2016 NFL Prediction
Accuracy
Microsoft's Cortana and Nate
Silver's
fivethirtyeight.com Elo are again
predicting
the outcomes of NFL games each week during the 2016 season. (See the
accuracy results for the two models for the 2015 NFL season here.)
The tables below use Brier scores to determine
the accuracy of the probabilistic predictions from both models.
Brier scores range from 0 to
1, with 0 meaning the probabilities in the models perfectly match the
outcomes of the games and 1 meaning no matches were made (so the closer
to 0.000, the the 95% confidence intervals
for both models for Wild Card Weekend include 0.250, making both
models no better or worse than assigning each team a 50% win
probability.better calibration of the model).
If a 50% (0.500) probability
were assigned to each team for each game, the Brier score would be
0.250.
For the 2016 season, the Elo model has a Brier score of 0.219 and the Cortana
model has a Brier score of 0.218. In
the 2015 season, the Elo model had a Brier score of 0.230 and the
Cortana model had a Brier score for 0.229. In the 2014 season, the Elo
model had a Brier score of 0.208 and the Cortana model had a Brier
score of 0.213. Both
the Elo model (win probability of 0.61) and the Cortana model (win
probability of 0.52) correctly picked New England in the Super Bowl.
|
Brier
Scores |
2016 NFL Season Summary |
538 Elo |
Cortana |
|
Week 1 |
0.210 |
0.230 |
Week 2 |
0.229 |
0.247 |
Week 3 |
0.213 |
0.247 |
Week 4 |
0.264 |
0.271 |
Week 5 |
0.260 |
0.235 |
Week 6 |
0.242 |
0.219 |
Week 7 |
0.228 |
0.239 |
Week 8 |
0.238 |
0.239 |
Week 9 |
0.233 |
0.216 |
Week 10 |
0.233 |
0.252 |
Week 11 |
0.175 |
0.163 |
Week 12 |
0.222 |
0.187 |
Week 13 |
0.177 |
0.217 |
Week 14 |
0.203 |
0.173 |
Week 15 |
0.215 |
0.195 |
Week 16 |
0.248 |
0.239 |
Week 17 |
0.174 |
0.187 |
Wild Card Weekend |
0.141 |
0.137 |
Divisional Round |
0.225 |
0.190 |
Championship Sunday |
0.121 |
0.113 |
Super Bowl Sunday | 0.152 | 0.230 |
|
Season totals |
0.219 |
0.218 |
The 95% confidence intervals
for the season totals for both models exclude 0.250, making both
models slightly better than assigning each team a 50% win
probability. The average win probabilities
among winning teams
are 56% for the Elo model and 56% for the Cortana model. For Championship Sunday, the
Elo model had a Brier score of 0.121 and the Cortana model had a Brier
score of 0.113.
2016 NFL - Championship Sunday |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Atlanta |
0.61 |
0.152 |
0.13 |
0.137 |
New England |
0.70 |
0.090 |
0.70 |
0.090 |
|
Brier scores for Championship Sunday |
0.121 |
|
0.113 |
The 95% confidence interval
for the Elo model for Championship Sunday includes
0.250, making it no better or worse than assigning each team a 50%
win
probability. The 95% confidence interval for the Cortana
model does not include 0.250. For the Divisional Round, the
Elo model had a Brier score of 0.225 and the Cortana model had a Brier
score of 0.190.
2016 NFL - Divisional Round |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Atlanta |
0.62 |
0.144 |
0.73 |
0.073 |
New England |
0.85 |
0.023 |
0.83 |
0.029 |
Pittsburgh |
0.36 |
0.410 |
0.37 |
0.397 |
Green Bay |
0.43 |
0.325 |
0.49 |
0.260 |
|
Brier scores for Divisional Round |
0.225 |
|
0.190 |
The 95% confidence intervals
for both models for the Divisional Round include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. For Wild Card Weekend, the
Elo model had a Brier score of 0.141 and the Cortana model had a Brier
score of 0.137.
2016 NFL - Wild Card Weekend |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Houston |
0.46 |
0.292 |
0.47 |
0.281 |
Seattle |
0.71 |
0.084 |
0.73 |
0.073 |
Pittsburgh |
0.73 |
0.073 |
0.69 |
0.096 |
Green Bay |
0.66 |
0.116 |
0.69 |
0.096 |
|
Brier scores for Wild Card Weekend |
0.141 |
|
0.137 |
While this was the best weekly performance for both models so far in the season, the 95% confidence intervals
for both models for Wild Card Weekend include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 17 of the season, the
Elo model had a Brier score of 0.174 and the Cortana model had a Brier
score of 0.187.
2016 NFL - Week 17 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Cincinnati |
0.58 |
0.176 |
0.63 |
0.137 |
Tennessee |
0.47 |
0.281 |
0.63 |
0.137 |
Tampa Bay |
0.53 |
0.221 |
0.57 |
0.185 |
Indianapolis |
0.81 |
0.036 |
0.70 |
0.090 |
New England |
0.65 |
0.123 |
0.55 |
0.203 |
Minnesota |
0.77 |
0.053 |
0.66 |
0.116 |
NY Jets |
0.42 |
0.336 |
0.29 |
0.504 |
Philadelphia |
0.34 |
0.436 |
0.57 |
0.185 |
Pittsburgh |
0.93 |
0.005 |
0.80 |
0.040 |
Atlanta |
0.76 |
0.058 |
0.80 |
0.040 |
NY Giants |
0.41 |
0.348 |
0.28 |
0.518 |
Arizona |
0.63 |
0.137 |
0.69 |
0.096 |
Denver |
0.53 |
0.221 |
0.48 |
0.270 |
Kansas City |
0.75 |
0.063 |
0.61 |
0.152 |
Seattle |
0.82 |
0.032 |
0.72 |
0.078 |
Green Bay |
0.49 |
0.260 |
0.51 |
0.240 |
|
Brier scores for week 17 |
0.174 |
|
0.187 |
The 95% confidence intervals
for both models for week 17 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 16 of the season, the
Elo model had a Brier score of 0.248 and the Cortana model had a Brier
score of 0.239.
2016 NFL - Week 16 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Philadelphia |
0.45 |
0.303 |
0.47 |
0.281 |
Miami |
0.41 |
0.348 |
0.36 |
0.410 |
New England |
0.90 |
0.010 |
0.82 |
0.032 |
Jacksonville |
0.35 |
0.423 |
0.33 |
0.449 |
Green Bay |
0.70 |
0.090 |
0.64 |
0.130 |
Cleveland |
0.31 |
0.476 |
0.37 |
0.397 |
Washington |
0.62 |
0.144 |
0.63 |
0.137 |
Atlanta |
0.44 |
0.314 |
0.58 |
0.176 |
Oakland |
0.67 |
0.109 |
0.63 |
0.137 |
New Orleans |
0.55 |
0.203 |
0.47 |
0.281 |
Arizona |
0.25 |
0.563 |
0.26 |
0.548 |
San Francisco |
0.26 |
0.548 |
0.42 |
0.336 |
Houston |
0.57 |
0.185 |
0.52 |
0.230 |
Pittsburgh |
0.73 |
0.073 |
0.67 |
0.109 |
Kansas City |
0.70 |
0.090 |
0.73 |
0.073 |
Dallas |
0.70 |
0.090 |
0.69 |
0.096 |
|
Brier scores for week 16 |
0.248 |
|
0.239 |
The 95% confidence intervals
for both models for week 16 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability.
In week 15 of the season, the
Elo model had a Brier score of 0.215 and the Cortana model had a Brier
score of 0.195.
2016 NFL - Week 15 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Seattle |
0.84 |
0.026 |
0.83 |
0.029 |
Miami |
0.51 |
0.240 |
0.58 |
0.176 |
Green Bay |
0.69 |
0.096 |
0.67 |
0.109 |
Houston |
0.83 |
0.029 |
0.67 |
0.109 |
Buffalo |
0.88 |
0.014 |
0.75 |
0.063 |
Baltimore |
0.66 |
0.116 |
0.69 |
0.096 |
Tennessee |
0.15 |
0.723 |
0.31 |
0.476 |
NY Giants |
0.53 |
0.221 |
0.57 |
0.185 |
Indianapolis |
0.32 |
0.462 |
0.40 |
0.360 |
Pittsburgh |
0.53 |
0.221 |
0.55 |
0.203 |
New Orleans |
0.30 |
0.490 |
0.43 |
0.325 |
Atlanta |
0.90 |
0.010 |
0.84 |
0.026 |
New England |
0.54 |
0.212 |
0.55 |
0.203 |
Oakland |
0.59 |
0.168 |
0.55 |
0.203 |
Dallas |
0.72 |
0.078 |
0.75 |
0.063 |
Carolina |
0.42 |
0.336 |
0.30 |
0.490 |
|
Brier scores for week 15 |
0.215 |
|
0.195 |
The 95% confidence intervals
for both models for week 15 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 14 of the season, the
Elo model had a Brier score of 0.203 and the Cortana model had a Brier
score of 0.173.
2016 NFL - Week 14 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Kansas City |
0.69 |
0.096 |
0.66 |
0.116 |
Pittsburgh |
0.47 |
0.281 |
0.52 |
0.230 |
Tennessee |
0.31 |
0.476 |
0.60 |
0.160 |
Washington |
0.46 |
0.292 |
0.53 |
0.221 |
Miami |
0.50 |
0.250 |
0.64 |
0.130 |
Carolina |
0.69 |
0.096 |
0.61 |
0.152 |
Cincinnati |
0.75 |
0.063 |
0.74 |
0.068 |
Detroit |
0.82 |
0.032 |
0.80 |
0.040 |
Houston |
0.36 |
0.410 |
0.39 |
0.372 |
Minnesota |
0.73 |
0.073 |
0.66 |
0.116 |
NY Jets |
0.57 |
0.185 |
0.48 |
0.270 |
Tampa Bay |
0.66 |
0.116 |
0.60 |
0.160 |
Atlanta |
0.63 |
0.137 |
0.64 |
0.130 |
Green Bay |
0.41 |
0.348 |
0.42 |
0.336 |
NY Giants |
0.41 |
0.348 |
0.53 |
0.221 |
New England |
0.79 |
0.044 |
0.80 |
0.040 |
|
Brier scores for week 14 |
0.203 |
|
0.173 |
The 95% confidence interval
for the Elo model for week 14 includes 0.250, making it no better or worse than assigning each team a 50% win
probability. The 95% confidence interval for the Cortana
model does not include 0.250, for the second time this season. In week 13 of the season, the
Elo model had a Brier score of 0.177 and the Cortana model had a Brier
score of 0.217.
2016 NFL - Week 13 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Dallas |
0.54 |
0.212 |
0.52 |
0.230 |
Kansas City |
0.52 |
0.230 |
0.47 |
0.281 |
Detroit |
0.48 |
0.270 |
0.30 |
0.490 |
New England |
0.86 |
0.020 |
0.81 |
0.036 |
Denver |
0.79 |
0.044 |
0.69 |
0.096 |
Green Bay |
0.63 |
0.137 |
0.53 |
0.221 |
Cincinnati |
0.56 |
0.194 |
0.52 |
0.230 |
Baltimore |
0.56 |
0.194 |
0.52 |
0.230 |
Chicago |
0.65 |
0.123 |
0.57 |
0.185 |
Oakland |
0.59 |
0.168 |
0.57 |
0.185 |
Pittsburgh |
0.64 |
0.130 |
0.63 |
0.137 |
Arizona |
0.58 |
0.176 |
0.53 |
0.221 |
Tampa Bay |
0.42 |
0.336 |
0.36 |
0.410 |
Seattle |
0.71 |
0.084 |
0.69 |
0.096 |
Indianapolis |
0.42 |
0.336 |
0.55 |
0.203 |
|
Brier scores for week 13 |
0.177 |
|
0.217 |
The 95% confidence interval
for the Elo model for week 13 does not include 0.250 for the first
time this season. The 95% confidence interval for the Cortana
model includes 0.250, making it no better or worse than assigning each team a 50% win
probability. In week 12 of the season, the
Elo model had a Brier score of 0.222 and the Cortana model had a Brier
score of 0.187.
2016 NFL - Week 12 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Detroit |
0.56 |
0.194 |
0.63 |
0.137 |
Dallas |
0.69 |
0.096 |
0.72 |
0.078 |
Pittsburgh |
0.47 |
0.281 |
0.55 |
0.203 |
Tennessee |
0.43 |
0.325 |
0.55 |
0.203 |
Buffalo |
0.86 |
0.020 |
0.78 |
0.048 |
Baltimore |
0.55 |
0.203 |
0.55 |
0.203 |
Atlanta |
0.59 |
0.168 |
0.64 |
0.130 |
NY Giants |
0.75 |
0.063 |
0.72 |
0.078 |
New Orleans |
0.62 |
0.144 |
0.67 |
0.109 |
Miami |
0.81 |
0.036 |
0.79 |
0.044 |
San Diego |
0.33 |
0.449 |
0.36 |
0.410 |
Tampa Bay |
0.29 |
0.504 |
0.36 |
0.410 |
Oakland |
0.58 |
0.176 |
0.64 |
0.130 |
New England |
0.69 |
0.096 |
0.69 |
0.096 |
Kansas City |
0.43 |
0.325 |
0.47 |
0.281 |
Green Bay |
0.31 |
0.476 |
0.34 |
0.436 |
|
Brier scores for week 12 |
0.222 |
|
0.187 |
The 95% confidence intervals
for both models for week 12 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 11 of the season, the
Elo model had a Brier score of 0.175 and the Cortana model had a Brier
score of 0.163.
2016 NFL - Week 11 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Carolina |
0.69 |
0.096 |
0.67 |
0.109 |
Pittsburgh |
0.74 |
0.068 |
0.75 |
0.063 |
Dallas |
0.75 |
0.063 |
0.77 |
0.053 |
Detroit |
0.83 |
0.029 |
0.75 |
0.063 |
Indianapolis |
0.69 |
0.096 |
0.57 |
0.185 |
Buffalo |
0.44 |
0.314 |
0.40 |
0.360 |
Tampa Bay |
0.16 |
0.706 |
0.26 |
0.548 |
NY Giants |
0.75 |
0.063 |
0.69 |
0.096 |
Minnesota |
0.54 |
0.212 |
0.57 |
0.185 |
Miami |
0.44 |
0.314 |
0.48 |
0.270 |
New England |
0.82 |
0.032 |
0.83 |
0.029 |
Seattle |
0.73 |
0.073 |
0.73 |
0.073 |
Washington |
0.64 |
0.130 |
0.64 |
0.130 |
Oakland |
0.50 |
0.250 |
0.66 |
0.116 |
|
Brier scores for week 11 |
0.175 |
|
0.163 |
The 95% confidence interval
for the Elo model for week 11 includes 0.250, making it no
better or worse than assigning each team a 50% win
probability. The 95% confidence interval for the Cortana model
does not include 0.250, which is a first for this model this season. In week 10 of the season, the
Elo model had a Brier score of 0.233 and the Cortana model had a Brier
score of 0.252.
2016 NFL - Week 10 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Baltimore |
0.81 |
0.036 |
0.67 |
0.109 |
Houston |
0.69 |
0.096 |
0.58 |
0.176 |
Denver |
0.59 |
0.168 |
0.39 |
0.372 |
Los Angeles |
0.36 |
0.410 |
0.47 |
0.281 |
Philadelphia |
0.54 |
0.212 |
0.63 |
0.137 |
Kansas City |
0.52 |
0.230 |
0.40 |
0.360 |
Tampa Bay |
0.59 |
0.168 |
0.53 |
0.221 |
Washington |
0.53 |
0.221 |
0.53 |
0.221 |
Tennessee |
0.35 |
0.423 |
0.47 |
0.281 |
Miami |
0.38 |
0.384 |
0.43 |
0.325 |
Arizona |
0.85 |
0.023 |
0.79 |
0.044 |
Dallas |
0.48 |
0.270 |
0.39 |
0.372 |
Seattle |
0.34 |
0.436 |
0.36 |
0.410 |
NY Giants |
0.57 |
0.185 |
0.53 |
0.221 |
|
Brier scores for week 10 |
0.233 |
|
0.252 |
The 95% confidence interval
for the Elo model for week 10 includes 0.250, making it no better or worse than assigning each team a 50% win
probability. The Cortana model performed worse than assigning each team a 50% win probability. In week 9 of the season, the
Elo model had a Brier score of 0.233 and the Cortana model had a Brier
score of 0.216.
2016 NFL - Week 9 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Atlanta |
0.56 |
0.176 |
0.55 |
0.203 |
Detroit |
0.29 |
0.504 |
0.28 |
0.518 |
NY Giants |
0.50 |
0.250 |
0.52 |
0.230 |
Miami |
0.53 |
0.221 |
0.69 |
0.096 |
Kansas City |
0.90 |
0.010 |
0.86 |
0.020 |
Dallas |
0.78 |
0.048 |
0.77 |
0.053 |
Baltimore |
0.43 |
0.325 |
0.52 |
0.230 |
New Orleans |
0.56 |
0.194 |
0.67 |
0.109 |
Carolina |
0.55 |
0.203 |
0.60 |
0.160 |
Indianapolis |
0.26 |
0.548 |
0.30 |
0.490 |
San Diego |
0.72 |
0.078 |
0.53 |
0.221 |
Oakland |
0.39 |
0.372 |
0.47 |
0.281 |
Seattle |
0.69 |
0.096 |
0.55 |
0.203 |
|
Brier scores for week 9 |
0.233 |
|
0.216 |
The 95% confidence intervals
for both models for week 9 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 8 of the season, the
Elo model had a Brier score of 0.238 and the Cortana model had a Brier
score of 0.239.
2016 NFL - Week 8 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Tennessee |
0.59 |
0.168 |
0.61 |
0.152 |
Washington/Cincinnati* |
0.52 |
0.250 |
0.66 |
0.276 |
New England |
0.52 |
0.230 |
0.53 |
0.221 |
Carolina |
0.51 |
0.240 |
0.52 |
0.230 |
NY Jets |
0.65 |
0.123 |
0.57 |
0.185 |
Houston |
0.59 |
0.168 |
0.52 |
0.230 |
Kansas City |
0.62 |
0.144 |
0.42 |
0.336 |
New Orleans |
0.30 |
0.490 |
0.47 |
0.281 |
Oakland |
0.45 |
0.303 |
0.48 |
0.270 |
Denver |
0.78 |
0.048 |
0.72 |
0.078 |
Atlanta |
0.56 |
0.194 |
0.55 |
0.203 |
Dallas |
0.59 |
0.168 |
0.60 |
0.160 |
Chicago |
0.25 |
0.563 |
0.30 |
0.490 |
|
Brier scores for week 8 |
0.238 |
|
0.239 |
*The Brier scores for the Washington/Cincinnati
tie were calculated by giving one-half of the Brier scores for the win for each team (the
Elo model had Cincinnati with a 52% win probability and the Cortana model had Cincinnati with
a 66% win probability). The 95% confidence intervals
for both models for week 8 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 7 of the season, the
Elo model had a Brier score of 0.228 and the Cortana model had a Brier
score of 0.239.
2016 NFL - Week 7 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Green Bay |
0.81 |
0.036 |
0.75 |
0.063 |
NY Giants |
0.48 |
0.270 |
0.39 |
0.372 |
Kansas City |
0.80 |
0.040 |
0.77 |
0.053 |
Indianapolis |
0.52 |
0.230 |
0.43 |
0.325 |
Philadelphia |
0.43 |
0.325 |
0.48 |
0.270 |
Cincinnati |
0.82 |
0.032 |
0.77 |
0.053 |
Detroit |
0.56 |
0.194 |
0.58 |
0.176 |
Oakland |
0.51 |
0.240 |
0.55 |
0.203 |
Miami |
0.36 |
0.410 |
0.34 |
0.436 |
NY Jets |
0.59 |
0.168 |
0.45 |
0.303 |
Tampa Bay |
0.45 |
0.303 |
0.47 |
0.281 |
San Diego |
0.26 |
0.548 |
0.26 |
0.548 |
New England |
0.46 |
0.292 |
0.58 |
0.176 |
Seattle/Arizona* |
0.51 |
0.250 |
0.55 |
0.253 |
Denver |
0.71 |
0.084 |
0.74 |
0.068 |
|
Brier scores for week 7 |
0.228 |
|
0.239 |
*The Brier scores for the Seattle/Arizona
tie were calculated by giving one-half of the Brier scores for the win for each team (the
Elo model had Arizona with a 51% win probability and the Cortana model had Seattle with
a 55% win probability). The 95% confidence intervals
for both models for week 7 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 6 of the season, the
Elo model had a Brier score of 0.242 and the Cortana model had a Brier
score of 0.219.
2016 NFL - Week 6 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
San Diego |
0.28 |
0.518 |
0.47 |
0.281 |
Buffalo |
0.81 |
0.036 |
0.83 |
0.029 |
Washington |
0.50 |
0.250 |
0.60 |
0.160 |
Tennessee |
0.65 |
0.123 |
0.72 |
0.078 |
NY Giants |
0.55 |
0.203 |
0.37 |
0.397 |
New Orleans |
0.42 |
0.336 |
0.45 |
0.303 |
Jacksonville |
0.38 |
0.384 |
0.48 |
0.270 |
Detroit |
0.59 |
0.168 |
0.67 |
0.109 |
Miami |
0.25 |
0.563 |
0.37 |
0.397 |
New England |
0.70 |
0.090 |
0.79 |
0.044 |
Kansas City |
0.57 |
0.185 |
0.42 |
0.336 |
Seattle |
0.68 |
0.102 |
0.66 |
0.116 |
Dallas |
0.31 |
0.476 |
0.31 |
0.476 |
Houston |
0.67 |
0.109 |
0.55 |
0.203 |
Arizona |
0.71 |
0.084 |
0.70 |
0.090 |
|
Brier scores for week 6 |
0.242 |
|
0.219 |
The 95% confidence intervals
for both models for week 6 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 5 of the season, the
Elo model had a Brier score of 0.260 and the Cortana model had a Brier
score of 0.235.
2016 NFL - Week 5 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Arizona |
0.60 |
0.160 |
0.60 |
0.160 |
New England |
0.75 |
0.063 |
0.78 |
0.048 |
Detroit |
0.46 |
0.292 |
0.45 |
0.303 |
Indianapolis |
0.66 |
0.116 |
0.61 |
0.152 |
Tennessee |
0.28 |
0.518 |
0.34 |
0.436 |
Washington |
0.36 |
0.410 |
0.42 |
0.336 |
Minnesota |
0.73 |
0.073 |
0.64 |
0.130 |
Pittsburgh |
0.75 |
0.063 |
0.79 |
0.044 |
Atlanta |
0.23 |
0.593 |
0.33 |
0.449 |
Dallas |
0.51 |
0.240 |
0.52 |
0.230 |
Buffalo |
0.46 |
0.292 |
0.48 |
0.270 |
Oakland |
0.64 |
0.130 |
0.74 |
0.068 |
Green Bay |
0.76 |
0.058 |
0.77 |
0.053 |
Tampa Bay |
0.20 |
0.640 |
0.22 |
0.608 |
|
Brier scores for week 5 |
0.260 |
|
0.235 |
The Elo model performed worse than assigning each team a 50% win
probability and the 95% confidence interval
for the Cortana model includes 0.250, making it no better or worse than assigning each team a 50% win
probability. In week 4 of the season, the
Elo model had a Brier score of 0.264 and the Cortana model had a Brier
score of 0.271.
2016 NFL - Week 4 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Cincinnati |
0.73 |
0.073 |
0.69 |
0.096 |
Jacksonville |
0.34 |
0.436 |
0.37 |
0.397 |
Houston |
0.82 |
0.032 |
0.67 |
0.109 |
Washington |
0.73 |
0.073 |
0.70 |
0.090 |
Seattle |
0.55 |
0.203 |
0.48 |
0.270 |
Buffalo |
0.24 |
0.578 |
0.26 |
0.548 |
Atlanta |
0.46 |
0.292 |
0.45 |
0.303 |
Oakland |
0.33 |
0.449 |
0.47 |
0.281 |
Chicago |
0.43 |
0.325 |
0.39 |
0.372 |
Denver |
0.76 |
0.058 |
0.61 |
0.152 |
Los Angeles |
0.30 |
0.490 |
0.27 |
0.533 |
New Orleans |
0.37 |
0.397 |
0.36 |
0.410 |
Dallas |
0.45 |
0.303 |
0.52 |
0.230 |
Pittsburgh |
0.55 |
0.203 |
0.61 |
0.152 |
Minnesota |
0.78 |
0.048 |
0.64 |
0.130 |
|
Brier scores for week 4 |
0.264 |
|
0.271 |
Both models performed worse than assigning each team a 50% win
probability.
In week 3 of the season, the
Elo model had a Brier score of 0.213 and the Cortana model had a Brier
score of 0.247.
2016 NFL - Week 3 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
New England |
0.70 |
0.090 |
0.64 |
0.130 |
Buffalo |
0.40 |
0.360 |
0.43 |
0.325 |
Oakland |
0.53 |
0.221 |
0.47 |
0.281 |
Washington |
0.35 |
0.423 |
0.36 |
0.410 |
Miami |
0.68 |
0.102 |
0.74 |
0.068 |
Baltimore |
0.62 |
0.144 |
0.63 |
0.137 |
Green Bay |
0.69 |
0.096 |
0.64 |
0.130 |
Denver |
0.52 |
0.230 |
0.39 |
0.372 |
Minnesota |
0.34 |
0.436 |
0.26 |
0.548 |
Los Angeles |
0.47 |
0.281 |
0.48 |
0.270 |
Seattle |
0.78 |
0.048 |
0.67 |
0.109 |
Kansas City |
0.68 |
0.102 |
0.58 |
0.176 |
Indianapolis |
0.60 |
0.160 |
0.57 |
0.185 |
Philadelphia |
0.44 |
0.314 |
0.47 |
0.281 |
Dallas |
0.67 |
0.109 |
0.55 |
0.203 |
Atlanta |
0.46 |
0.292 |
0.42 |
0.336 |
|
Brier scores for week 3 |
0.213 |
|
0.247 |
The 95% confidence intervals
for both models for week 3 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 2 of the season, the
Elo model had a Brier score of 0.229 and the Cortana model had a Brier
score of 0.247.
2016 NFL - Week 2 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
NY Jets |
0.43 |
0.325 |
0.43 |
0.325 |
Carolina |
0.78 |
0.048 |
0.74 |
0.068 |
Dallas |
0.37 |
0.397 |
0.31 |
0.476 |
Pittsburgh |
0.64 |
0.130 |
0.60 |
0.160 |
NY Giants |
0.62 |
0.144 |
0.63 |
0.137 |
New England |
0.81 |
0.036 |
0.78 |
0.048 |
Houston |
0.44 |
0.314 |
0.47 |
0.281 |
Tennessee |
0.19 |
0.656 |
0.20 |
0.640 |
Baltimore |
0.57 |
0.185 |
0.55 |
0.203 |
Los Angeles |
0.33 |
0.449 |
0.42 |
0.336 |
Arizona |
0.78 |
0.048 |
0.73 |
0.073 |
San Diego |
0.66 |
0.116 |
0.67 |
0.109 |
Atlanta |
0.40 |
0.360 |
0.21 |
0.624 |
Denver |
0.80 |
0.040 |
0.81 |
0.036 |
Minnesota |
0.58 |
0.176 |
0.60 |
0.160 |
Philadelphia |
0.51 |
0.240 |
0.48 |
0.270 |
|
Brier scores for week 2 |
0.229 |
|
0.247 |
The 95% confidence intervals
for both models for week 2 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability. In week 1 of the season, the
Elo model had a Brier score of 0.206 and the Cortana model had a Brier
score of 0.229.
2016 NFL - Week
1 |
Winner |
538
Elo Win |
Brier
Score |
Cortana
Win |
Brier
Score |
|
Denver |
0.60 |
0.160 |
0.57 |
0.185 |
Green Bay |
0.68 |
0.102 |
0.55 |
0.203 |
Kansas
City |
0.81 |
0.036 |
0.74 |
0.068 |
Oakland |
0.40 |
0.360 |
0.42 |
0.336 |
Cincinnati |
0.49 |
0.260 |
0.47 |
0.281 |
Philadelphia |
0.71 |
0.084 |
0.67 |
0.109 |
Minnesota |
0.71 |
0.084 |
0.60 |
0.160 |
Houston |
0.67 |
0.109 |
0.58 |
0.176 |
Baltimore |
0.53 |
0.221 |
0.47 |
0.281 |
Tampa Bay |
0.31 |
0.476 |
0.42 |
0.336 |
Seattle |
0.82 |
0.032 |
0.78 |
0.048 |
NY Giants |
0.44 |
0.314 |
0.47 |
0.281 |
Detroit |
0.43 |
0.325 |
0.43 |
0.325 |
New England |
0.40 |
0.360 |
0.33 |
0.449 |
Pittsburgh |
0.54 |
0.212 |
0.53 |
0.221 |
San Francisco |
0.53 |
0.221 |
0.53 |
0.221 |
|
Brier scores for week 1 |
0.210 |
|
0.230 |
The 95% confidence intervals
for both models for week 1 include 0.250, making both
models no better or worse than assigning each team a 50% win
probability.
|