Research paper + PPT

NFL Power Rankings Through Bradley-Terry Modeling TO N Y W U! STATS 50

NOTE: I don’t have express written consent from the NFL to use their logos so blame them if this presentation is a little bare

Bradley-Terry Modeling  

The Data! ARI ATL BAL BUF CAR CHI CIN CLE DAL DEN DET GB HOU IND JAC KC MIA MIN NE NO NYG NYJ OAK PHI PIT SD SF SEA STL TB TEN WAS

ARI 11 8 3 11 11 4 1 9 17 10 ATL 11 2 3 52 BAL 22 28 2 15 7 20 31 14 BUF 3 16 3 19 1 55 CAR 7 0 17 31 6 CHI 14 8 8 8 8 CIN 14 10 0 9 10 17 1 26 CLE 2 21 2 10 21 5 1 DAL 13 3 14 21 13 7 3 16 DEN 21 7 7 20 3 14 24 14 25 DET 1 17 12 4 14 1 21 7 17 GB 6 21 62 3 35 5 7 33 HOU 6 16 14 16 38 11 IND 7 27 1 5 47 16 24 22 JAC 18 1 KC 4 19 27 14 3 4 27 MIA 13 3 14 13 3 24 37 MIN 13 18 6 28 4 3 NE 15 28 26 22 25 22 23 2 7 9 NO 18 21 11 3 6 NYG 10 13 29 31 NYJ 5 7 OAK 4 11 PHI 24 21 23 10 3 17 27 6 19 3 PIT 20 18 3 7 17 8 3 SD 1 12 19 31 10 9 3 SF 11 5 3 6 5 14 4 SEA 16 4 6 20 21 6 10 16 10 STL 15 52 3 2 2 24 TB 3 20 TEN 2 16 WAS 3 31 2

Win/Loss Rankings

0

0.035

0.07

0.105

0.14

0 10 20 30 40

0.1243

0.10510.1032

0.076 0.0721

0.0571

0.04370.0428 0.03890.03790.0358 0.0303 0.03 0.02740.02650.0223 0.0212

0.0120.01190.01180.01150.0112 0.00890.00740.0072 0.00630.00450.004 0.0030.0021 0.00190.0019

DEN 10 0.124346 NE 19 0.105084 ARI 1 0.103176 GB 12 0.075982 SEA 28 0.072122 PHI 24 0.057056 SD 26 0.043723 DET 11 0.042774 KC 16 0.038942 IND 14 0.037868 DAL 9 0.035815 CIN 7 0.03025 MIA 17 0.029967 SF 27 0.027368 BUF 4 0.026498 BAL 3 0.022254 STL 29 0.02118 PIT 25 0.012017 MIN 18 0.011938 HOU 13 0.011835 CHI 6 0.011467 CLE 8 0.011197 NO 20 0.008889 ATL 2 0.007361 CAR 5 0.00722 NYG 21 0.006302 OAK 23 0.004487 NYJ 22 0.004024 WAS 32 0.002952 JAC 15 0.002079 TEN 31 0.001937 TB 30 0.00189

Point Differential Rankings

0

0.035

0.07

0.105

0.14

0 10 20 30 40

0.132 0.1239

0.0976

0.076 0.0652

0.0601

0.0433 0.040.0381 0.0364 0.03490.0348 0.034 0.0277

0.0221 0.01710.0163 0.0144 0.01310.0125 0.0108 0.00970.008 0.00730.00710.0068 0.00250.0024 0.00160.00160.0015 0.0014

NE 19 0.131954 DEN 10 0.12393 SEA 28 0.097586 KC 16 0.075993 GB 12 0.06518 ARI 1 0.060062 MIA 17 0.043296 PHI 24 0.039981 BAL 3 0.038111 IND 14 0.036377 SD 26 0.034861 BUF 4 0.034798 DET 11 0.034041 PIT 25 0.02769 DAL 9 0.022065 SF 27 0.017103 CIN 7 0.01629 MIN 18 0.014445 STL 29 0.013053 CLE 8 0.012489 HOU 13 0.010834 NO 20 0.009734 CAR 5 0.008045 NYG 21 0.007271 CHI 6 0.00714 ATL 2 0.006801 NYJ 22 0.002469 OAK 23 0.002395 WAS 32 0.001621 TB 30 0.00156 JAC 15 0.001465 TEN 31 0.001358

Model Comparison

0

0.035

0.07

0.105

0.14

0 10 20 30 40

0.132 0.1239

0.0976

0.076 0.0652

0.0601

0.0433 0.040.0381 0.0364 0.03490.0348 0.034 0.0277

0.0221 0.01710.0163 0.0144 0.01310.0125 0.0108 0.00970.008 0.00730.00710.0068 0.00250.0024 0.00160.00160.0015 0.0014

Power Rankings Comparison The BT models are generally consistent with news analyst rankings ◦ Does well when comparing the “echelons” of teams

◦ Elite Teams ◦ Good Teams ◦ Fringe Playoff Teams ◦ Not Good Teams ◦ Bottom Feeder Teams

BT W/L BT PtDiff ESPN NFL.com DEN NE GB GB NE DEN NE NE ARI SEA DEN DEN GB KC SEA SEA SEA GB ARI IND PHI ARI IND ARI SD MIA PHI PHI DET PHI DAL DAL KC BAL DET DET IND IND PIT PIT DAL SD SD KC CIN BUF BAL CIN MIA DET CIN SD SF PIT KC BAL BUF DAL MIA MIA BAL SF SF HOU STL CIN BUF BUF PIT MIN HOU STL MIN STL CLE SF HOU CLE STL CLE CHI HOU ATL MIN CLE NO MIN ATL NO CAR NO CAR ATL NYG CAR NO CAR CHI CHI NYG NYG ATL NYG CHI OAK NYJ WAS TB NYJ OAK NYJ WAS WAS WAS OAK OAK JAC TB TB JAC TEN JAC JAC NYJ TB TEN TEN TEN

Playoff Predictions

NFC Wk 15 Wk 16 Wk 17 1 ARI STL SEA SF

gamma 0.1032441 0.021184 0.072156 0.027371 E(wins) prob 0.829751 0.588619 0.790446 2.208816 12.21

2 GB BUF TB DET 0.0759796 0.026495 0.001888 0.042766

0.741449 0.975757 0.639851 2.357057 12.36 3 PHI DAL WAS NYG

0.0570369 0.035806 0.002949 0.006298 0.614338 0.950835 0.900565 2.465737 11.47

4 ATL PIT NO CAR 0.0073539 0.012 0.00888 0.007214

0.379976 0.452987 0.504814 1.337778 6.34 5 SEA SF ARI STL

0.0721564 0.027371 0.103244 0.021184 0.724991 0.411381 0.773048 1.90942 10.91

6 DET MIN CHI GB 0.0427662 0.011932 0.011463 0.07598

0.781854 0.788622 0.360149 1.930625 10.93 7 DAL PHI IND WAS

0.035806 0.057037 0.037821 0.002949 0.385662 0.486317 0.923901 1.795881 10.8

8 SF SEA SD ARI 0.0273709 0.072156 0.043738 0.103244

0.275009 0.384915 0.209554 0.869478 7.87 9 NO CHI ATL TB

0.0088804 0.011463 0.007354 0.001888 0.436528 0.547013 0.824691 1.808231 6.81

10 CAR TB CLE ATL 0.0072137 0.001888 0.011181 0.007354

0.792587 0.392153 0.495186 1.679927 5.68

W/L Pt. Diff E(wins) NFC E(wins)

12.36 GB 1 GB 12.29 12.21 ARI 2 ARI 11.98 11.47 PHI 3 PHI 11.45 6.81 NO 4 NO 7.02

10.93 DET 5 SEA 11.35 10.91 SEA 6 DET 10.87

AFC 12.54 NE 1 DEN 12.65 12.51 DEN 2 NE 12.53 11.23 IND 3 IND 11.36 10.23 BAL 4 BAL 10.49 9.13 KC 5 PIT 9.7 8.64 CIN 6 KC 9.39

Predictions

Things Learned from the Models Strengths ◦ Very good at factoring in “Strength of Schedule” ◦ Doesn’t rely on “eye test” and other non-quantitative measurements

Weaknesses ◦ Equal weighting for early season results ◦ Doesn’t account for key factors

◦ Injuries ◦ On field performance vs. stat-sheet scores (i.e. garbage time TDs)

Extensions Home Field Advantage

Turnover Luck ◦ Recovering fumbles

A better point differential adjustment ◦ Get rid of garbage time scoring?