to Elo, 2015 might be one of the most typical seasons in recent NFL history.


Hot Takedown: Are NFL wins weirdly bunched?


In addition to calculating how closely the season’s distribution of wins mimics historical norms, we applied the same computation to the distribution of team Elo ratings across the NFL, tracking how much deviation there was between the frequency of teams at any given rating3 and what we’d expect from the long-term historical average. (Elo is a good proxy for talent because it attempts to measure each team’s true strength at any given moment for the specific purpose of predicting the outcomes of future games.) And, amazingly, the distribution of Elo ratings this season is the fourth-closest match to the typical dispersion in any full-length NFL season since 1978!

All of this underscores how weird the NFL has been in the early going this season. Judging by the parity of talent in the league, this season hews particularly close to average. But the distribution of outcomes has been anything but typical: A bunch of teams are winning and losing more than they should, and there’s less of a football middle class than we’d expect. All we can say is, stay tuned to see if the wackiness continues.

Read more: 2015 NFL Predictions

Footnotes

  1. During the 16-game-schedule era, meaning we went back to 1978 and also excluded the 1982 and 1987 seasons, which were shortened by labor conflicts.

  2. Specifically, the sum of squared deviations from the expected number of teams in each record category.

  3. Splitting up teams into buckets by rounding their ratings to the nearest 50 points; i.e., we calculated what percentage of teams we’d expect to have Elo ratings of 1450, 1500, 1550, etc.

Andrew Flowers writes about economics and sports for FiveThirtyEight.

Neil Paine is a senior sportswriter for FiveThirtyEight.

Comments