Previous predictions:
_________________________________________________________________________________
Week 13: Recap
My predictions went 10-6 (63%). Again, not great, but also not terrible. Lot's of exciting games and upsets, and many were close calls. Biggest upsets: Falcons and Patriots. Biggest surprise? Cowboys. Good luck in week 14!
_________________________________________________________________________________
Week 13: Recap
My predictions went 10-6 (63%). Again, not great, but also not terrible. Lot's of exciting games and upsets, and many were close calls. Biggest upsets: Falcons and Patriots. Biggest surprise? Cowboys. Good luck in week 14!
_________________________________________________________________________________
Week 13: Fourth Published Predictions (Win/Lose)
Below is the next round of predictions I was able to make. I continued to clean and automate my code, trying to gather new data and test its usefulness in my model. I also tested different kinds of algorithms for classification. I was using a logistic regression that was capping out at 68% accurate. I also tried a decision tree (65%) and random forests (66%). I then tried an SVM, and that got up to 81%! I have yet to do thorough testing on it so I am a bit skeptical as overfitting is a concern. Still, my predictions are consistent with the other models I used. I will continue to do more testing.
What is odd about the below predictions? The biggest oddity is the Cowboys/Redskins game. My model predicts a Cowboys win (67%), even though they are playing away, have a worse record, and are missing QB Tony Romo. Pro-football-reference.com gives the Redskins a 61% chance of winning. My model is probably off in this prediction, but I'd need to look into the details to figure out why that is. Other than that, the predictions seem reasonable.
Note: I included 3 decimal places in the probability to highlight the Seahawks/Vikings game. That is as close as it gets.
Note: I included 3 decimal places in the probability to highlight the Seahawks/Vikings game. That is as close as it gets.
Here are the predictions for week 13:
Week
|
Date
|
Team
|
Home Away
|
Opponent
|
Probability Win
|
Predicted Team Win
|
Actual Team Win
|
13
|
12/3/2015
|
Lions
|
Packers
|
0.416
|
0
|
0
| |
13
|
12/3/2015
|
Packers
|
@
|
Lions
|
0.584
|
1
|
1
|
13
|
12/6/2015
|
49ers
|
@
|
Bears
|
0.484
|
0
|
1
|
13
|
12/6/2015
|
Bears
|
49ers
|
0.516
|
1
|
0
| |
13
|
12/6/2015
|
Bengals
|
@
|
Browns
|
0.766
|
1
|
1
|
13
|
12/6/2015
|
Bills
|
Texans
|
0.549
|
1
|
1
| |
13
|
12/6/2015
|
Broncos
|
@
|
Chargers
|
0.639
|
1
|
1
|
13
|
12/6/2015
|
Browns
|
Bengals
|
0.234
|
0
|
0
| |
13
|
12/6/2015
|
Buccaneers
|
Falcons
|
0.394
|
0
|
1
| |
13
|
12/6/2015
|
Cardinals
|
@
|
Rams
|
0.781
|
1
|
1
|
13
|
12/6/2015
|
Chargers
|
Broncos
|
0.361
|
0
|
0
| |
13
|
12/6/2015
|
Chiefs
|
@
|
Raiders
|
0.618
|
1
|
1
|
13
|
12/6/2015
|
Colts
|
@
|
Steelers
|
0.311
|
0
|
0
|
13
|
12/6/2015
|
Dolphins
|
Ravens
|
0.421
|
0
|
1
| |
13
|
12/6/2015
|
Eagles
|
@
|
Patriots
|
0.371
|
0
|
1
|
13
|
12/6/2015
|
Falcons
|
@
|
Buccaneers
|
0.606
|
1
|
0
|
13
|
12/6/2015
|
Giants
|
Jets
|
0.539
|
1
|
0
| |
13
|
12/6/2015
|
Jaguars
|
@
|
Titans
|
0.432
|
0
|
0
|
13
|
12/6/2015
|
Jets
|
@
|
Giants
|
0.461
|
0
|
1
|
13
|
12/6/2015
|
Panthers
|
@
|
Saints
|
0.789
|
1
|
1
|
13
|
12/6/2015
|
Patriots
|
Eagles
|
0.629
|
1
|
0
| |
13
|
12/6/2015
|
Raiders
|
Chiefs
|
0.382
|
0
|
0
| |
13
|
12/6/2015
|
Rams
|
Cardinals
|
0.219
|
0
|
0
| |
13
|
12/6/2015
|
Ravens
|
@
|
Dolphins
|
0.579
|
1
|
0
|
13
|
12/6/2015
|
Saints
|
Panthers
|
0.211
|
0
|
0
| |
13
|
12/6/2015
|
Seahawks
|
@
|
Vikings
|
0.499
|
0
|
1
|
13
|
12/6/2015
|
Steelers
|
Colts
|
0.689
|
1
|
1
| |
13
|
12/6/2015
|
Texans
|
@
|
Bills
|
0.451
|
0
|
0
|
13
|
12/6/2015
|
Titans
|
Jaguars
|
0.568
|
1
|
1
| |
13
|
12/6/2015
|
Vikings
|
Seahawks
|
0.501
|
1
|
0
| |
13
|
12/7/2015
|
Cowboys
|
@
|
Redskins
|
0.673
|
1
|
1
|
13
|
12/7/2015
|
Redskins
|
Cowboys
|
0.327
|
0
|
0
|
_________________________________________________________________________________
No comments:
Post a Comment