NFL Rankings

January 16, 2017

Team Wins Losses Win Probability Rank Win Probability v. Generic Team Point Margin Rank Point Margin v. Generic Team Close Game Luck Close Game Luck Rank
New England 15.0 2.0 1 0.717 1 8.94 0 15
Dallas 13.0 4.0 2 0.648 3 5.63 1 10
Kansas City 12.0 5.0 3 0.614 4 4.94 1 10
Atlanta 12.0 5.0 7 0.584 2 8.39 -5 25
Pittsburgh 13.0 5.0 6 0.600 5 4.39 -1 18
Green Bay 12.0 6.0 4 0.612 7 3.21 3 8
Oakland 12.0 5.0 5 0.611 10 1.49 5 7
Denver 9.0 7.0 11 0.527 6 3.98 -5 25
Seattle 11.5 6.5 9 0.531 9 2.36 0 15
Philadelphia 7.0 9.0 16 0.491 8 3.15 -8 29
Washington 8.5 7.5 12 0.520 12 1.25 0 15
NY Giants 11.0 6.0 8 0.575 18 0.08 10 3
Baltimore 8.0 8.0 18 0.468 14 1.07 -4 24
Cincinnati 6.5 9.5 22 0.418 11 1.29 -11 32
Minnesota 8.0 8.0 18 0.468 17 0.09 -1 18
Tampa Bay 9.0 7.0 14 0.506 21 -0.63 7 5
Houston 10.0 8.0 10 0.530 25 -2.70 15 1
Arizona 7.5 8.5 23 0.413 13 1.15 -10 30
New Orleans 7.0 9.0 21 0.428 15 0.88 -6 28
Tennessee 9.0 7.0 13 0.512 23 -1.09 10 3
Indianapolis 8.0 8.0 18 0.468 19 -0.36 1 10
Miami 10.0 7.0 15 0.505 26 -2.84 11 2
Detroit 9.0 8.0 17 0.481 24 -1.36 7 5
San Diego 5.0 11.0 26 0.389 16 0.21 -10 30
Buffalo 7.0 9.0 25 0.398 20 -0.50 -5 25
Carolina 6.0 10.0 24 0.406 22 -0.65 -2 22
Jacksonville 3.0 13.0 28 0.315 27 -4.49 -1 18
NY Jets 5.0 11.0 27 0.347 29 -8.47 2 9
Chicago 3.0 13.0 29 0.303 28 -7.49 -1 18
Los Angeles 4.0 12.0 30 0.283 31 -11.23 1 10
Cleveland 1.0 15.0 32 0.199 30 -10.35 -2 22
San Francisco 2.0 14.0 31 0.205 32 -11.66 1 10

NFL Predictions

January 16, 2017

Home Visitor Prediction Line
Atlanta Green Bay Atlanta 60.5% Atlanta -7.5
New England Pittsburgh New England 75.7% New England -7

Description

The Jarratt NFL Rankings are a combination of two statistical models. Both models attempt to best explain each National Football League game result using several explanatory factors. One model explains wins and losses. The other model explains points on the court, specifically margin of victory and defeat.

The explanatory factors are:

  • home advantage
  • individual team strengths

Both models use the same data; only the representation of the outcome changes. To estimate home advantage, division advantage, and individual team strengths, the Jarratt Rankings use Markov Chain Monte Carlo sampling via Stan in R through the rethinking R package by Richard McElreath.

The posterior distribution that MCMC sampling returns is used to simulate 20,000 games per team against a generic team, half using each model described above. This generic team’s strength is set at the league average, and the uncertainty about its strength is set to the average of all teams’ uncertainty (i.e., the standard deviations of their posterior distributions).

The Close Game Luck is the difference between a team’s Win Probability Rank and its Point Margin Rank. If a team wins more often than their expected point spread indicates, then that team makes more efficient use of their points or is lucky in close games.

Last updated January 16, 2017.