2008 Luckiest Teams

Here's the final ranking of lucky and unlucky teams for 2008. I estimate team luck by using my efficiency regression model to calculate each team's expected wins--how many wins a team can normally expect, on average, given their actual performance in offensive and defensive running, passing, turnovers, and penalties. Expected wins are adjusted for average opponent strength.

The difference between the expected wins and actual wins is what I loosely estimate as team luck. For a brief introduction on the concept, see this last post on the subject. For a more thorough but general discussion, see this essay on NFL luck. The executive summary is that there are good and bad breaks for every team in every game, and often they'll roughly even out, but many times they won't. The bottom line is that a 16-game season is far too short for all the breaks to even out. With 32 teams in the league, some are going to be luckier than others.

Tennessee, New England, and the New York Jets top the list of lucky dogs in 2008, while San Diego, Kansas City, and New Orleans appear to be among the unluckiest.

Now, I don't claim this is all luck. Part of it could be game-day coaching. For those critics out there of Norv Turner or Andy Reid, you can point to this and say they managed to take teams with some of the most spectacular statistical performances and make mediocre records. You could also make the opposite claim about Jeff Fisher or Bill Belichick.

Also of note are the 0-16 Detroit Lions. They could have been expected to win 1 or 2 games given their stats this year.

You can click on the headers any column to sort the table. I've included a division column so you can see how teams lucked-out compared to their division-mates.

Rank	Team	Exp. W	Act. W	Luck	Div
1	TEN	10.5	13	+2.5	AS
2	NE	8.9	11	+2.1	AE
3	NYJ	7.0	9	+2.0	AE
4	SF	5.2	7	+1.8	NW
5	BUF	5.3	7	+1.7	AE
6	MIN	8.4	10	+1.6	NN
7	IND	10.6	12	+1.4	AS
8	NYG	11.0	12	+1.0	NE
9	HOU	7.0	8	+1.0	AS
10	CLE	3.1	4	+0.9	AN
11	MIA	10.1	11	+0.9	AE
12	ARI	8.2	9	+0.8	NW
13	DEN	7.5	8	+0.5	AW
14	BAL	10.5	11	+0.5	AN
15	PIT	11.9	12	+0.1	AN
16	JAX	4.9	5	0.1	AS
17	CHI	9.0	9	0.0	NN
18	OAK	5.1	5	-0.1	AW
19	TB	9.1	9	-0.1	NS
20	CIN	4.7	4.5	-0.2	AN
21	DAL	9.3	9	-0.3	NE
22	CAR	12.5	12	-0.5	NS
23	ATL	11.5	11	-0.5	NS
24	SEA	4.5	4	-0.5	NW
25	STL	2.8	2	-0.8	NW
26	DET	1.8	0	-1.8	NN
27	WAS	10.0	8	-2.0	NE
28	GB	8.5	6	-2.5	NN
29	PHI	12.1	9.5	-2.6	NE
30	NO	10.7	8	-2.7	NS
31	KC	4.9	2	-2.9	AW
32	SD	11.4	8	-3.4	AW

2008 Final Efficiency Rankings

The ratings are listed below in terms of generic win 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, which can be considered to-date strength of schedule, and all ratings include adjustments for opponent strength.

Offensive rank (ORANK) is offensive generic win probability which is based on each team's offensive efficiency stats only. In other words, it's the team's GWP assuming it had a league-average defense. DRANK is is a team's generic win probability rank assuming it had a league-average offense.

GWP is based on a logistic regression model applied to current team stats. The model includes offensive and defensive passing and running efficiency, offensive turnover rates, and team penalty rates. A full explanation of the methodology can be found here. This year, however, I've made one important change based on research that strongly indicates that defensive interception rates are highly random and not consistent throughout the year. Accordingly, I've removed them from the model and updated the weights of the remaining stats.

RANK	TEAM	LAST WK	GWP	Opp GWP	O RANK	D RANK
1	CAR	1	0.79	0.55	4	6
2	PHI	2	0.77	0.52	11	2
3	PIT	3	0.75	0.53	20	1
4	ATL	4	0.73	0.54	3	17
5	SD	5	0.72	0.53	1	10
6	NYG	6	0.69	0.53	2	13
7	NO	8	0.68	0.55	5	19
8	IND	12	0.68	0.49	7	15
9	TEN	7	0.67	0.46	15	4
10	BAL	11	0.67	0.52	17	3
11	MIA	9	0.64	0.44	6	20
12	WAS	10	0.63	0.51	12	11
13	DAL	14	0.59	0.53	16	8
14	TB	13	0.58	0.55	19	9
15	CHI	15	0.57	0.51	22	7
16	NE	16	0.57	0.47	13	24
17	GB	17	0.54	0.51	14	18
18	MIN	19	0.53	0.52	23	5
19	ARI	18	0.52	0.50	9	23
20	DEN	20	0.48	0.50	8	27
21	HOU	22	0.45	0.50	10	28
22	NYJ	21	0.45	0.45	24	12
23	BUF	24	0.34	0.44	27	22
24	SEA	28	0.34	0.47	30	14
25	OAK	27	0.33	0.56	29	16
26	JAX	23	0.32	0.50	18	30
27	KC	26	0.31	0.53	21	26
28	CIN	29	0.30	0.55	32	21
29	SF	25	0.29	0.49	25	25
30	CLE	30	0.20	0.57	28	29
31	STL	31	0.18	0.52	26	31
32	DET	32	0.12	0.57	31	32

To-date efficiency stats below. As always, click on the headers to sort.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.1	3.5	0.024	0.028	6.5	4.0	0.025	0.39
ATL	7.4	4.4	0.025	0.015	6.0	4.9	0.018	0.29
BAL	6.0	4.0	0.028	0.025	5.1	3.6	0.049	0.40
BUF	5.9	4.2	0.031	0.034	6.3	4.3	0.020	0.28
CAR	7.3	4.8	0.029	0.014	5.7	4.4	0.022	0.32
CHI	5.5	3.9	0.027	0.016	5.9	3.4	0.035	0.29
CIN	4.3	3.6	0.029	0.027	6.3	3.9	0.024	0.30
CLE	4.6	3.9	0.041	0.024	7.1	4.5	0.052	0.35
DAL	6.6	4.3	0.037	0.032	5.3	4.2	0.016	0.49
DEN	7.1	4.8	0.029	0.019	7.0	5.0	0.012	0.37
DET	5.3	3.8	0.037	0.036	7.9	5.1	0.009	0.38
GB	6.6	4.1	0.024	0.023	6.0	4.6	0.042	0.49
HOU	7.3	4.3	0.036	0.028	6.9	4.5	0.025	0.34
IND	6.8	3.4	0.021	0.010	5.9	4.2	0.031	0.32
JAX	5.8	4.2	0.024	0.021	7.3	4.0	0.028	0.42
KC	5.4	4.8	0.030	0.022	7.0	5.0	0.025	0.32
MIA	7.0	4.2	0.014	0.017	6.2	4.2	0.033	0.34
MIN	6.0	4.5	0.038	0.028	6.0	3.3	0.023	0.35
NE	6.1	4.4	0.021	0.016	6.4	4.1	0.030	0.25
NO	7.7	4.0	0.028	0.018	6.4	4.2	0.029	0.39
NYG	6.1	5.0	0.020	0.017	5.8	4.0	0.034	0.42
NYJ	5.9	4.7	0.043	0.024	6.1	3.7	0.024	0.28
OAK	5.2	4.3	0.026	0.033	6.4	4.7	0.034	0.42
PHI	6.2	4.0	0.026	0.015	5.1	3.5	0.029	0.31
PIT	5.9	3.7	0.030	0.026	4.3	3.3	0.038	0.41
SD	7.7	4.1	0.023	0.017	6.3	4.0	0.025	0.38
SF	6.0	4.0	0.037	0.043	6.1	3.8	0.022	0.37
SEA	5.1	4.2	0.032	0.021	6.9	4.2	0.016	0.30
STL	5.2	4.0	0.037	0.023	7.3	4.9	0.027	0.37
TB	6.1	4.1	0.023	0.020	5.9	4.3	0.046	0.42
TEN	6.1	4.3	0.020	0.019	5.2	3.7	0.035	0.43
WAS	5.5	4.4	0.012	0.020	5.8	3.8	0.025	0.33
Avg	6.1	4.2	0.028	0.023	6.2	4.2	0.028	0.36

Establishing the Run?

"Establish the run" might be the three most over-used words in football analysis. Bad football analysis, that is. You would give yourself one heck of hangover playing one of those fraternity-type drinking games going bottoms-up every time you heard that phrase on a Sunday.

Establishing the run could mean a lot of different things. To most people I think it means that offenses will demonstrate the willingness to run frequently, hoping that defenses will bias towards stopping run later in the game. I think this interpretation has been debunked fairly thoroughly, so I'm going to look at it from a different angle.

Establishing the run could mean running often early in a game in hopes that a defense will weaken, and runs will be longer later in the game. I think this is more plausible theory. It makes sense on the surface. Running plays tend to batter the defense while passing plays allow the defense to go on the attack. By the fourth quarter, it's believable that defenders would be battered and fatigued. Defenders would get off of blocks slower and tackles would be sloppier. Watching Baltimore's two consecutive long TD runs late in the fourth quarter at Dallas in week 16 made me wonder: Does running frequently lead to longer gains?

To answer the question, I compared the average gains from running plays based on when they took place in a game. By "when," I mean which run play it was, not what time it took place in the game. In other words, if running frequently fatigues a defense, then the gain of a team's 30th run should tend to be longer than its first run.

We'd expect that the more runs an offense calls, the longer the subsequent runs should tend to be. The graph below depicts the average gain of each rushing play based on its order in which it was run. It plots the average gain of each team's first run of a game, 2nd run, 3rd run, ...etc.

(Data is from all regular season games from 2000-2007. All runs except kneel-downs were included.)

There's no increase in average gain as the number of runs increase. A team's very first run of a game is just as long as, if not longer than, the 20th, 30th or even 40th.

This result is despite the expectation that teams that are good at running would naturally run more often. Teams that are ahead toward the end of the game Those teams would therefore be the ones that we'd expect would accumulate more attempts. If so, we'd see the average gains increase as the run attempts mount. But we don't.

So this is evidence that runs are just runs, no matter how many have come before them. An offense can expect the same average gain on the first snap of the game as on the 80th, after 30 previous run plays.

The run may not set up the run, but does it set up the pass? Does running frequently allow longer gains on passes later in the game? I'll look at that question next.

Edit: Follow-up here that addresses many of the comments below.

Weekly Roundup

In his Tuesday Morning Quarterback column from a week ago, Gregg Easterbrook tells us about his favorite high school state champions, the Pulaski Academy of Arkansas. They never punt on 4th downs, and they almost always win. To his credit, Easterbrook has been out in front of the go-for-it movement. (I'm not a regular reader of Easterbrook, and now I remember why. His columns are the length of short novels, frequently meandering into bad politics and pop-culture commentary.) Thanks to all the readers who alerted me to the Pulaski article at Rivals.

FO has a good article on the hidden value of pass interference penalties. Devin Hester is used to illustrate how the value of speedy deep-threat WRs is obscured by the fact that they aren't credited with the penalty yards they draw.

Dave Berri from Wages of Wins posted his weekly QB and RB rankings based on econometric models like my own. Note who the top QB is--Pennington. Dave may have an updated ranking by the time you read this. He also adds his thoughts on an ongoing discussion between amateur/internet sabermetricians and academic researchers.

Last week I wrote that the Redskins collapse this year may have been due to too few interceptions. In other words, their very low interception rate may have become an end in itself, rather than a by-product of a good passing attack.

I wrote, "You can guarantee zero interceptions by playing in an extremely conservative way, tossing short passes, taking sacks, or throwing the ball away anytime a defender is in the same zip code as the receiver. You can minimize interceptions, but you'll lose every game doing it. At some point in risk-reward continuum, there is an optimum level of risk in passing strategy."

The PFR blog appears to have picked up on the theory a few days later and added some evidence that supports it. We appear to be of like mind on the topic, as the PFR post says, "If you never throw an interception, you’re taking too many sacks, throwing too many balls out of bounds, and getting too many four yard gains on 3rd-and-9. So if zero is not the optimal turnover rate, then what is?"

I have a couple small constructive suggestions. The post finds that teams that have high rates of turnovers per non-scoring drive score more points than teams with low turnover rates. First, I'd suggest looking at interceptions per non-scoring drive instead of all turnovers. Interceptions more than fumbles are functions of an offense's risk-reward balance. I'm not sure if it really matters, though. The effect is the same no matter what kind of turnover it is. But it's worth looking at.

Second, and more importantly, I'd look at point differential rather than points scored. In other words, do teams with more turnovers per non-scoring drive outscore opponents? It may be that they score more points themselves, but they may be allowing even more points due to handing favorable field position over to their opponents.

Smart Football, the best Xs and Os site on the web, dissects Paul Johnson's 'Flexbone' offense at Georgia Tech. I'm a big Paul Johnson fan because he brought Navy's program a lot of success over the past several years. I'm not a big Xs and Os guy, meaning I'm not an expert. But I'd like to learn more. Can anyone suggest a good book that digs deep into NFL-style offensive or defensive systems?

Super Bowl Probabilities

The updated Super Bowl probabilities are out. Despite the #1 vs. #2 match-ups in week 16, there aren't any big changes this week. The big drop-off between the top 5 and everyone else is mostly due to two factors: 1) just clinching a playoff spot, and 2) getting a first round bye.

Team	Conf Champ	SB Champ
PIT	39	20
CAR	33	20
TEN	35	15
NYG	29	13
ATL	25	15
MIA	5	2
IND	5	2
BAL	5	2
TB	3	1
MIN	2	1
ARI	2	1
PHI	1	1
DAL	1	1
DEN	1	1

Thanks as always to NFL-Forecast.com.

Playoff Probabilities Week 16

Courtesy of Chris at NFL-Forecast.com, here are the latest playoff probabilities for each team.

These are calculated using the NFL-Forecast software mini-app that runs thousands of simulated seasons. The outcomes are based on game-by-game probabilities with every crazy tie-breaking scenario factored in. Chris has used the probabilities from Advanced NFL Stats as his default game probabilities for the past two seasons.

There are two tables below. The first lists the probability that each team will finish in each place in their division. The second table lists the overall playoff probabilities, broken down by seed.

Quite a few people thought I was nuts when I said Miami and San Diego had the inside tracks on winning their divisions halfway through the season. If there was ever a week to have 'Sunday Ticket,' this would be it. Eight of the 16 games have do-or-die playoff consequences--not just seeding implications. Eleven of the 32 teams are still on the playoff bubble in the final week of the season.

AFC EAST
Team	1st	2nd	3rd	4th
MIA	61	15	24	0
NE	24	61	15	0
NYJ	15	24	61	0
BUF	0	0	0	100
AFC NORTH
Team	1st	2nd	3rd	4th
PIT	100	0	0	0
BAL	0	100	0	0
CIN	0	0	54	46
CLE	0	0	46	54
AFC SOUTH
Team	1st	2nd	3rd	4th
TEN	100	0	0	0
IND	0	100	0	0
HOU	0	0	100	0
JAX	0	0	0	100
AFC WEST
Team	1st	2nd	3rd	4th
SD	78	22	0	0
DEN	22	78	0	0
OAK	0	0	100	0
KC	0	0	0	100
NFC EAST
Team	1st	2nd	3rd	4th
NYG	100	0	0	0
PHI	0	75	7	19
WAS	0	0	73	27
DAL	0	25	21	54
NFC NORTH
Team	1st	2nd	3rd	4th
MIN	66	34	0	0
CHI	34	66	0	0
GB	0	0	100	0
DET	0	0	0	100
NFC SOUTH
Team	1st	2nd	3rd	4th
CAR	55	46	0	0
ATL	46	55	0	0
TB	0	0	100	0
NO	0	0	0	100
NFC WEST
Team	1st	2nd	3rd	4th
ARI	100	0	0	0
SF	0	100	0	0
SEA	0	0	100	0
STL	0	0	0	100

AFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
TEN	100	0	0	0	0	0	100
PIT	0	100	0	0	0	0	100
IND	0	0	0	0	100	0	100
BAL	0	0	0	0	0	89	89
SD	0	0	0	78	0	0	78
MIA	0	0	61	0	0	0	61
NE	0	0	24	0	0	7	31
DEN	0	0	0	22	0	0	22
NYJ	0	0	15	0	0	4	19
NFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
NYG	100	0	0	0	0	0	100
CAR	0	55	0	0	46	0	100
ATL	0	46	0	0	54	1	100
ARI	0	0	0	100	0	0	100
MIN	0	0	66	0	0	0	66
TB	0	0	0	0	0	63	63
CHI	0	0	34	0	0	3	38
DAL	0	0	0	0	1	24	25
PHI	0	0	0	0	0	9	9

Week 17 Game Probabilities

Win probabilities for week 17 NFL games are listed below. Week 17 is notoriously unpredictable, so these probabilities should be understood to be based on the assumption both teams are playing at full strength. Adjust them as desired to account for teams resting starters or focusing on Caribbean beaches rather than the gameplan.

The probabilities are based on an efficiency win model explained here and here with some modifications. 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.

Pwin	GAME	Pwin
0.04	STL at ATL	0.96
0.18	JAX at BAL	0.82
0.63	NE at BUF	0.37
0.44	KC at CIN	0.56
0.08	DET at GB	0.92
0.59	CHI at HOU	0.41
0.47	TEN at IND	0.53
0.59	NYG at MIN	0.41
0.50	CAR at NO	0.50
0.59	MIA at NYJ	0.41
0.26	DAL at PHI	0.74
0.06	CLE at PIT	0.94
0.16	OAK at TB	0.84
0.25	SEA at ARI	0.75
0.22	DEN at SD	0.78
0.76	WAS at SF	0.24

Week 16 Efficiency Rankings

The ratings are listed below in terms of generic win 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, which can be considered to-date strength of schedule, and all ratings include adjustments for opponent strength.

Offensive rank (ORANK) is offensive generic win probability which is based on each team's offensive efficiency stats only. In other words, it's the team's GWP assuming it had a league-average defense. DRANK is is a team's generic win probability rank assuming it had a league-average offense.

GWP is based on a logistic regression model applied to current team stats. The model includes offensive and defensive passing and running efficiency, offensive turnover rates, and team penalty rates. A full explanation of the methodology can be found here. This year, however, I've made one important change based on research that strongly indicates that defensive interception rates are highly random and not consistent throughout the year. Accordingly, I've removed them from the model and updated the weights of the remaining stats.

RANK	TEAM	LAST WK	GWP	Opp GWP	O RANK	D RANK
1	CAR	2	0.78	0.55	6	7
2	ATL	3	0.76	0.57	1	16
3	PHI	4	0.76	0.52	10	2
4	PIT	1	0.74	0.54	20	1
5	SD	8	0.71	0.53	4	12
6	NO	6	0.70	0.54	2	14
7	NYG	5	0.70	0.53	3	11
8	TEN	7	0.69	0.44	15	4
9	WAS	10	0.66	0.52	11	10
10	MIA	9	0.64	0.42	5	20
11	BAL	13	0.64	0.52	19	3
12	IND	14	0.63	0.47	8	19
13	TB	12	0.62	0.58	18	8
14	DAL	11	0.61	0.52	16	9
15	CHI	15	0.59	0.52	22	6
16	NE	20	0.55	0.47	14	23
17	GB	17	0.54	0.55	12	17
18	MIN	18	0.53	0.51	25	5
19	DEN	19	0.50	0.49	7	26
20	ARI	16	0.50	0.51	9	24
21	NYJ	21	0.45	0.43	23	15
22	HOU	22	0.41	0.49	13	30
23	JAX	23	0.35	0.49	17	27
24	SF	26	0.33	0.48	24	25
25	BUF	24	0.33	0.43	27	21
26	OAK	30	0.31	0.55	30	13
27	KC	27	0.31	0.54	21	28
28	SEA	25	0.30	0.46	31	18
29	CIN	28	0.29	0.56	32	22
30	CLE	29	0.22	0.55	26	29
31	STL	31	0.16	0.50	28	31
32	DET	32	0.13	0.58	29	32

To-date efficiency stats below. As always, click on the headers to sort.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.0	3.3	0.024	0.028	6.5	4.0	0.023	0.37
ATL	7.4	4.1	0.022	0.015	6.1	4.9	0.019	0.31
BAL	5.7	4.0	0.029	0.026	5.1	3.5	0.048	0.39
BUF	6.0	4.2	0.033	0.035	6.3	4.4	0.021	0.29
CAR	7.1	4.8	0.030	0.015	5.6	4.4	0.022	0.33
CHI	5.5	3.9	0.029	0.015	5.8	3.4	0.038	0.31
CIN	4.3	3.5	0.031	0.028	6.4	4.0	0.026	0.29
CLE	4.8	3.9	0.038	0.024	7.1	4.5	0.053	0.34
DAL	6.7	4.3	0.038	0.029	5.2	4.3	0.016	0.50
DEN	7.1	4.7	0.028	0.020	6.9	4.8	0.013	0.35
DET	5.3	3.8	0.036	0.039	7.9	5.0	0.010	0.37
GB	6.6	3.9	0.025	0.021	6.0	4.7	0.042	0.49
HOU	7.1	4.3	0.039	0.029	7.0	4.5	0.028	0.33
IND	6.8	3.4	0.022	0.010	6.0	4.1	0.032	0.32
JAX	5.8	4.2	0.021	0.018	7.0	4.0	0.030	0.43
KC	5.4	4.9	0.032	0.022	7.2	5.0	0.026	0.30
MIA	7.1	4.2	0.015	0.016	6.2	4.2	0.029	0.35
MIN	5.9	4.5	0.038	0.028	5.9	3.2	0.024	0.36
NE	6.1	4.5	0.021	0.017	6.5	4.1	0.031	0.26
NO	7.7	4.0	0.029	0.019	6.2	4.1	0.030	0.39
NYG	6.1	5.0	0.022	0.018	5.6	4.0	0.034	0.42
NYJ	5.9	4.8	0.041	0.022	6.1	3.7	0.026	0.28
OAK	5.1	4.2	0.025	0.036	6.3	4.7	0.034	0.43
PHI	6.2	4.0	0.027	0.014	5.1	3.5	0.029	0.33
PIT	5.9	3.6	0.029	0.028	4.4	3.3	0.035	0.40
SD	7.6	3.8	0.024	0.019	6.2	3.9	0.023	0.37
SF	5.9	4.0	0.038	0.042	6.2	3.8	0.023	0.38
SEA	5.1	4.3	0.030	0.020	6.8	4.1	0.015	0.28
STL	5.2	3.8	0.039	0.024	7.3	4.7	0.024	0.38
TB	6.0	4.0	0.023	0.020	5.9	4.2	0.046	0.40
TEN	6.2	4.3	0.021	0.019	5.0	3.7	0.037	0.44
WAS	5.6	4.4	0.013	0.020	5.7	3.8	0.025	0.35
Avg	6.1	4.1	0.028	0.023	6.2	4.1	0.028	0.36

Realtime Win Probabilities

If you haven't checked out the win probability site during a game in a while, you should check it out. The win probability model itself has been improved incrementally over the past few weeks, plus I've added a few additional features.

For each upcoming play, the probability of a first down is displayed. The Expected Points for each field position, down, and distance are also updated live. Additionally, the probabilities of the current drive culminating in a touchdown and a field goal is displayed.

Any questions, suggestions, or other comments are welcome.

Anatomy of a Collapse

Going into week 9 of the 2008 season, the Washington Redskins were 6-2. Their two losses came in a low scoring affair against the Giants at the Meadowlands, and in a fluky 19-17 game hosting the Rams. They had beaten division rivals and pre-season darlings Dallas and Philadelphia, both on the road. They outscored the prolific Saints and Cardinals offenses. The playoffs were a virtual certainty, and challenging the Giants for NFC supremacy wasn't out of the question.

Then, the Redskins dropped the next 5 out 6 games, falling to 7-7 and eliminating themselves from playoff contention in all but the most improbable scenarios. In fact, despite a late-season win over the Eagles this week, they have been officially eliminated. What went wrong?

Before I go any further, keep in mind that maybe nothing went wrong. Say the Redskins had about a 50/50 shot at winning every one of their games--essentially a coin flip. It's not terribly uncommon for 8 coin flips to turn up 6 heads and 2 tails. And it wouldn't be unfathomable that another few flips wouldn't result in 5 tails and a head or two. In fact, it's looking like the Redskins will finish either 8-8 or 9-7, essentially a coin flip of a team. There's nothing that says a team's 8 wins have to be evenly paced throughout a season.

But teams aren't coins, so I'll take a deeper look at the Redskins' performance game by game. I'll look at their four major components--running and passing, on both offense and defense.

I measure passing performance by adjusted net yards per attempt. This is passing yards minus sack yards, minus 40 yards per interception, per drop back. Running performance is measured simply by yards per run.

Opponent strength is an important factor. For each game I adjust the Redskins' performance in each phase of the game according to how strong the opposing team is in the opposite phase. I simply added the opponent's difference from average to the Redskins' efficiencies. For example, if the Redskins' offensive passing efficiency was 5.0 yds per attempt, and their opponent's season-long yds per att allowed was 1.5 yards above league average, I gave the Redskins 6.5 yds per attempt.

The graphs below plot performance as a 4-game moving average. This is to account for the natural game-to-game variation in performance. We shouldn't expect a team to maintain steady efficiencies in every game they play. Ultimately, the moving average allows us to see team-wide season-long improvements or declines while accounting for opponent strength.

The first graph illustrates opponent-adjusted offensive efficiency for passing and running, and the second illustrates defensive efficiencies.





We can see that the defense was relatively steady all year. Oddly, it looks like the pass defense improved following the eighth game, which is when the 1-5 slide started. (Remember lower is better for defense.) The run defense was rock-steady all year.

The rushing offense was equally steady and remained above-average all year. But Washington's passing game showed a steady decline throughout the season.

So it was the passing game, not Portis, not the secondary, not the pass rush. My theory is that they threw too few interceptions early in the year. That's not a typo. I think head coach Jim Zorn or quarterback Jason Campbell became captive to the idea that they were winning because they weren't throwing interceptions. While a lack of turnovers certainly helps win games, if it becomes a passer's primary goal it could be harmful.

Interceptions are a part of the bargain, a natural consequence to throwing the ball. You can guarantee zero interceptions by playing in an extremely conservative way, tossing short passes, taking sacks, or throwing the ball away anytime a defender is in the same zip code as the receiver. You can minimize interceptions, but you'll lose every game doing it.

At some point in the risk-reward continuum, there is an optimum level of risk in passing strategy. I'm guessing the Redskins found themselves too far on the conservative side of optimum.

Weekly Roundup

This week we see just how bad the NFC West is in 2008, and how team penalties can be an overlooked factor. Plus, a cool Wikipedia page about the Patriots' offensive and defensive systems.

The PFR blog has two good posts this week. First, they compare the intra- and inter-division records of the NFC West. The bottom line is it may be one of the weakest divisions in a long time. The NFC South, on the other hand, may be one of the strongest.

PFR is also trying to establish a sound way of measuring the best defenses of all time. There's points allowed, yards allowed, yards per play allowed, plus turnovers and other considerations. This year's Steelers defense may rank among the best. If you ask me, the 2000 Ravens hold the crown.

Regular readers know I'm no big fan of Football Outsiders. But when they have interesting stuff, I'll take note. This week they look at one of the most overlooked team stats--penalties. They look at playoff contender team penalties and at "penalties against." In my own research, a team 1 standard deviation better than average in penalty yards per play would win and additional 0.4 wins per season. Penalties against appear slightly more important at 0.5 wins per standard devation.

The Dolphins look like favorites against the struggling Chiefs, except that they'll be playing in some of the coldest temperatures on record. With the weather getting cold, here are a few articles on weather from last season. Warm climate teams don't seem to have as big a problem as dome teams. Dome teams playing in the cold have only won 12% of their games in the last several seasons.

Super Bowl Probabilities

The updated Super Bowl probabilities are out. The Steelers' win over Baltimore strengthens their odds considerably, while Carolina becomes the NFC favorite.

Team	Conf Champ	SB Champ
PIT	48	26
CAR	39	22
TEN	32	15
NYG	29	13
ATL	13	7
MIA	6	3
IND	5	2
TB	5	2
BAL	4	2
PHI	3	2
ARI	4	1
DAL	2	1
MIN	3	1
DEN	3	1

Thanks to NFL-Forecast.com.

ANNOUNCEMENT: Adv NFL Stats Community

I get emails periodically from fellow stat-heads asking me to review some football research. Often they’re professors or statistics students writing an academic paper, or an amateur enthusiast like me. There have also been a number of guys who have started up a website very similar to mine, and have asked for a link to their sites which I’m always happy to do. Very often however, they aren’t able to sustain the effort and the sites go dark after a few weeks.

I thought a good idea would be to provide a place where stat-heads can post their own research. Ideally, this would be a place with an existing readership base that can review and comment. It would be a place where we could exchange ideas and even data. Unlike some other stat-oriented sites, it would become a collaborative open community without “premium” content or “proprietary” black-box stats.

So today I’m announcing the creation of Advanced NFL Stats Community, an adjunct site where stats guys can post their own research, analysis, or even just ‘fact-backed’ opinions.

The call goes out. Everyone is invited, and I won’t filter posts that disagree with my own theories. If you’ve done some of your own research and want to share it or find out what other people think, send me what you’ve got and I’ll be happy to add it. Or if you’re trying to build readership for your own stats site, that’s fine too. Just shoot me an email at the address in the About | Contact page in the menu above, or you can post a submission directly to the site by emailing hatch113.statscommunity@blogger.com.

To get the ball rolling, I’m going to make two of my primary databases available for anyone to use as a basis for research. Both my team statistic database and my game result database for the 2002 through 2007 seasons are now published. These are the databases I use for my primary efficiency regression models for team rankings and game predictions.

My only request is that if you use the data for other purposes you credit this site. Lots of other data can be found all across the web, at sites such as nfl.com espn.com, and myway.com and can be easily copied into Excel (Copy/Paste Special/Text).

Welcome.

Giants, Lions, Home Field Advantage, and Time Travel

Some of the greatest breakthroughs in math and science have come when people question the rules. For example, negative numbers questioned the premise that there could not be quantities less than zero. To us in modern times, negative numbers seem intuitive, but to ancient thinkers it was a very difficult idea to contemplate. In fact, the existence of zero itself was very controversial for ages.

There are plenty of other examples. One of the greatest breakthroughs in geometry came when the foundations of Euclidian geometry were questioned. For millennia, mathematicians worked with the 2-dimensional Cartesian plane (x, y) and then 3-dimensional spaces (x, y, z). But it wasn’t until well into the 19th Century that anyone wondered what 4- or n-dimensional math would be like.

Euclid taught us that the interior angles of a triangle always add up to 180 degrees. But when we plot a triangle on, say a globe of the Earth, the triangle appears to bulge slightly and the angles add up to greater than 180 degrees. On surfaces with negative curvature, the angles sum to less than 180 degrees. Questioning that single assumption gave birth to new fields of science that help us understand our universe.

Remember “imaginary numbers,” like 4i or -3i? We were taught that you can never take a square root of a negative number. Then one day in algebra class, they told us “but if you do, just throw an i after the number and keep going.” At some point, someone must have asked, what if you could take the square root of an negative number? What would math be like then? And so a whole new field of mathematics opened up. Imaginary numbers are an essential concept in applications such as systems engineering and quantum theory.

What does this have to do with the Lions and Giants? Recently, I was trying to explain why the effect of home field advantage (HFA) is stronger for closely matched teams and weaker for mis-matched teams. Let’s say for closely matched teams, who would each have a .50/.50 shot at winning at a neutral site, HFA makes the game a .60/.40 proposition. We could describe the strength of HFA as +.10/-.10.

But now take a game where the Lions are playing the Giants. At a neutral site, the game might be something like a .95/.05 proposition in favor of New York. In other words, the Lions would pull of an upset in 1 out of 20 games. But if the game were at the Meadowlands and we apply the same +.10/-.10 adjustment for HFA, the probability the Giants would win would be 1.05, and the probability the Lions would win would be -0.05.

But because probabilities can never be greater than 1 or less than 0, this obviously can’t be the case. Therefore, the effect of HFA must diminish for mis-matched teams. (But if anyone could have a negative probability of winning, it might be the Lions this year!)

Then I thought, let’s question the assumption. Why can’t probabilities be greater than 1 or less than zero? As with negative numbers, or non-Euclidian triangles, or imaginary numbers, let’s just throw away the assumption and keep chugging. What would a universe be like with “imaginary” probabilities?

It’s almost impossible to wrap your brain around such a concept. I don’t know what it would mean. Metaphysical do-overs? Branching timelines? Can the future affect the past?

A quick Google search for “negative probabilities” turns up a number of results, and clearly this has been thought of before. It appears to be an alternative way to explain quantum mechanics. But it was just a weird, stray thought that I thought I’d share.

Right now I've got the Lions with a 10% chance they'll win this weekend, and a 9% chance they'll win next week. This equates to an 18% chance they'll win at least 1 of those 2 games, and an 82% chance they'll finish 0-16.

Playoff Probabilities Week 15

Courtesy of Chris at NFL-Forecast.com, here are the latest playoff probabilities for each team.

These are calculated using the NFL-Forecast software mini-app that runs thousands of simulated seasons. The outcomes are based on game-by-game probabilities with every crazy tie-breaking scenario factored in. Chris has used the probabilities from Advanced NFL Stats as his default game probabilities for the past two seasons.

There are two tables below. The first lists the probability that each team will finish in each place in their division. The second table lists the overall playoff probabilities, broken down by seed.

AFC EAST
Team	1st	2nd	3rd	4th
MIA	56	19	25	0
NE	9	55	36	0
NYJ	35	26	39	0
BUF	0	0	0	100
AFC NORTH
Team	1st	2nd	3rd	4th
PIT	100	0	0	0
BAL	0	100	0	0
CLE	0	0	78	22
CIN	0	0	22	78
AFC SOUTH
Team	1st	2nd	3rd	4th
TEN	100	0	0	0
IND	0	100	0	0
HOU	0	0	98	2
JAX	0	0	2	98
AFC WEST
Team	1st	2nd	3rd	4th
DEN	90	10	0	0
SD	10	90	0	0
OAK	0	0	93	7
KC	0	0	7	93
NFC EAST
Team	1st	2nd	3rd	4th
NYG	100	0	0	0
DAL	0	50	41	10
PHI	0	50	40	10
WAS	0	0	20	80
NFC NORTH
Team	1st	2nd	3rd	4th
MIN	84	16	0	0
CHI	16	84	0	0
GB	0	0	100	0
DET	0	0	0	100
NFC SOUTH
Team	1st	2nd	3rd	4th
CAR	80	12	8	0
TB	11	49	41	0
ATL	10	39	52	0
NO	0	0	0	100
NFC WEST
Team	1st	2nd	3rd	4th
ARI	100	0	0	0
SF	0	98	2	0
SEA	0	2	97	1
STL	0	0	1	99

AFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
TEN	52	48	0	0	0	0	100
PIT	48	52	0	0	0	0	100
IND	0	0	0	0	92	7	99
DEN	0	0	2	88	0	0	90
BAL	0	0	0	0	6	70	76
MIA	0	0	56	0	0	2	58
NYJ	0	0	33	2	0	5	40
NE	0	0	9	0	2	16	27
SD	0	0	0	10	0	0	10
NFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
NYG	50	44	6	0	0	0	100
ARI	0	0	11	89	0	0	100
CAR	50	27	2	0	10	9	99
MIN	0	8	65	11	0	0	84
ATL	0	10	0	0	30	29	69
TB	0	11	0	0	31	23	65
DAL	0	0	0	0	23	18	41
PHI	0	0	0	0	6	20	26
CHI	0	0	16	0	0	0	16

Week 16 Game Probabilities

Win probabilities for week 16 NFL games are listed below. The probabilities are based on an efficiency win model explained here and here with some modifications. 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.

Pwin	GAME	Pwin
0.65	IND at JAX	0.35
0.40	BAL at DAL	0.60
0.42	CIN at CLE	0.58
0.90	NO at DET	0.10
0.77	MIA at KC	0.23
0.67	ATL at MIN	0.33
0.48	ARI at NE	0.52
0.50	CAR at NYG	0.50
0.67	SF at STL	0.33
0.51	PIT at TEN	0.49
0.53	PHI at WAS	0.47
0.27	BUF at DEN	0.73
0.60	HOU at OAK	0.40
0.57	NYJ at SEA	0.43
0.45	SD at TB	0.55
0.32	GB at CHI	0.68

Pro Bowler? Really?

I know I'm piling on Brett Favre lately, but he was named to the 2009 Pro Bowl today.Did he deserve it?

Well, no. He didn't. I wouldn't be writing this if I thought he did. His rankings in various stats among the AFC's 16 top qualifying quarterbacks are:

Passer Rating: 7th
Total Yards: 7th
Yards Per Attempt: 9th
Interceptions: 1st
Int Per Attempt: 1st

The one thing Favre can claim is TD passes. He ranks 4th in TD passes, and 2nd in TDs Per Attempt. But since I'm an incurable cynic, these stats tell me he's calling his own number near the goal line!

Some might argue that Favre should go to the Pro Bowl because the Jets' resurgence in 2008 can be credited to him, but I'd disagree. Some might say, "Dude, those are just stats." But every time someone says "just stats," I hear "just facts."

Luckiest Teams Through Week 15

There's this guy in my fantasy league who finished 10th out of 10 teams in terms of total points scored--dead last. But somehow he finished with a 7-6-1 record and nabbed the 4th and final playoff spot. This past weekend was our league's semi-final round, and he won his game by...1 point. A meaningless sack of a 3rd string QB in the final seconds of a 20-point game on Monday night game put his team over the top. Despite managing the very worst team in our league, one more lucky win means he'll takes home all the marbles.

But that's just fantasy football. What about real football? Can a team get lucky like that, where every punt stays out of the end zone, every loose ball bounces into their hands, and Ed Hochuli rules their QB's obvious fumble was an incomplete pass?

Explanation

This is one of my most fun stats--"team luck"--and last year it got plenty of criticism. I estimate team luck by using my efficiency regression model to calculate each team's expected wins--how many wins a team can normally expect, on average, given their actual performance in offensive and defensive running, passing, turnovers, and penalties. The difference between the expected wins and actual wins is what I loosely call team luck.

There are plenty of things my model does not consider, special teams being the most prominent. But special teams plays are the most random events in the sport, save for the coin flip. Luck is a punt that lands on the 5 and skids into the end zone for touchback instead of bouncing into the air and being downed at the 1. A kick or punt return for a touchdown certainly requires skill, but when the kick return (or missed field goal or anything else) occurs means everything.

A kick return when a team is already ahead by 20 points doesn't mean much, but when a team is behind by 3 in the 4th quarter, it means the game. Teams and players can't control when those events occur, or else they'd save them up for when they matter most. So in a very substantial way, they are luck, at least when it comes to deciding game outcomes. (For a more thorough discussion, see this essay on NFL luck.)

Lucky and Unlucky Teams

So far this year, the luckiest teams are the Jets, Patriots, Titans, Bills, Broncos and Vikings. Basically, these teams have 2 or more wins more than their stats suggest they should have.

On the other side of the coin are the Packers, the Chargers, the Saints, and the Chiefs. Of particular interest are the Lions, who are on the verge of the first 0-16 season. Normally, a team with their stats would have won a game or two by now.

One interesting thing about this analysis is that Miami's turnaround isn't so miraculous as it seems. Despite their 1-15 season in 2007, their stats indicated they normally would have won about 5 or 6 games, but were unlucky in the extreme. I think their improvement this year is real and substantial, but not as drastic as their record indicates. According to their stats they should be a 9- or 10-win team this year, which would be a 4 or 5 game improvement rather than a 9 game improvement.

The Falcons showed the same pattern, but to a lesser degree. Although they notched only 4 wins, they "should" have won 6. That shouldn't take anything away from their comeback this year, though. In 2008, they continue to be one of the more unlucky teams.

The Charmed One

Looking back at 2007's luckiest/unluckiest teams, one thing stood out as most remarkable. Last year Green Bay was the #1 luckiest team while the Jets were near the bottom at #31. But this year the Packers are the least lucky team while the Jets are the luckiest. So what changed? According some minor media reports, I vaguely recall an obscure quarterback was traded from the Packers to the Jets...

"Ahah!" my brother-in-law and noted Favre lover is saying right now. Your model's residual doesn't measure luck. It measures Favre-ness. You know, that intangible mumbo-jumbo that John Madden blathers on about every other Sunday night.

But my model does include passing, interceptions, and sacks. What it doesn't include are things like J.P. Losman fumbling on a 2nd and 5 on his own 16-yard line with a 3-point lead and less than 2 minutes remaining...which was instantly returned for a game-winning touchdown. If you can make a case that Brett somehow intangibled his way to that win, then I'll gladly stand corrected.


The Full List

RANK	TEAM	GWP	Curr W	Exp W	Luck
1	NYJ	0.42	9	6.0	+3.0
2	NE	0.45	9	6.5	+2.5
3	TEN	0.68	12	9.7	+2.3
4	BUF	0.27	6	4.0	+2.0
5	DEN	0.42	8	6.0	+2.0
6	MIN	0.49	9	7.1	+1.9
7	NYG	0.66	11	9.4	+1.6
8	DAL	0.54	9	7.8	+1.2
9	SF	0.26	5	3.8	+1.2
10	ARI	0.48	8	6.9	+1.1
11	IND	0.63	10	9.0	+1.0
12	CAR	0.72	11	10.2	+0.8
13	MIA	0.59	9	8.4	+0.6
14	PIT	0.74	11	10.5	+0.5
15	HOU	0.47	7	6.7	+0.3
16	STL	0.12	2	1.8	+0.2
17	SEA	0.24	3	3.5	-0.5
18	BAL	0.68	9	9.6	-0.6
19	TB	0.68	9	9.7	-0.7
20	JAX	0.40	5	5.7	-0.7
21	OAK	0.26	3	3.8	-0.8
22	CHI	0.62	8	8.8	-0.8
23	CIN	0.24	2.5	3.5	-1.0
24	ATL	0.72	9	10.2	-1.2
25	CLE	0.37	4	5.3	-1.3
26	WAS	0.59	7	8.4	-1.4
27	PHI	0.71	8.5	10.1	-1.6
28	DET	0.11	0	1.7	-1.7
29	KC	0.27	2	4.0	-2.0
30	NO	0.65	7	9.3	-2.3
31	SD	0.60	6	8.5	-2.5
32	GB	0.55	5	7.9	-2.9

Week 15 Efficiency Rankings

The ratings are listed below in terms of generic win 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, which can be considered to-date strength of schedule, and all ratings include adjustments for opponent strength.

Offensive rank (ORANK) is offensive generic win probability which is based on each team's offensive efficiency stats only. In other words, it's the team's GWP assuming it had a league-average defense. DRANK is is a team's generic win probability rank assuming it had a league-average offense.

GWP is based on a logistic regression model applied to current team stats. The model includes offensive and defensive passing and running efficiency, offensive turnover rates, and team penalty rates. A full explanation of the methodology can be found here. This year, however, I've made one important change based on research that strongly indicates that defensive interception rates are highly random and not consistent throughout the year. Accordingly, I've removed them from the model and updated the weights of the remaining stats.

RANK	TEAM	LAST WK	GWP	Opp GWP	O RANK	D RANK
1	PIT	3	0.76	0.53	20	1
2	CAR	5	0.76	0.54	7	5
3	ATL	1	0.76	0.56	1	17
4	PHI	2	0.75	0.52	12	4
5	NYG	4	0.69	0.52	6	10
6	NO	9	0.68	0.56	3	13
7	TEN	7	0.68	0.43	16	2
8	SD	6	0.67	0.51	5	12
9	MIA	11	0.66	0.43	2	19
10	WAS	8	0.65	0.51	11	11
11	DAL	15	0.64	0.52	14	7
12	TB	12	0.63	0.56	18	6
13	BAL	10	0.63	0.52	19	3
14	IND	13	0.62	0.48	9	18
15	CHI	16	0.61	0.51	21	8
16	ARI	14	0.56	0.51	4	22
17	GB	17	0.52	0.55	13	20
18	MIN	20	0.51	0.50	24	9
19	DEN	18	0.49	0.49	8	27
20	NE	19	0.48	0.46	15	24
21	NYJ	21	0.47	0.43	23	14
22	HOU	22	0.44	0.51	10	28
23	JAX	24	0.38	0.48	17	25
24	BUF	23	0.34	0.43	28	21
25	SEA	25	0.33	0.48	27	16
26	SF	26	0.32	0.49	25	26
27	KC	29	0.29	0.52	22	30
28	CIN	30	0.28	0.60	31	23
29	CLE	28	0.27	0.57	26	29
30	OAK	27	0.27	0.55	32	15
31	STL	31	0.15	0.53	29	31
32	DET	32	0.14	0.57	30	32

To-date efficiency stats below. As always, click on the headers to sort.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.2	3.3	0.023	0.025	6.4	4.0	0.025	0.38
ATL	7.6	4.2	0.023	0.015	6.1	4.9	0.021	0.29
BAL	5.8	3.8	0.031	0.025	5.1	3.4	0.048	0.41
BUF	5.9	4.2	0.035	0.036	6.2	4.2	0.021	0.29
CAR	6.9	4.8	0.032	0.015	5.6	4.1	0.023	0.35
CHI	5.5	3.9	0.026	0.016	5.7	3.5	0.038	0.31
CIN	4.2	3.4	0.032	0.027	6.6	3.9	0.018	0.30
CLE	5.0	3.9	0.031	0.025	7.1	4.5	0.054	0.32
DAL	6.9	4.3	0.037	0.030	5.3	4.0	0.017	0.49
DEN	7.1	4.5	0.029	0.020	6.9	4.9	0.013	0.34
DET	5.3	3.7	0.034	0.040	7.8	4.9	0.011	0.38
GB	6.5	4.0	0.025	0.023	6.1	4.8	0.040	0.50
HOU	7.2	4.4	0.039	0.029	6.9	4.5	0.029	0.34
IND	6.6	3.4	0.023	0.010	6.0	4.2	0.033	0.32
JAX	5.7	4.2	0.021	0.018	6.8	4.1	0.032	0.43
KC	5.3	4.7	0.028	0.020	7.3	4.9	0.026	0.30
MIA	7.1	4.1	0.014	0.016	6.1	3.9	0.026	0.37
MIN	5.9	4.5	0.041	0.023	6.0	3.2	0.025	0.38
NE	5.9	4.5	0.023	0.018	6.7	4.2	0.031	0.24
NO	7.6	3.8	0.031	0.019	6.3	4.0	0.027	0.41
NYG	6.1	4.8	0.023	0.020	5.5	3.9	0.035	0.45
NYJ	6.0	4.8	0.039	0.024	6.1	3.7	0.027	0.28
OAK	4.9	4.2	0.027	0.037	6.4	4.7	0.035	0.45
PHI	6.3	4.0	0.030	0.014	5.2	3.4	0.031	0.34
PIT	5.8	3.6	0.027	0.023	4.3	3.2	0.037	0.42
SD	7.5	3.8	0.026	0.020	6.2	3.9	0.021	0.38
SF	6.0	3.9	0.034	0.042	6.3	3.8	0.023	0.40
SEA	5.0	4.4	0.032	0.016	6.9	4.1	0.012	0.28
STL	5.2	3.9	0.040	0.027	7.6	4.7	0.018	0.39
TB	6.0	4.0	0.020	0.020	5.7	4.3	0.050	0.41
TEN	6.2	4.4	0.022	0.020	4.9	3.7	0.036	0.44
WAS	5.7	4.5	0.013	0.021	5.8	3.8	0.028	0.36
Avg	6.1	4.1	0.029	0.023	6.2	4.1	0.028	0.37

Weekly Roundup

I'm starting a new regular feature here at Advanced NFL Stats. There are lots of great sites that feature interesting football and sports research. So each week or so, I'll try to provide a list of links to articles that catch my eye along with some of my own thoughts. If you have any suggestions or additions, send 'em in.

I'll start of with this article from Football Outsiders. Bill Connelly improves his Equivalent Points concept for college football. It's a lot like Expected Points, but only considers current-drive points. It's unusual for FO because the methodology is actually (somewhat) described.

The Pro-Football-Reference.com blog has been unusually active this week. First, Doug Drinen discusses the computer rankings that go into the BCS formula, particularly Jeff Sagarin's Elo system. The article does a good job of explaining how such a system works in layman's terms. I wrote about the Elo and Sagarin systems earlier this year.

Chase Stuart looks at how rookie QBs Matt Ryan and Joe Flacco stand up in historical context. Basically, Ryan's year is freakish. I wish the networks would give the Falcons a national game. I haven't even been able to watch him play. Here is my own take from last week on Ryan and Flacco and their season-long improvement.

Doug also looks at season sweeps and splits. Which is more common: a home-and-home split or a 2-game sweep? In the end, the author is really looking for whether teams "have the number" of an opponent, or whether sweeps are just a part of the natural variation in team strengths. I strongly disagree with the conclusion--I do think it's an interesting post.

Turning to other sports, Hawerchuck looks at Malcolm Gladwell's observation about NHL players and their birth month distribution. (More discussion here.) Hockey players in Canada are grouped by birth year, so throughout their childhood development the early-month-born players are the bigger, faster, and stronger ones. (Remember when you were 11? 12-year olds seemed like giants.) By the time players get to the pros, fewer late-month players are around. But the ones that do make it tend to outperform the others. In case you're curious, I took a quick look at this for NFL players and found no differences in birth-months on current rosters.

I can sympathize. I'm a late-birth-month guy who had to play rec-league basketball against older guys, usually a grade or two ahead of me. I held my own for the most part, but was usually at a big disadvantage that took a lot of the fun out of it. But when I played against my friends and classmates, I was a half-step ahead of them--kind of like swinging with two bats I guess. (End of therapy session.)

Speaking of basketball, Phil Birnbaum discusses a study that finds that traditional, qualitative measures of NBA player ability are at least as good as the advanced quantitative measures. Phil points out that the more scientific quantitative measures don't consider the defense half of the sport very well--or at all. But quantitative analysis is still useful, at least for the offensive side of the sport. This should give us pause in football circles. Team-level analysis is one thing, but the player-level analysis in football has to be taken with a grain of salt. We can talk about how "good a year" a player is having, but only part of that performance belongs to the player himself. And individual defensive performance is, for now, nearly impossible to measure quantitatively.

Lots to chew on. Hope you like the roundup. Again, if you have any suggestions for additions, even your own stuff, post it below.

Super Bowl Probabilities

The updated Super Bowl probabilities are out. Could we see a match-up of two of the league's most venerable franchises? The front runners are:

Team	Conf Champ	SB Champ
PIT	39.4	20.9
NYG	37.7	20.5
TEN	37.0	17.6
CAR	28.6	15.5
ATL	10.9	6.4
BAL	8.4	3.3
TB	7.9	3.7
IND	6.4	2.5
ARI	5.9	2.6
MIA	4.5	1.6
PHI	3.5	1.8
DEN	2.5	0.7
CHI	2.2	0.9
WAS	1.7	0.7
SD	1.0	0.6

Thanks to NFL-Forecast.com.

Bears' Fake Punt vs. Saints?

With a 7-point lead and 7:44 left in the 4th quarter, the Bears faced a 4th and 4 from their own 47 yard line. Chicago faked the punt as punter Brad Maynard threw to receiver Adrian Peterson. It was close to being caught, but ultimately ruled incomplete. Was this a good call by head coach Lovie Smith?

I don’t think we have enough information to have a solid idea how likely that exact play would succeed in that particular game situation. But what I will do is estimate how often a fake punt would have to be successful for Smith’s play call to make sense.

With a 7-point lead and about 7 minutes to go in the game, a team that punts from its own 47 can expect a win probability (WP) of 0.79. That is, they’d win about 79% of the time. A team that turns the ball over on downs at the 47 would have a WP of 0.57.

If a fake punt-pass were successful, a 1st down and 10 on the opponent’s 40 yard line would yield a WP of about 0.95. A completed pass might have put the game on ice.

The break-even probability of success (P_s) for the fake punt to make it worth-while can be calculated as follows:

0.79 = 0.95*P_s + 0.57*(1-P_s)
0.38*P_s = 0.22
P_s = 0.58

So the fake punt play would have to be successful at least 58% of the time to make the risk worth it. That’s pretty high, and it’s only the break-even point. So I doubt it was a good call, but can't say conclusively.

This analysis doesn’t factor in the potency of the Saint’s offense, which was what I’d bet Lovie Smith was thinking of when he made the call. But that consideration works on both sides of the equation. Failing to convert for the 1st down would have handed the ball to the Saints offense in very favorable field position. Plus, the Saints would have had a fair chance of scoring even with the punt. Chicago’s defense isn’t exactly filled with slouches, and had been holding their own all game. I can’t be certain it’s a complete wash, but it’s pretty close.

Playoff Probabilities Week 14

Courtesy of Chris at NFL-Forecast.com, here are the latest playoff probabilities for each team.

These are calculated using the NFL-Forecast software mini-app that runs thousands of simulated seasons. The outcomes are based on game-by-game probabilities with every crazy tie-breaking scenario factored in. Chris has used the probabilities from Advanced NFL Stats as his default game probabilities for the past two seasons.

There are two tables below. The first lists the probability that each team will finish in each place in their division. The second table lists the overall playoff probabilities, broken down by seed.

AFC EAST
Team	1st	2nd	3rd	4th
Miami	61	22	17	0
New England	9	51	39	1
NY Jets	30	27	41	2
Buffalo	0	0	3	97
AFC NORTH
Team	1st	2nd	3rd	4th
Pittsburgh	87	13	0	0
Baltimore	13	87	0	0
Cleveland	0	0	97	3
Cincinnati	0	0	3	97
AFC SOUTH
Team	1st	2nd	3rd	4th
Tennessee	100	0	0	0
Indianapolis	0	100	0	0
Houston	0	0	95	5
Jacksonville	0	0	5	95
AFC WEST
Team	1st	2nd	3rd	4th
Denver	92	8	0	0
San Diego	8	92	0	0
Oakland	0	0	86	14
Kansas City	0	0	14	86
NFC EAST
Team	1st	2nd	3rd	4th
NY Giants	100	0	0	0
Philadelphia	0	48	31	20
Washington	0	30	30	40
Dallas	0	21	38	40
NFC NORTH
Team	1st	2nd	3rd	4th
Minnesota	62	36	2	0
Chicago	33	53	14	0
Green Bay	4	12	84	0
Detroit	0	0	0	100
NFC SOUTH
Team	1st	2nd	3rd	4th
Carolina	64	27	9	0
Tampa Bay	24	41	34	1
Atlanta	12	32	43	14
New Orleans	0	0	14	85
NFC WEST
Team	1st	2nd	3rd	4th
Arizona	100	0	0	0
San Francisco	0	98	2	0
Seattle	0	1	76	22
St Louis	0	0	22	78

AFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
Tennessee	81	19	0	0	0	0	100
Pittsburgh	19	67	0	0	2	11	100
Indianapolis	0	0	0	0	84	14	98
Baltimore	0	13	0	0	13	67	94
Denver	0	0	6	86	0	0	92
Miami	0	0	60	1	0	1	62
NY Jets	0	0	25	5	0	2	32
New England	0	0	9	0	0	5	14
San Diego	0	0	0	8	0	0	8
Buffalo	0	0	0	0	0	0	0
Cincinnati	0	0	0	0	0	0	0
Cleveland	0	0	0	0	0	0	0
Houston	0	0	0	0	0	0	0
Jacksonville	0	0	0	0	0	0	0
Kansas City	0	0	0	0	0	0	0
Oakland	0	0	0	0	0	0	0
NFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
NY Giants	72	27	0	0	0	0	100
Arizona	0	4	73	23	0	0	100
Carolina	27	36	0	0	24	9	96
Tampa Bay	0	20	3	0	28	19	72
Minnesota	0	1	18	43	0	0	62
Atlanta	0	12	0	0	28	16	57
Chicago	0	0	4	29	0	0	33
Philadelphia	0	0	0	0	6	22	29
Washington	0	0	0	0	5	18	23
Dallas	0	0	0	0	8	13	21
Green Bay	0	0	0	4	0	0	4
New Orleans	0	0	0	0	0	2	2
Detroit	0	0	0	0	0	0	0
St Louis	0	0	0	0	0	0	0
San Francisco	0	0	0	0	0	0	0
Seattle	0	0	0	0	0	0	0

Week 15 Game Probabilities

Win probabilities for week 15 NFL games are listed below. The probabilities are based on an efficiency win model explained here and here with some modifications. 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.

Pwin	GAME	Pwin
0.48	NO at CHI	0.52
0.27	TB at ATL	0.73
0.53	PIT at BAL	0.47
0.20	DEN at CAR	0.80
0.83	WAS at CIN	0.17
0.69	TEN at HOU	0.31
0.05	DET at IND	0.95
0.57	GB at JAX	0.43
0.83	SD at KC	0.17
0.16	SF at MIA	0.84
0.37	BUF at NYJ	0.63
0.65	SEA at STL	0.35
0.26	MIN at ARI	0.74
0.61	NE at OAK	0.39
0.56	NYG at DAL	0.44
0.08	CLE at PHI	0.92

Week 14 Efficiency Rankings

The ratings are listed below in terms of generic win 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, which can be considered to-date strength of schedule, and all ratings include adjustments for opponent strength.

Offensive rank (ORANK) is offensive generic win probability which is based on each team's offensive efficiency stats only. In other words, it's the team's GWP assuming it had a league-average defense. DRANK is is a team's generic win probability rank assuming it had a league-average offense.

GWP is based on a logistic regression model applied to current team stats. The model includes offensive and defensive passing and running efficiency, offensive turnover rates, and team penalty rates. A full explanation of the methodology can be found here. This year, however, I've made one important change based on research that strongly indicates that defensive interception rates are highly random and not consistent throughout the year. Accordingly, I've removed them from the model and updated the weights of the remaining stats.

RANK	TEAM	LAST WK	GWP	Opp GWP	O RANK	D RANK
1	ATL	1	0.76	0.55	1	20
2	PHI	3	0.76	0.55	7	3
3	PIT	6	0.75	0.52	21	1
4	NYG	2	0.74	0.52	3	9
5	CAR	5	0.74	0.54	10	6
6	SD	4	0.72	0.52	2	13
7	TEN	8	0.70	0.41	14	2
8	WAS	7	0.69	0.54	8	11
9	NO	9	0.67	0.56	4	17
10	BAL	13	0.65	0.51	19	4
11	MIA	10	0.64	0.45	6	22
12	TB	12	0.64	0.54	18	5
13	IND	15	0.63	0.50	11	14
14	ARI	11	0.62	0.51	5	15
15	DAL	14	0.61	0.51	15	8
16	CHI	16	0.61	0.49	20	7
17	GB	19	0.52	0.55	13	21
18	DEN	17	0.50	0.46	9	27
19	NE	20	0.47	0.48	16	24
20	MIN	21	0.46	0.49	26	10
21	NYJ	18	0.44	0.44	24	18
22	HOU	23	0.41	0.48	12	29
23	BUF	22	0.40	0.42	27	19
24	JAX	24	0.36	0.48	17	26
25	SEA	26	0.32	0.48	29	16
26	SF	29	0.32	0.53	25	25
27	OAK	25	0.27	0.56	32	12
28	CLE	27	0.27	0.56	23	28
29	KC	28	0.27	0.52	22	30
30	CIN	30	0.24	0.59	31	23
31	STL	31	0.15	0.56	28	31
32	DET	32	0.12	0.56	30	32

To-date efficiency stats below. As always, click on the headers to sort.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.4	3.3	0.024	0.024	6.3	3.8	0.026	0.37
ATL	7.5	4.2	0.019	0.012	6.2	4.9	0.020	0.30
BAL	6.1	3.8	0.028	0.025	5.1	3.4	0.053	0.42
BUF	6.2	4.1	0.031	0.033	6.1	4.0	0.017	0.29
CAR	6.7	4.8	0.032	0.016	5.7	4.0	0.022	0.33
CHI	5.7	4.0	0.024	0.016	5.7	3.5	0.038	0.32
CIN	4.1	3.4	0.034	0.029	6.7	4.0	0.019	0.30
CLE	5.0	3.9	0.029	0.025	7.1	4.5	0.051	0.34
DAL	6.9	4.2	0.039	0.029	5.4	3.9	0.014	0.45
DEN	7.3	4.5	0.028	0.020	6.7	4.9	0.012	0.35
DET	5.2	3.7	0.037	0.040	7.7	5.0	0.012	0.38
GB	6.5	4.1	0.025	0.023	6.1	4.8	0.043	0.51
HOU	7.3	4.4	0.043	0.028	7.0	4.6	0.029	0.29
IND	6.4	3.4	0.025	0.010	5.9	4.2	0.036	0.32
JAX	5.6	4.1	0.023	0.018	6.7	4.2	0.032	0.42
KC	5.2	4.8	0.028	0.022	7.4	5.0	0.027	0.31
MIA	7.1	4.1	0.015	0.017	6.3	3.9	0.028	0.36
MIN	5.8	4.4	0.043	0.024	6.1	3.2	0.026	0.39
NE	5.9	4.3	0.022	0.019	6.7	4.1	0.031	0.25
NO	7.8	3.8	0.030	0.021	6.5	4.1	0.025	0.39
NYG	6.4	4.9	0.020	0.018	5.4	3.9	0.037	0.45
NYJ	5.9	4.7	0.037	0.025	6.4	3.5	0.023	0.28
OAK	4.7	4.2	0.026	0.037	6.4	4.5	0.035	0.46
PHI	6.2	4.0	0.026	0.015	5.3	3.5	0.028	0.36
PIT	5.9	3.6	0.030	0.021	4.4	3.2	0.035	0.42
SD	7.7	3.8	0.026	0.015	6.2	3.9	0.020	0.36
SF	6.2	3.9	0.038	0.046	6.2	3.8	0.024	0.41
SEA	4.8	4.4	0.034	0.016	7.0	4.0	0.013	0.28
STL	5.1	3.8	0.043	0.026	7.6	4.7	0.019	0.39
TB	6.0	4.0	0.020	0.022	5.6	4.3	0.048	0.38
TEN	6.3	4.4	0.021	0.018	4.8	3.7	0.039	0.44
WAS	5.7	4.6	0.014	0.018	5.7	3.8	0.030	0.34
Avg	6.2	4.1	0.029	0.023	6.2	4.1	0.029	0.37

No Man's Land

I recently looked at how often offenses go for it on 4th down. But I think most 4th down conversion attempts are a function of how desperate a team is in the final minutes of a game. I'm convinced teams should be going for it far more often than as a matter of doctrine, not just as a last ditch tactic when trailing late in the 4th quarter.

In this article I'll look at tendencies on 4th down. To filter out the element of desperation and only look at how coaches and coordinators make 4th down decisions as a matter of routine, I'll look at first quarter situations only. Except in the rarest of games, teams are not outscored so badly that they become desperate before the end of the 1st quarter. First quarter situations would therefore reveal the baseline NFL doctrine for 4th down decisions.

The graph below illustrates how frequently teams punt, attempt a field goal, or go for it on 4th down situations at each yard line on the field.

The 34 yard line is the "no man's land" of 4th down decision-making--too close to punt, too far to try a field goal. Outside of the 34, offenses tend to punt. Inside the 34, teams tend to kick field goals. The 34 yard line is where field goals become a 50/50 proposition. Outside the 34, the success rate for field goals drops off very quickly. The 34 is also where offenses will go for the conversion most often, nearly 40% of the time.

It seems that coaches aren't really going for the conversion because they believe it works out better for them in the long run, but because they don't have a viable kicking option. They'll go for it only when the safe and conventional alternative is taken away.

I think that graph is pretty cool for some reason. It says a mouthful. But it leaves out perhaps the most important consideration on 4th down--to go distance. There are a lot of variables, so they all can't go on a single graph. The following graphs break out the play-calls by to go distance. Passes and runs are broken out too. Again, these are all for the 1st quarter.

The Super Bowl Champion Will Be...

The NFL-Forecast.com software now includes predictions through the Super Bowl. Front runners are:

Giants 23.98%
Pittsburgh 18.12%
Tennessee 17.8%
Atlanta 12.4%
Carolina 9.9%

No one else is above 5%.

The Learning Curve

One of the storylines this season is the solid performance of two rookie quarterbacks. Ravens QB Joe Flacco and Falcons QB Matt Ryan are having surprisingly good seasons for any QB, much less a rookie. Ryan has done particularly well, especially given the state of the Falcons in 2007.

Another storyline is how well Matt Cassel of the Patriots has stepped into Tom Brady’s shoes. Prior to a tough outing against the NFL’s best pass defense last week, he threw for 400 yards in two consecutive games. Although not a rookie, Cassel had only 39 attempts in his previous three seasons as a back up.

Cassel is likely to end up as a starter on another team next year, but fans of the Ravens and Falcons can be optimistic about the prospects for their teams in coming years. Previous research has looked at the year-to-year trends of QB performance and found that the big jump comes in a passer’s second year of playing. But what about during the rookie year?

Some prominent QBs showed improvement in the second half of their first year playing. In 1998 Peyton Manning’s yards per attempt (YPA) went from 6.4 to 6.6 from his first 8 games to his second 8. Ben Roethlisberger went from 8.5 to 9.3(!) YPA. Tom Brady went from 6.6 to 7.2 YPA. (These are gross YPA, not accounting for sacks or interceptions).

In this article, I’ll look at Flacco’s and Ryan’s game-by-game performance through week 12 to see if we can detect an upward trend. I’ll also look at Cassel’s numbers to see if he exhibits a similar trend.

Performance is measured by adjusted YPA, which is passing yards – 40 yards per interception + 10 yards per touchdown. The 40 yard penalty for interceptions is a commonly accepted adjustment, partly because an interception precludes a punt. The 10 yard increase for each touchdown is not a bonus, but simply accounts for the depth of the end zone and the difficulty of most goal line passes due to the compression of the field.

Game-by-game performance can be affected by a variety of factors such as team injuries or weather. Random variation due to small sample size and the strength of opponent pass defenses are probably two of the most important. To account for the small sample size of an individual game, I’ll use a 4-game moving average to chart performance. To account for opponent strength I’ll factor-in defensive pass efficiency. For every yard per attempt more than the league average an opponent yields, I’ll subtract that amount from the individual game performance of each QB. For every yard under the average, I’ll add that amount. In effect, the QBs get extra credit for facing tougher pass defenses, and are penalized when facing weaker ones.

So far this season, Flacco and Ryan have faced slightly above average pass defenses, each averaging -0.1 YPA better than average (lower is better). Cassel has faced slightly weaker opponents, at +0.2 YPA worse than average.

The graph below plots all three QB’s performance in “defense-adjusted” adjusted YPA by game. Game-to-game variations make it very hard to discern any trends, so you might want to skip down to the next graph.


This next graph is a four-game moving average, and it shows the trends clearly.


For comparison, the NFL average going into week 14 is 5.8 adj. YPA. Ryan’s season average is 7.6, Flacco’s is 6.3, and Cassel’s is 6.2.

Ryan started the season with the strongest performance, but it’s Flacco who has shown the most improvement. Perhaps that shouldn't be too surprising given Flacco's Division 1-AA pedigree. Cassel, who had already been in the league for three years, started off somewhere between Ryan and Flacco, but has shown a shallower rate of improvement.

I’m not going to draw any concrete conclusions, but it is interesting that all three QBs do show a similar improving trend.

Playoff Probabilities Week 13

Courtesy of Chris at NFL-Forecast.com, here are the latest playoff probabilities for each team.

These are calculated using the NFL-Forecast software mini-app that runs thousands of simulated seasons. The outcomes are based on game-by-game probabilities with every crazy tie-breaking scenario factored in. Chris has used the probabilities from Advanced NFL Stats as his default game probabilities for the past two seasons.

There are two tables below. The first lists the probability that each team will finish in each place in their division. The second table lists the overall playoff probabilities, broken down by seed.

AFC EAST
Team	1st	2nd	3rd	4th
NY Jets	54	27	17	2
Miami	37	32	28	3
New England	9	38	47	7
Buffalo	0	3	9	88
AFC NORTH
Team	1st	2nd	3rd	4th
Pittsburgh	89	11	0	0
Baltimore	11	89	0	0
Cleveland	0	0	97	3
Cincinnati	0	0	3	97
AFC SOUTH
Team	1st	2nd	3rd	4th
Tennessee	100	0	0	0
Indianapolis	0	100	0	0
Houston	0	0	79	21
Jacksonville	0	0	21	79
AFC WEST
Team	1st	2nd	3rd	4th
Denver	87	13	0	0
San Diego	13	86	1	0
Oakland	0	1	80	19
Kansas City	0	0	19	81
NFC EAST
Team	1st	2nd	3rd	4th
NY Giants	100	0	0	0
Washington	0	50	32	18
Dallas	0	32	41	27
Philadelphia	0	18	27	55
NFC NORTH
Team	1st	2nd	3rd	4th
Minnesota	54	39	7	0
Chicago	28	36	35	0
Green Bay	18	25	58	0
Detroit	0	0	0	100
NFC SOUTH
Team	1st	2nd	3rd	4th
Tampa Bay	34	28	37	0
Carolina	35	34	29	3
Atlanta	31	37	26	6
New Orleans	0	1	8	91
NFC WEST
Team	1st	2nd	3rd	4th
Arizona	100	0	0	0
San Francisco	0	78	20	2
Seattle	0	16	65	20
St Louis	0	6	15	79

AFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
Tennessee	85	13	1	0	0	0	100
Pittsburgh	15	68	6	0	2	8	99
Indianapolis	0	0	0	0	85	11	97
Denver	0	0	5	82	0	0	87
Baltimore	0	10	1	0	10	55	76
NY Jets	0	7	43	4	2	10	65
Miami	0	1	36	0	0	7	44
New England	0	0	8	0	1	9	18
San Diego	0	0	0	13	0	0	13
Buffalo	0	0	0	0	0	1	1
Oakland	0	0	0	0	0	0	0
Houston	0	0	0	0	0	0	0
Cincinnati	0	0	0	0	0	0	0
Cleveland	0	0	0	0	0	0	0
Jacksonville	0	0	0	0	0	0	0
Kansas City	0	0	0	0	0	0	0
NFC Percent Probability Playoff Seeding
Team	1st	2nd	3rd	4th	5th	6th	Total
Arizona	0	3	76	21	0	0	100
NY Giants	84	15	0	0	0	0	100
Carolina	12	23	0	0	27	18	80
Atlanta	1	28	2	0	27	16	74
Tampa Bay	3	30	1	0	18	21	73
Minnesota	0	1	17	37	0	0	54
Washington	0	0	0	0	15	21	37
Chicago	0	0	3	25	0	0	29
Dallas	0	0	0	0	11	15	27
Green Bay	0	0	0	17	0	0	18
Philadelphia	0	0	0	0	1	7	8
New Orleans	0	0	0	0	0	1	1
Detroit	0	0	0	0	0	0	0
St Louis	0	0	0	0	0	0	0
San Francisco	0	0	0	0	0	0	0
Seattle	0	0	0	0	0	0	0

Game Probabilities Week 14

Win probabilities for week 14 NFL games are listed below. The probabilities are based on an efficiency win model explained here and here with some modifications. 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.

Pwin	GAME	Pwin
0.09	OAK at SD	0.91
0.54	WAS at BAL	0.46
0.19	JAX at CHI	0.81
0.87	MIN at DET	0.13
0.31	HOU at GB	0.69
0.10	CIN at IND	0.90
0.53	ATL at NO	0.47
0.39	PHI at NYG	0.61
0.09	CLE at TEN	0.91
0.62	MIA at BUF	0.38
0.16	KC at DEN	0.84
0.65	NYJ at SF	0.35
0.04	STL at ARI	0.96
0.29	DAL at PIT	0.71
0.69	NE at SEA	0.31
0.27	TB at CAR	0.73

Change extra points?

Reader Eddy Elfenbein makes a great point about extra points. I also like the idea of narrowing the field goal posts, something Bill Cowher advocated last Sunday. It would not only make extra points less certain, it would make fourth down conversion attempts more common and make overtime less susceptible to the 'lose the coin toss, never touch the ball' phenomenon.

Ideas like these really aren't that revolutionary. They would just be returning the game to its 'natural' balance. Kickers have become so accurate in recent decades that it has warped the game from its original intent. But NFL football has certainly evolved in many ways, and its unparalleled success makes tinkering with it a tough sell.

Extra points have become so automatic, I don't even pay attention. They're just going to be surrounded by commercials featuring that stiff Sprint CEO and the Bud Light drinkability chick in the green jersey. The only reason to watch them is when there is a possibility of a challenge on the touchdown.

Kicking field goals is such a peculiar and specialized thing. It has almost nothing to do with the rest of the sport but can be so decisive. It would be like getting extra runs in baseball by lacing up some skates and slapping a shoot-out shot after every home run.

Week 13 Efficiency Rankings

The ratings are listed below in terms of generic win 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, which can be considered to-date strength of schedule, and all ratings include adjustments for opponent strength.

Offensive rank (ORANK) is offensive generic win probability which is based on each team's offensive efficiency stats only. In other words, it's the team's GWP assuming it had a league-average defense. DRANK is is a team's generic win probability rank assuming it had a league-average offense.

GWP is based on a logistic regression model applied to current team stats. The model includes offensive and defensive passing and running efficiency, offensive turnover rates, and team penalty rates. A full explanation of the methodology can be found here. This year, however, I've made one important change based on research that strongly indicates that defensive interception rates are highly random and not consistent throughout the year. Accordingly, I've removed them from the model and updated the weights of the remaining stats.

RANK	TEAM	LAST WK	GWP	Opp GWP	O RANK	D RANK
1	ATL	4	0.79	0.54	1	19
2	NYG	6	0.76	0.49	2	8
3	CAR	3	0.75	0.54	10	6
4	PHI	7	0.74	0.52	9	5
5	SD	1	0.73	0.56	3	15
6	PIT	8	0.73	0.50	21	1
7	WAS	2	0.73	0.52	7	9
8	TEN	11	0.70	0.41	13	2
9	NO	8	0.70	0.55	4	17
10	MIA	9	0.66	0.46	5	23
11	ARI	10	0.65	0.55	6	18
12	TB	13	0.63	0.52	20	4
13	DAL	14	0.62	0.48	12	10
14	BAL	18	0.61	0.48	18	3
15	IND	15	0.58	0.52	11	14
16	CHI	12	0.58	0.50	19	11
17	DEN	19	0.53	0.49	8	26
18	GB	20	0.52	0.55	14	13
19	NYJ	16	0.51	0.46	25	16
20	NE	17	0.50	0.50	17	24
21	MIN	21	0.50	0.51	24	7
22	BUF	22	0.44	0.40	26	21
23	HOU	23	0.42	0.46	15	29
24	JAX	24	0.32	0.47	16	25
25	OAK	26	0.29	0.56	30	12
26	SEA	25	0.28	0.47	29	20
27	CLE	28	0.26	0.55	23	28
28	KC	29	0.24	0.54	22	30
29	SF	27	0.23	0.53	27	27
30	CIN	30	0.19	0.59	32	22
31	STL	31	0.11	0.56	28	31
32	DET	32	0.09	0.57	31	32

To-date efficiency stats below. As always, click on the headers to sort.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.3	3.3	0.023	0.026	6.4	3.8	0.026	0.38
ATL	7.4	4.3	0.018	0.011	6.1	4.8	0.022	0.30
BAL	6.0	3.8	0.027	0.024	5.2	3.5	0.052	0.43
BUF	6.5	4.0	0.030	0.030	6.1	4.1	0.018	0.30
CAR	6.7	4.4	0.027	0.018	5.5	4.1	0.024	0.36
CHI	5.6	4.0	0.023	0.017	5.9	3.4	0.039	0.34
CIN	4.1	3.4	0.029	0.030	6.5	4.0	0.021	0.30
CLE	5.2	4.1	0.029	0.023	7.1	4.4	0.049	0.32
DAL	7.1	4.3	0.035	0.030	5.5	4.0	0.015	0.48
DEN	7.3	4.5	0.029	0.021	6.9	4.9	0.013	0.37
DET	5.1	3.8	0.039	0.038	7.7	5.1	0.006	0.39
GB	6.3	4.0	0.024	0.024	5.7	4.8	0.045	0.51
HOU	7.1	4.4	0.045	0.026	6.9	4.5	0.029	0.29
IND	6.3	3.5	0.027	0.009	6.0	4.2	0.031	0.33
JAX	5.8	4.1	0.023	0.018	6.8	4.3	0.032	0.45
KC	5.2	4.8	0.030	0.023	7.4	5.0	0.027	0.33
MIA	7.1	4.2	0.016	0.018	6.6	3.9	0.028	0.36
MIN	5.7	4.4	0.040	0.021	6.1	3.2	0.027	0.40
NE	5.9	4.3	0.024	0.019	6.7	4.1	0.034	0.26
NO	7.9	3.6	0.032	0.023	6.3	4.2	0.025	0.40
NYG	6.5	4.9	0.022	0.017	5.4	3.9	0.040	0.46
NYJ	6.1	4.7	0.038	0.027	6.3	3.6	0.023	0.27
OAK	4.9	4.3	0.019	0.037	6.2	4.6	0.037	0.45
PHI	6.3	4.1	0.028	0.015	5.3	3.4	0.030	0.35
PIT	6.0	3.7	0.033	0.020	4.3	3.1	0.031	0.45
SD	7.5	3.9	0.028	0.017	6.4	4.0	0.015	0.35
SF	6.1	4.0	0.039	0.046	6.4	3.8	0.023	0.43
SEA	4.7	4.3	0.037	0.016	7.1	4.1	0.014	0.28
STL	5.0	3.9	0.045	0.024	7.5	4.8	0.018	0.42
TB	5.9	4.1	0.021	0.022	5.5	3.8	0.045	0.41
TEN	6.2	4.2	0.017	0.018	4.9	3.8	0.041	0.39
WAS	5.8	4.7	0.010	0.017	5.7	3.8	0.029	0.34
Avg	6.2	4.1	0.028	0.023	6.2	4.1	0.028	0.37

4th Down Trend?

As most regular readers know, I'm a big proponent of going for it on 4th down far more often than is currently practiced in the NFL. There is good research on when teams should go for it on 4th down, but what about the "currently practiced in the NFL" part? How often do offenses roll the dice? And are teams going for it more frequently?

This two-year-old article by Len Pasquarelli of ESPN.com says that as of late 2006, the trend in the NFL was that offenses were going for it on 4th down more often. I like much of what the article has to say--that going for it on 4th down isn't just for desperate situations and that teams with leads should sometimes go for it.

But in 2006, there was no trend. Pasquarelli's misguided observation was just the product of a small sample size and failing to account for the varying amount of 4th down opportunities there were in each year. Here are the 4th down conversion attempt numbers since 2002 accounting for 4th down opportunities, including projected(*) numbers though 11 games for 2008.

Year	4th Downs	Conv. Attempts	Rate (%)	Conversions	Success (%)
2002	4205	497	12	252	51
2003	4305	501	12	232	46
2004	4185	454	11	219	48
2005	4248	465	11	223	48
2006	4218	473	11	239	51
2007	4152	533	13	261	49
2008*	4096	468	17	252	54

If you just looked at 2004 through 2006, yes, you'd see an increase in absolute numbers of 4th down attempts. But as a percentage of 4th down opportunities, 2006 marked the end of a small downward trend.

Since Pasquarelli's article however, we've seen a remarkable increase in 4th down attempts. 2007 marked the first increase since at least 2002 with a 13% rate. So far 2008 has shown an even larger increase with a 17% rate. The season's not over yet, and two seasons does not a make trend. But something could be changing.