2007 Final Regular Season Team Efficiency Ratings

Final regular season NFL team efficiency rankings are listed below in terms of generic winning probability. The GWP is the probability a team would beat the league average team at a neutral site. Each team's opponent's average GWP is also listed (Opp GWP), which can be considered to-date strength of schedule. GWP is adjusted to reflect the strength of past opponents. Offensive ranking (OGWP/O Rank) is based on each team's offensive GWP, i.e. it's the team's GWP assuming it had a league-average defense. D Rank is vice-versa. Rankings are based on a logistic regression model applied to data from all 256 regular season games. A full explanation of the methodology can be found here.

The Bills, Texans, and Dolphins had the toughest schedules. The easiest belonged to the 49ers, Cardinals, and Packers. The best defense this year belonged to the Colts, followed closely by the Titans. The Chargers came on strong in the second half of the year to claim third. The Saints had the worst. The best offense belonged to the Patriots, followed by the Jaguars, Colts, Cowboys, and Packers. The 49ers edged out the Raiders for the worst offense.

Click on the table headers to sort.

Rank	Team	GWP	Opp GWP	OGWP	O Rank	DGWP	D Rank
1	NE	0.87	0.53	0.80	1	0.61	4
2	IND	0.85	0.54	0.68	3	0.69	1
3	JAX	0.77	0.54	0.72	2	0.55	9
4	DAL	0.75	0.50	0.66	4	0.58	8
5	TB	0.73	0.45	0.63	6	0.61	6
6	SD	0.66	0.49	0.53	15	0.64	3
7	PIT	0.65	0.47	0.55	13	0.61	5
8	GB	0.64	0.44	0.63	5	0.50	13
9	PHI	0.62	0.54	0.60	7	0.49	15
10	TEN	0.62	0.55	0.43	22	0.68	2
11	SEA	0.60	0.39	0.56	11	0.52	12
12	WAS	0.60	0.54	0.50	17	0.60	7
13	DEN	0.52	0.50	0.59	8	0.42	28
14	NYG	0.52	0.52	0.44	20	0.55	10
15	HOU	0.49	0.56	0.54	14	0.42	29
16	MIN	0.49	0.45	0.53	16	0.46	20
17	CLE	0.48	0.45	0.56	10	0.44	23
18	NYJ	0.47	0.55	0.43	23	0.49	16
19	CIN	0.46	0.47	0.58	9	0.42	30
20	BUF	0.46	0.57	0.48	19	0.49	18
21	NO	0.40	0.50	0.55	12	0.34	32
22	BAL	0.40	0.51	0.41	25	0.50	14
23	ATL	0.38	0.51	0.48	18	0.43	26
24	CAR	0.37	0.53	0.39	27	0.52	11
25	ARI	0.35	0.42	0.44	21	0.46	19
26	KC	0.34	0.51	0.35	28	0.49	17
27	MIA	0.34	0.55	0.39	26	0.42	31
28	CHI	0.29	0.50	0.33	30	0.45	21
29	DET	0.28	0.51	0.41	24	0.43	25
30	STL	0.25	0.47	0.34	29	0.45	22
31	OAK	0.24	0.52	0.33	31	0.43	27
32	SF	0.15	0.45	0.22	32	0.43	24

Playoff Predictions - Wildcard Round

Game probabilities for the NFL wildcard playoff round are listed below. The probabilities are based on an efficiency win model explained here and here. The model considers offensive and defensive efficiency stats including running, passing, sacks, turnover rates, and penalty rates. Team stats are adjusted for previous opponent strength. Click here for a sortable comparison of playoff team stats.

V Prob	Game	H Prob
0.55	JAX at PIT	0.45
0.31	TEN at SD	0.69
0.36	WAS at SEA	0.64
0.22	NYG at TB	0.78

Road Winners in the Playoffs

One notable difference between the regular season and the playoffs is that the higher seeded playoff team hosts the game. If the seedings reflect the actual strength of the teams, we should expect to see a larger winning percentage for home teams than in the regular season. Teams that win on the road through the playoffs would be truly remarkable teams.

In this post I'll look at road winners in each round of the playoffs from 1994, the year the salary cap began, through 2006, the most recent year. I'll also look at differences in road winners from two different playoff periods. The first period is from 1994-2001, when the playoffs featured 3 division winners and 1 wildcard team from each conference. The second period is from 2002-2006 when the playoffs featured a 4th division winner and two wildcard teams. The sample sizes aren't terribly large, but large enough to begin to see some possible trends.

The first table lists each round of the playoffs for the entire post-salary cap period, and the number and percentage of road winners. Overall road teams won 33% of games in the playoffs. This is in contrast to the normal 42-43% rate in the regular season.

1994-2006	Games	Road Wins	Road Win%
Conference	26	11	42
Division	52	13	25
Wildcard	52	15	29
All Rounds	130	39	32

The next two tables break up the data into the two playoff periods. The first table lists the road winners from '94-'01. The second table lists them from '02-'06, featuring the NFL's current playoff structure.

1994-2001	Games	Road Wins	Road Win%
Conference	16	6	38
Division	32	7	22
Wildcard	32	7	22
All Rounds	75	20	29

2002-2006	Games	Road Wins	Road Win%
Conference	10	5	50
Division	20	6	30
Wildcard	20	8	40
All Rounds	50	19	40

Again, the sample sizes aren't large, but we can make some observations. First, the playoff seedings do appear to reflect a stronger home team winning percentage. The better team is probably seeded higher and hosts the game. But this effect could also be due to the "dome at cold" effect. With bitter cold weather in most cities in January, dome teams would be at a severe disadvantage when playing on the road outdoors. In fact, of the 14 "dome at cold" match-ups, dome teams have won only 2 (14%), which is same the rate as in December in the regular season.

The conference championship games reflect a relatively high number of road winners (40%). This makes sense for a couple reasons. Low-seeded dome teams have already been eliminated. Also, the games feature the closest match-up between relative team strength. They ostensibly would feature the two best teams at the time (the true 1 and 2 seeds), regardless of their official seeding.

Third, there is a noticeable difference between the '94-'01 wildcard rounds and those of '02-'06. Road teams have been winning more frequently lately in the wildcard round (40%, vs 29%). There isn't enough data yet to make any firm conclusions, but it makes sense intuitively.

I think we're seeing more road winners or "seed"-upsets due to the new division structure. With four division winners in each conference, it's more likely an average team can luck into the division championship, especially with a weak division. With only four teams in each (compared to five or six previously), it's more likely to have a thoroughly weak division. The wildcard teams, which are road teams by rule, are probably from the stronger divisions and possibly stronger teams than their seeding indicates. Similar to the conference championship games, the modern wildcard round feature games between closely seeded teams (5 vs. 4) and (3 vs. 6).

This year, the two 5 vs. 4 seed games are expected to be toss-ups. JAX is slightly favored over division-winner PIT by 2.5 points, and TB is favored over NYG by 2.5 points. SEA, the division winner of the clearly weakest division, is only favored over WAS by 3.5 points.

Does Defense Win Championships?

Well, of course. And so does offense. But the conventional wisdom that "defense wins championships" implies that defense is particularly more important than offense in the playoffs and the Super Bowl. This post will begin to look at whether defense really does matter more than offense in the NFL by comparing the right tails of the performance distributions of offenses and defenses--where playoff teams come from.

This time of year we are helped to the standard slew of articles declaring that defense is more important. Here is the latest example from ESPN.com. It's a good example because it suffers from some fatal flaws (itemized by Phil Birnbaum here). Typically, these articles look at past examples of NFL champions and comparing the offensive and defensive rankings of each team. I don't think this kind of analysis is necessarily very valid--

• They are anecdotal
• The sample size is usually very small, and results are probably not statistically significant--just due to chance
  
• If the sample size is large, it covers very distinct periods of NFL passing and blocking rules, confounding any results
  
• If the sample size is limited to one period of NFL rules, it can be dominated by one or two particular teams would skew the results (PIT in the 70s or NE currently, for example)
• Often, the analysis shows that defense is indeed important, but not more important than offense
• The rankings of each squad is almost always based on points scored or total yards, which are more often than not deceiving about the true performance of a squad

The biggest flaw may be using total points or total yards to compare squads. Offensive points scored is dependent on defense ability and special teams. Defenses that provide good field position allow an offense an easier time scoring. Teams with poor defenses provide poor field position, which allows their offense to gain more yards but probably fewer points scored. Additionally, teams that are winning will usually sacrifice total yards for chewing up time on the clock.

There is also an assumption that playoff football is somehow systematically different than regular season football. I'm not comfortable with that assumption--the rules are the same, the refs are the same, the field is still 100 yds long, and a touchdown is still 7 points. There are some unique things about the playoffs, but for now I'm going to set them aside and begin to look at how offenses and defenses compare in general by using regular season stats. (Using post-season stats is possible, but very problematic. About half of all playoff teams only play a single game before being eliminated, yielding very erratic and extreme team averages).

To best compare offensive and defensive performance and ability, squad efficiency--yards per play--should be used. Offensive and defensive efficiency are independent of each other. They aren't dependent on special teams or field position considerations. And perhaps most importantly, they aren't confused by the direction of causation. For example, total rushing yards correlates strongly with winning, but it's actually the winning that allows teams to inflate their rushing yards by eating the clock at the end of a winning game.

Efficiency stats are usually divided into passing and running efficiencies here at NFL Stats. But in this article we're interested in offenses and defenses as a package and not pass/run balance, so we'll use offensive efficiency (yards gained per offensive play) and defensive efficiency (yards allowed per offensive play). Data is from the 2002-2006 NFL seasons (N=160).

The graph below illustrates the actual distribution of team offensive and defensive efficiencies. The vertical axis represents the number of squads with the noted efficiency level (horizontal axis).

Team efficiencies are expected to be distributed approximately normally. That is, there are a few really great offenses, an equally few really poor ones, and a bunch of average ones. League offensive and defensive efficiency must be balanced--there has to be equally as many yards gained as allowed. Accordingly, the average offensive efficiency and the average defensive efficiency are both 5.3 yards per play. The standard deviations are different, however. Offensive efficiency is 0.49 yds and defensive efficiency is about 20% smaller at 0.38 yds.

With enough observations, we would expect the distributions to smooth out, resembling more of a typical bell-shaped normal distribution. With the given averages and standard deviations, the theoretical efficiencies are approximated in the figure below.

The actual and theoretical distributions show the same tendency--offensive efficiencies are spread wider than defensive efficiencies. Notice the right tails of the distributions (below). This is where playoff and championship teams come from. (Actually, we should be comparing the right tail of the offense and the left tail of the defense. But the normal distribution is symmetrical, so rotating the defense about the league average yields the same comparison.)

This means that great offenses tend to be "better" than great defenses, and terrible offenses tend to be "worse" than terrible defenses. If my offense is 2 standard deviations (SD) above the mean and your defense is 2 SD above the mean, my offense would tend to prevail because a great offense tends to gain more yards above the NFL average than an equally great defense allows below the NFL average.

So if a great offense usually trumps a great defense, where does the perception that "defense wins championships" come from? Truly dominant defenses such as the 2000 Ravens, 2002 Buccaneers, or 1985 Bears are relatively rare, and are therefore more memorable. Also, defense has traditionally been overlooked, at least by the mainstream hype-laden media. Even football insiders seem to focus on offense, demonstrated by who is inducted in the Hall of Fame, or who the MVPs tend to be. So the phrase "defense wins championships" may really mean "defense helps win championships more than most people think they do."

One thing that this analysis does not do is focus on the specific case of great offensive vs. great defense. It considers team efficiencies as a whole. While this analysis indicates the likely outcomes of strong offenses vs. strong defenses, it is an indirect inference. A case by case study could look at playoff-type match-ups of good teams only, and tell us more.

Admittedly, there are some special qualities about the playoffs. The outdoor weather in northern cities can be extreme, and the home team is more often the better team. Weather may indeed affect the balance of offense and defense, but it likely affects the balance between running and passing games more. And weather affects both opponents in a game, so it's not clear if it really matters. Playoff weather could also be analyzed in further research.

So when looking at the NFL as a whole, offense and defense balances symmetrically. But when focusing on the right tails of performance, where playoff teams come from, we see that great offenses out-pace equally great defenses.

Playoff Team Efficiency Comparison

Below is a comparison of the efficiency stats for playoff teams that make up the factors in the prediction model. Data from week 17 is not included because most teams rested starters for some or all of the game.

O Pass, D Run, etc. is in yards per attempt. O Int, and D Int are the percent of pass attempts resulting in an interception. O Fum is percentage of all plays in which a fumble occurs. Pen is penalty yards per all plays.

Click on the table headers to sort.

Team	O Pass	O Run	O Int	O Fum	D Pass	D Run	D Int	Pen
DAL	7.5	4.3	3.6	2.1	5.4	4.0	3.5	0.31
GB	7.3	3.9	2.7	2.9	6.0	3.9	3.5	0.33
IND	7.3	3.8	2.7	1.3	5.1	3.9	4.7	0.29
JAX	6.6	4.6	1.9	1.6	5.9	4.0	3.6	0.34
NE	7.8	4.2	1.7	1.2	5.2	4.4	3.6	0.35
NYG	5.4	4.6	3.7	2.8	5.6	4.0	3.1	0.31
PIT	6.2	4.3	2.9	2.0	4.9	3.9	2.2	0.34
SD	6.1	4.2	3.6	1.9	5.7	4.1	5.5	0.37
SEA	6.3	3.6	2.2	1.8	5.6	3.9	3.7	0.18
TB	6.4	4.2	1.5	1.9	5.1	3.7	3.2	0.32
TEN	5.7	4.0	3.9	2.8	5.4	4.0	4.2	0.39
WAS	6.2	3.8	2.2	3.6	5.4	3.9	2.3	0.36