January 4, 2016

Exponential Smoothing Ratings

Filed under: Uncategorized — wwinston @ 9:04 am

Here I started with team ratings based on Vegas preseason over under wins. Then I used exponential smoothing with alpha=.10 to update ratings. for example, if a team was favored by 10 in a game and won by 20, their rating would increase by .1*(20-10) = 1 point. This method gives more weight to more recent games. For example weight for last week’s game = .9*weight for current game. With these ratings I cannot predict the score of a game. only margin of victory. NFC is looking tough with top 3 teams.  These ratings seem consistent with the late season declines of NE, GB, DEN, and CINN. They also show the great improvement made by teams like the Lions and Seahawks. I will use these ratings  to estimate playoff probabilities by tomorrow.

 

RANK TEAM RATING
1 SEA 11.7
2 CAR 9.2
3 ARI 8.3
4 PITT 7.2
5 CINN 6.3
6 KC 6.1
7 MINN 5.5
8 NE 5.3
9 DEN 3.9
10 JETS 2.8
11 GB 2.6
12 HOU 2.3
13 WASH 2
14 BUFF 1.7
15 DET -0.2
16 STL -1.5
17 OAK -2.4
18 GIANTS -2.5
19 ATL -2.6
20 BALT -3.1
21 SD -3.2
22 PHI -3.2
23 CHI -3.3
24 NO -3.6
25 DALL -4
26 MIA -4.3
27 INDY -4.9
28 SF -6
29 TAMPA -6.3
30 CLE -6.6
31 JACK -6.7
32 TENN -10.5

No Comments »

No comments yet.

RSS feed for comments on this post. TrackBack URL

Leave a comment

Powered by WordPress