The NFL playoffs begin Saturday, and if recent reports are any indication, this season could mark the end of the 12-team postseason era. In March, the league will vote on whether to expand its tournament field to 14 teams for the 2015 season — after which the floodgates may open for further expansion in subsequent seasons.
This is not necessarily a bad thing. The NFL playoffs are a TV ratings bonanza, and it doesn’t seem as though our appetite for football is waning (despite a trying year off the field). Plus, teams such as the Philadelphia Eagles, Buffalo Bills and Houston Texans — all of which missed the playoffs this season despite winning records — make a case for creating more playoff spots. (And the 7-8-1 Carolina Panthers hosting a home playoff game makes the case for re-seeding the field.)
But which expanded format is best? As I’ve done in the past, I’ll take the perspective that the best format is the one that sees the most deserving team win the most often, using a Monte Carlo simulation to test playoff fields of various sizes.
More specifically, I generated random preseason Elo ratings for every team (based on the historical distribution of real-life preseason Elo ratings) and simulated the real-life 2014 regular season schedule 1,000 times — an exercise similar to that performed by Doug Drinen in this classic post at the Pro-Football-Reference.com blog. For each of those simulations, I tracked the regular season standings, seeding the teams within each conference using simulated point differential as the tiebreaker.
The preseason ratings represent the starting talent levels for every team, but they can go up or down depending on the simulated game outcomes, in accordance with the Elo formula. For instance, a team assigned an initial Elo below the league-average mark of 1500 could tear off an improbably great regular season and finish above 1700. That would be used as the team’s strength rating going into the postseason.
For potential expanded fields of 14, 16, 20, and 32 teams, I tracked how often the most talented preseason and end-of-regular-season Elo team won the Super Bowl, as well as the average preseason and pre-playoff Elo ratings (and rankings) of the simulated Super Bowl champs. For comparison, I also ran this test for contracted fields of 2, 4, and 8 teams, as well as the current 12-team setup.
Here’s a little more to help decipher that chart. When the playoffs contained just 2 teams, the average simulated Super Bowl winner had an Elo of 1565.3 before the season, which gave it an average ranking of 9.2 among the NFL’s 32 teams. Also, 12.8 percent of those Super Bowl winners were ranked No. 1 in the preseason. After the regular season was simulated, those teams averaged an Elo rating of 1690.1, with an average ranking of 2.5 within the league, and they led the league in post-regular season Elo 44.7 percent of the time. Finally, following the Super Bowl the average winner from our two-team-playoff universe had an Elo rating of 1711.6.
Comparing those categories across all formats, the irony is that a BCS-style two-team playoff produces the most talented champion from the perspective of both preseason and end-of-regular-season Elo ratings. But since that’s clearly neither realistic nor desirable, it appears the 14-team bracket is the superior option. On average, it yields the most talented team of any expanded format, and enables that team to win the Super Bowl with quite a bit more frequency — perhaps due to first-round byes only being given to the top seed in each conference.
Interestingly, a 14-team bracket also yields the best average post-Super Bowl rating for the champ among any format tested, expanded or not. (Granting that the current 12-team setup sees the better regular-season team win more often.) Based on this research, then, a 14-team playoff seems to strike the best balance between letting teams settle things on the field and putting the most deserving teams in a position to succeed.