Preliminary 2013 Ratings

By Michael Gertz
Wednesday, August 21, 2013

Once the season begins, our 2013 team ratings will be based solely on games from the 2013 season. Rather than treating all teams equal prior to then, for the 2013 season we have devised a system of preliminary ratings based on our 2012 ratings. The system analyzes how team ratings change on average from one season to the next, and then applies these expected changes to the 2012 ratings.

Overall team ratings typically regress to the mean from one season to the next by about half, whether positive or negative. For instance, a team rated 4.0 can expect to be rated 2.0 on average the following year, while a team rated -3.6 can expect a -1.8 rating. However, further information can be gained by looking as well at individual play types. Just as our mid-season ratings factor in how likely different types of success or failure are to continue, certain play types are more likely to continue at a similar level of success rather than regress to the mean in the next season. Below are the correlation coefficients for different play types between 2 consecutive seasons of ratings. Preliminary 2013 ratings are formed by taking league average values plus the product of the coefficient and the distance from average that a team was in 2012 for each play type.

Correlation Coefficient of Consecutive Season Ratings
	Int	Sack	Pass	Run	KO	Abo	FG	Punt	Pen
Offense	0.15	0.20	0.55	0.16	0.35	0.00	0.03	0.28	0.34
Defense	0.11	0.20	0.28	0.25	0.18	0.00	0.25	0.14	0.00

The predictiveness of different play types is relatively similar to their in-season impacts on our ratings, but some categories like kickoffs are relatively reliable indicators for the next season despite not having a big effect on overall mid-season ratings. And because all categories are not equally predictive for the following season, we can distinguish differences in 2013 expectations even from 2 teams with equal 2012 ratings. Below we compare 2012 ratings to our new preliminary 2013 ratings.

Comparing 2012 and Preliminary 2013 Ratings
Rank	Team	2012	2013
1	DEN	6.6	3.1
2	NE	6.9	2.9
3	ATL	4.2	2.6
4	NO	3.5	2.5
5	GB	4.4	2.3
6	SEA	5.5	1.8
7	DAL	2.1	1.7
8	SF	5.0	1.6
9	BAL	3.1	1.1
10	DET	0.0	0.8
11	NYG	2.7	0.8
12	PIT	0.7	0.8
13	HOU	2.4	0.7
14	IND	-0.3	0.6
15	WAS	2.4	0.6
16	CAR	1.2	0.5
17	CHI	0.9	-0.1
18	PHI	-1.8	-0.1
19	CIN	0.7	-0.1
20	SD	-1.1	-0.3
21	OAK	-2.9	-0.3
22	TB	-0.1	-0.4
23	STL	-0.8	-0.5
24	BUF	-1.8	-0.7
25	MIN	-0.2	-0.8
26	MIA	-1.4	-0.8
27	TEN	-3.2	-1.3
28	CLE	-3.6	-1.7
29	NYJ	-4.2	-1.9
30	JAC	-5.6	-2.1
31	KC	-5.9	-2.5
32	ARI	-5.9	-2.7

The new ratings don't provide many huge surprises, but do produce some interesting quirks. Teams with a strong offensive passing game, the most predictive play type, generally climb up the ratings compared to 2012. Teams that happened to string together strong 2012 performances in several less predictive play types appear more likely to regress.

Keep in mind that these preliminary ratings are based solely on efficiency stats, and do not account for offseason roster changes. We will examine the impacts of individual players such as new starting quarterbacks and how to make rough adjustments that deviate a bit from the above baseline figures in our upcoming division and league preview articles.