Welcome! I intend this to be an ongoing project of predicting NFL game outcomes, point spreads, and final scores. My hope is that my models will improve over time as they become more sophisticated and use better data. I will try to regularly publish predictions to keep myself accountable, and will make observations along the way about what is working and what isn't. See below for latest updates. Enjoy!
_________________________________________________________________________________
Recap Week 7:
8.5 - 6.5 for the week (counting the Seahawks/Cardinals tie as 0.5 for both wins and losses). So the running total is 51.5 - 39.5 (57%). The ensemble (depending on how you treat the predicted ties), went 8 - 7. The probabilities I gave expected 9.46 wins, so another disappointing week. What happened?
- Dolphins beat the Bills
- All of my models agreed that the Bills should have won, but they didn't.
- Seahawks tied the Cardinals
- My models were conflicted but sided with the Seahawks barely. I think the Seahawks were fortunate to get a tie out of this one.
- Chargers beat the Falcons
- Another overtime game. My models all agreed on a Falcon's win, but they didn't.
- Patriots beat the Steelers
- I suspected they would but my model disagreed, so this was expected.
- Eagles beat the Vikings
- Again, all of my models agreed on a Vikings win, but they lost.
Meh. Another week of surprises, but I guess by this point that shouldn't be surprising.
How are other sites doing in their predictions? A lot of sites don't list their long-run success. I couldn't find an article for bing.com, for example. So that makes it hard to compare and keep track of how other sites are doing and whose picks are trustworthy. But here are some sites I found:
- fivethirtyeight.com
- 62 - 45 (58%)
- John Beech, CBS
- 59-47-1 (56%)
- Gil Alcarez, NFL Spin Zone
- 59-47-1 (56%)
- Seth Dunlap
- 40-42-2 (48%)
- Sports Insights
- 28 - 25 (53%)
I will take fivethirtyeight.com to be my standard because they are the easiest to compare to and are definitely trustworthy. They are having a bit of a rough time picking too, so I am in good company.
_________________________________________________________________________________Week 7:
It's been a bit of a rough start. I'm still above .500 but not by much. Hopefully my small improvements will, by the end of the season, yield something more significant.
The main change this week was to add rolling win ratios for the past 16, 32, and 48 games for each team. The thought behind this is that I am wanting to look at each team's recent success, including past season success. A win ratio at the beginning of the season doesn't say much since half of the teams will have a winratio of 100% after the first game. So this should do a better job of capturing the "winningness" of a team in recent history.
I also added a running point total for each team, both since 1976 and for the past 15 years. This is partly inspired by the methodology at fivethirtyeight.com. Basically, each team starts at 0, and gets 1 point for a win and -1 points for a loss. A franchise with a great history of wins will have a high positive running total whereas as a team with a terrible history with many losses will have a very low negative running total.
Since 1976, the top three teams by best win vs. loss running totals are:
- Broncos: 145
- Steelers: 140
- Patriots: 132
The worst teams are:
32. Buccaneers: -149
31. Lions: -140
30. Browns: -108
While such values are interesting from a historical perspective, they don't mean much for today's teams and for predicting wins and losses, since all of that is in the past and likely does not reflect the current team's win potential. So I also calculated this value for the past 15 years, which is the span of my dataset used in predictions.
Since 2001 (15 years ago), the top three teams by best win vs. loss running totals are:
- Patriots: 141
- Steelers: 78
- Colts: 75
The worst teams are:
32. Lions: -86
31. Browns: -83
30. Raiders: -69
These appear to be much more relevant for predicting this seasons' game outcomes.
With these additions the outcome model goes to about 75% accuracy on past data. This is a great improvement if I have not overfitted the model. I will remain skeptical, especially given my model performance so far this season. But I am hopeful that I am on the right track again and am making some progress. We shall see.
Oddities and Bold Predictions:
- Rams vs. Giants
- For the first time, Rams are predicted to win!! My outcome model gives them a 53% chance of winning, but my other two models favor the Giants. Not by much though. So the ensemble predicts the Giants to win, but I'll stick with my outcome model.
- Patriots vs. Steelers
- My outcome model predicts the Patriots to lose while the other two models predict ties. With a healthy Brady and an injured Roethlisberger, I am definitely picking the Patriots to win this. So I expect my model to be wrong on this one.
Good luck in Week 7!
Here are the predictions for Week 7:
Week
|
Date
|
Team
|
Home Away
|
Opponent
|
Prob. Win
|
Pred. Team Win
|
Actual Team Win
|
Pred. Team PD
|
Actual Team PD
|
Pred. Team
Score
|
Pred. Opp.
Score
|
Actual Team
Score
|
Actual Opp.
Score
|
Ensemble
Prediction Win
|
7
|
10/20/2016
|
Bears
|
@
|
Packers
|
0.3
|
0
|
0
|
-3
|
-16
|
19
|
24
|
10
|
26
|
0
|
7
|
10/20/2016
|
Packers
|
Bears
|
0.7
|
1
|
1
|
3
|
16
|
24
|
19
|
26
|
10
|
1
| |
7
|
10/23/2016
|
49ers
|
Buccaneers
|
0.555
|
1
|
0
|
0
|
-17
|
19
|
20
|
17
|
34
|
0.5
| |
7
|
10/23/2016
|
Bengals
|
Browns
|
0.875
|
1
|
1
|
9
|
14
|
20
|
16
|
31
|
17
|
1
| |
7
|
10/23/2016
|
Bills
|
@
|
Dolphins
|
0.841
|
1
|
0
|
9
|
-3
|
28
|
22
|
25
|
28
|
1
|
7
|
10/23/2016
|
Browns
|
@
|
Bengals
|
0.125
|
0
|
0
|
-9
|
-14
|
16
|
20
|
17
|
31
|
0
|
7
|
10/23/2016
|
Buccaneers
|
@
|
49ers
|
0.445
|
0
|
1
|
0
|
17
|
20
|
19
|
34
|
17
|
0.5
|
7
|
10/23/2016
|
Cardinals
|
Seahawks
|
0.434
|
0
|
0.5
|
-3
|
0
|
25
|
25
|
6
|
6
|
0.166666667
| |
7
|
10/23/2016
|
Chargers
|
@
|
Falcons
|
0.335
|
0
|
1
|
-3
|
3
|
23
|
25
|
33
|
30
|
0
|
7
|
10/23/2016
|
Chiefs
|
Saints
|
0.674
|
1
|
1
|
7
|
6
|
26
|
26
|
27
|
21
|
0.833333333
| |
7
|
10/23/2016
|
Colts
|
@
|
Titans
|
0.804
|
1
|
1
|
7
|
8
|
26
|
20
|
34
|
26
|
1
|
7
|
10/23/2016
|
Dolphins
|
Bills
|
0.159
|
0
|
1
|
-9
|
3
|
22
|
28
|
28
|
25
|
0
| |
7
|
10/23/2016
|
Eagles
|
Vikings
|
0.397
|
0
|
1
|
-3
|
11
|
21
|
24
|
21
|
10
|
0
| |
7
|
10/23/2016
|
Falcons
|
Chargers
|
0.665
|
1
|
0
|
3
|
-3
|
25
|
23
|
30
|
33
|
1
| |
7
|
10/23/2016
|
Giants
|
@
|
Rams
|
0.469
|
0
|
1
|
1
|
7
|
24
|
21
|
17
|
10
|
0.666666667
|
7
|
10/23/2016
|
Jaguars
|
Raiders
|
0.483
|
0
|
0
|
-3
|
-17
|
24
|
26
|
16
|
33
|
0
| |
7
|
10/23/2016
|
Jets
|
Ravens
|
0.514
|
1
|
1
|
-2
|
8
|
21
|
23
|
24
|
16
|
0.333333333
| |
7
|
10/23/2016
|
Lions
|
Redskins
|
0.503
|
1
|
1
|
-1
|
3
|
24
|
26
|
20
|
17
|
0.333333333
| |
7
|
10/23/2016
|
Patriots
|
@
|
Steelers
|
0.402
|
0
|
1
|
0
|
11
|
23
|
23
|
27
|
16
|
0.333333333
|
7
|
10/23/2016
|
Raiders
|
@
|
Jaguars
|
0.517
|
1
|
1
|
3
|
17
|
26
|
24
|
33
|
16
|
1
|
7
|
10/23/2016
|
Rams
|
Giants
|
0.531
|
1
|
0
|
-1
|
-7
|
21
|
24
|
10
|
17
|
0.333333333
| |
7
|
10/23/2016
|
Ravens
|
@
|
Jets
|
0.486
|
0
|
0
|
2
|
-8
|
23
|
21
|
16
|
24
|
0.666666667
|
7
|
10/23/2016
|
Redskins
|
@
|
Lions
|
0.497
|
0
|
0
|
1
|
-3
|
26
|
24
|
17
|
20
|
0.666666667
|
7
|
10/23/2016
|
Saints
|
@
|
Chiefs
|
0.326
|
0
|
0
|
-7
|
-6
|
26
|
26
|
21
|
27
|
0.166666667
|
7
|
10/23/2016
|
Seahawks
|
@
|
Cardinals
|
0.566
|
1
|
0.5
|
3
|
0
|
25
|
25
|
6
|
6
|
0.833333333
|
7
|
10/23/2016
|
Steelers
|
Patriots
|
0.598
|
1
|
0
|
0
|
-11
|
23
|
23
|
16
|
27
|
0.666666667
| |
7
|
10/23/2016
|
Titans
|
Colts
|
0.196
|
0
|
0
|
-7
|
-8
|
20
|
26
|
26
|
34
|
0
| |
7
|
10/23/2016
|
Vikings
|
@
|
Eagles
|
0.603
|
1
|
0
|
3
|
-11
|
24
|
21
|
10
|
21
|
1
|
7
|
10/24/2016
|
Broncos
|
Texans
|
0.516
|
1
|
1
|
2
|
18
|
24
|
23
|
27
|
9
|
1
| |
7
|
10/24/2016
|
Texans
|
@
|
Broncos
|
0.484
|
0
|
0
|
-2
|
-18
|
23
|
24
|
9
|
27
|
0
|
No comments:
Post a Comment