Standings Forecast Week 8

Season win totals and division standing projections are listed below. Projections are based on each team's opponent-adjusted generic win probability (GWP) which is the chance a team will win against a league-average opponent at a neutral site. The projections account for future opponent strength (Fut Opp), and projected wins (proj W) is a total of current and estimated future wins. The methodology is described more fully here.

TEAM	GWP	Fut Opp GWP	Proj GWP	Curr W	Proj W
AFC E
BUF	0.63	0.39	0.72	5	11.5
MIA	0.67	0.37	0.77	3	10.0
NYJ	0.46	0.47	0.49	4	8.4
NE	0.35	0.53	0.32	5	7.9
AFC N
PIT	0.63	0.51	0.62	5	10.6
BAL	0.46	0.55	0.41	4	7.7
CLE	0.35	0.52	0.33	3	6.0
CIN	0.19	0.52	0.18	0	1.4
AFC S
TEN	0.57	0.46	0.62	7	12.5
IND	0.49	0.43	0.56	3	8.1
JAX	0.43	0.44	0.48	3	7.4
HOU	0.43	0.48	0.45	3	7.0
AFC W
SD	0.74	0.44	0.78	3	9.2
DEN	0.45	0.53	0.42	4	7.8
OAK	0.38	0.53	0.35	2	5.1
KC	0.17	0.56	0.14	1	2.3
NFC E
WAS	0.82	0.50	0.82	6	12.5
PHI	0.80	0.52	0.79	4	11.1
NYG	0.68	0.66	0.52	6	10.7
DAL	0.62	0.58	0.54	5	9.3
NFC N
CHI	0.69	0.45	0.74	4	10.6
GB	0.49	0.54	0.45	4	8.0
MIN	0.50	0.49	0.50	3	7.5
DET	0.16	0.58	0.13	0	1.1
NFC S
CAR	0.75	0.50	0.75	6	12.0
TB	0.63	0.50	0.63	5	10.1
ATL	0.62	0.56	0.57	4	9.1
NO	0.65	0.51	0.63	4	9.1
NFC W
ARI	0.58	0.52	0.56	4	9.1
STL	0.32	0.50	0.32	2	4.9
SEA	0.31	0.58	0.25	2	4.2
SF	0.32	0.55	0.28	2	4.2

Week 9 Game Probabilities

Win probabilities for week 9 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
0.26	NYJ at BUF	0.74
0.06	DET at CHI	0.94
0.69	JAX at CIN	0.31
0.52	BAL at CLE	0.48
0.33	GB at TEN	0.67
0.66	ARI at STL	0.34
0.34	HOU at MIN	0.66
0.85	TB at KC	0.15
0.63	MIA at DEN	0.37
0.66	ATL at OAK	0.34
0.35	DAL at NYG	0.65
0.86	PHI at SEA	0.14
0.28	NE at IND	0.72
0.21	PIT at WAS	0.79

San Diego's Defensive Woes

Last July I wrote that the Chargers defense would appear to significantly decline in 2008, even with no change in skill or performance. The reason was that they could not possibly repeat their phenomenally high defensive interception rate. At the halfway mark this season, the Chargers evidently don't think they're doing very well because they just fired their defensive coordinator. But is the Chargers defense really that bad, or is it just unlucky?

Last season San Diego led the NFL with an insanely high 5.4% interception rate. That means they intercepted more than 1 out of every 20 passes attempted. But because defensive interceptions are almost completely random, I forecast that the overwhelming likelihood would be that their 2008 interception rate would be far closer to average. In fact, it's currently well below average at 2.2%. The average so far this year is 2.6%.

Using regression coefficients to weight the importance of each major efficiency stat, we can estimate that the difference in defensive interception rate from 2007 to '08 would cost San Diego 2.6 games over the course of a full season, or about 1.3 games so far this season.

Aside from their interceptions, the Chargers' pass defense is still above average but not as good as last year. So far in 2008, San Diego has allowed 6.2 net yards per attempt compared to an NFL average of 6.3. Last year, they allowed 5.7 net yards per attempt. On average, this would have cost them a difference of 0.8 wins over the course of a full season, or 0.4 wins so far.

Their defensive running efficiency is exactly the same at 4.2 yards per attempt. That's slightly worse than average, which is 4.1 yds per att.

In total, the Chargers interception rate (which is overwhelmingly random) accounts for an estimated 1.3 wins out of a total difference of 1.7, or about 75%.

So Ted Cotrell was fired as defensive coordinator for a drop-off of half a yard per pass attempt, or just less than half a win. The unfortunate loss of pass rusher Shawne Merriman to injury would easily account for such a difference. Now I'll make a new prediction. Because San Diego's interception rate has been so low this year, it's bound to improve (i.e. regress to the mean), and most people will think Cotrell's replacement, Ron Rivera, is to thank.

My point isn't that the Chargers have lost precisely 1.7 games fewer this year because of their defense. My main points are: 1) if you didn't buy how random defensive interceptions really are, maybe you'll reconsider; 2) Cotrell probably shouldn't have been fired; and 3) expect the Chargers defense to improve for the remainder of the year, if only because their interception rate is likely to improve.

Play of the Week

With a 21 to 17 lead midway through the 4th quarter, the NY Jets drove to the Kansas City 8 yard line. A virtually certain field goal would have given them a full touchdown lead and guaranteed New York at least one more full possession. This situation gave the Jets about an 85% chance of victory. But on 3rd and 2, Brett Favre was intercepted for the 3rd time in the game, this time by Chiefs CB Brandon Flowers who returned the ball 91 yards for a touchdown and a 3 point lead. Following the extra point, the Chiefs had an 80% chance of winning, for an amazing swing of 65%.

But the Jets got the last laugh. Their final drive culminated in a 15-yard TD pass to Laveranues Coles to take back the lead for good. The Chiefs had 1:00 to respond. With good field position following the kickoff, Kansas City made it to the Jets' 31 yard line with a 32% chance of pulling out the victory, before stalling on 4th and 1.


Congratulations, Brandon Flowers. Enjoy your free lifetime subscription to Advanced NFL Stats Premium.

Runner-up this week goes to the Dallas defense that stopped Tampa's final drive in the closing seconds to seal their 13-9 win.


The comeback of the week, however, is owned by the Panthers, who faced a 17-3 deficit against the Cardinals. With ten minutes left in the 3rd quarter, Carolina only had a 5% chance of winning before mounting their comeback. This was about 3 times more improbable than the Jets' amazing win.

Week 8 Efficiency Rankings

This week's 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.

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 variables.

Offensive rank (ORANK) is the ranking of 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.

RANK	TEAM	LAST WK	GWP	Opp GWP	ORANK	DRANK
1	WAS	1	0.82	0.52	3	5
2	PHI	2	0.81	0.58	5	4
3	CAR	4	0.75	0.58	9	3
4	SD	3	0.74	0.54	4	16
5	CHI	7	0.70	0.57	10	7
6	NYG	8	0.68	0.42	1	15
7	MIA	12	0.67	0.52	2	24
8	NO	9	0.65	0.58	6	18
9	TB	14	0.64	0.58	17	2
10	BUF	5	0.63	0.49	21	10
11	PIT	6	0.63	0.48	26	1
12	ATL	11	0.63	0.53	7	21
13	DAL	10	0.62	0.53	12	11
14	ARI	13	0.59	0.61	8	19
15	TEN	15	0.57	0.38	18	8
16	MIN	17	0.50	0.55	22	9
17	GB	18	0.50	0.48	14	13
18	IND	20	0.49	0.51	13	14
19	NYJ	16	0.46	0.44	30	12
20	BAL	23	0.46	0.47	25	6
21	DEN	21	0.45	0.48	16	29
22	HOU	24	0.43	0.45	15	30
23	JAX	19	0.43	0.51	11	28
24	OAK	22	0.38	0.51	23	20
25	NE	27	0.35	0.45	27	26
26	CLE	28	0.35	0.55	20	23
27	STL	26	0.33	0.60	19	27
28	SEA	25	0.32	0.49	24	25
29	SF	29	0.33	0.48	31	17
30	CIN	30	0.19	0.53	32	22
31	KC	31	0.18	0.51	29	31
32	DET	32	0.17	0.56	28	32

The to-date season efficiency stats are listed below.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.1	3.3	0.022	0.027	6.9	4.0	0.014	0.41
ATL	6.4	4.8	0.024	0.009	6.4	4.7	0.021	0.35
BAL	5.7	3.8	0.036	0.030	5.4	2.8	0.042	0.48
BUF	6.8	3.8	0.019	0.028	6.0	3.8	0.018	0.30
CAR	7.1	3.7	0.022	0.018	5.5	4.0	0.019	0.39
CHI	6.5	3.6	0.017	0.019	5.7	3.7	0.035	0.40
CIN	4.1	3.4	0.034	0.035	6.4	4.3	0.017	0.36
CLE	5.3	3.8	0.029	0.024	6.5	4.7	0.045	0.40
DAL	7.0	4.6	0.030	0.036	5.7	3.9	0.008	0.51
DEN	7.0	4.6	0.027	0.032	6.8	5.4	0.009	0.36
DET	5.3	4.2	0.031	0.036	8.4	4.7	0.005	0.42
GB	6.7	3.7	0.018	0.028	5.3	4.9	0.056	0.61
HOU	6.8	4.3	0.033	0.029	7.2	4.5	0.026	0.21
IND	6.2	3.4	0.034	0.009	5.8	4.2	0.025	0.44
JAX	5.8	4.1	0.018	0.014	7.0	4.4	0.027	0.38
KC	4.4	4.4	0.034	0.025	7.3	5.6	0.031	0.28
MIA	7.6	3.8	0.014	0.020	7.2	3.7	0.019	0.30
MIN	5.8	4.3	0.033	0.028	6.3	2.9	0.018	0.46
NE	5.5	4.2	0.028	0.016	6.8	4.4	0.041	0.29
NO	8.1	3.4	0.023	0.031	6.4	4.3	0.021	0.50
NYG	6.7	5.1	0.018	0.014	5.5	3.8	0.038	0.50
NYJ	5.9	4.6	0.050	0.023	6.0	3.2	0.020	0.31
OAK	5.3	4.3	0.015	0.031	6.5	4.5	0.032	0.41
PHI	6.8	4.1	0.016	0.017	5.6	3.5	0.035	0.34
PIT	5.9	3.9	0.036	0.033	4.4	2.8	0.025	0.43
SD	8.1	3.9	0.026	0.017	6.4	4.0	0.019	0.37
SF	6.0	4.5	0.045	0.044	6.4	3.6	0.031	0.48
SS	4.7	4.4	0.036	0.010	6.9	4.0	0.009	0.30
STL	5.3	4.0	0.020	0.023	7.3	4.8	0.029	0.45
TB	5.6	4.3	0.026	0.014	5.5	3.6	0.046	0.48
TEN	5.9	4.3	0.026	0.017	4.9	3.7	0.048	0.38
WAS	6.7	4.7	0.000	0.016	5.5	3.8	0.018	0.35
Avg	6.2	4.1	0.026	0.023	6.3	4.1	0.026	0.39

In-Game Win Probability Engine 0.5

A big (cosmetic) improvement for tonight's IND-TEN game. The win probability site now uses a new way of updating itself that makes it far smoother, especially for multiple games. Using AJAX, the site will now update itself automatically in the background without having to refresh and redraw the screen every 30 seconds. It might not sound like such an improvement, but the site will lose a lot of the quaint 1990's homemade feel. The updates will be much quicker. No more annoying flicker while the file downloads. And no more scrolling back down to the game you were following (in some browsers).

Final games from Sunday are here.

In-Game Win Probability Engine 0.4

Improvements to the (near) live win probability scoreboard this week include a model for overtime and a fix for some halftime bugs. The site will also update quicker and feature some minor cosmetic upgrades. The are also a number of behind-the-scenes improvements in the programming. The link is wp.advancednflstats.com. You can see graphs of last week's games here and week 6 games here. Feedback is always more than welcome. I'm currently drafting a post explaining how it works in case anyone is interested. Also, feel free to send in your suggestion for play of the week.

Week 8 Game Probabilities

Win probabilities for week 8 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
0.47	BUF at MIA	0.53
0.43	OAK at BAL	0.57
0.31	ARI at CAR	0.69
0.46	TB at DAL	0.54
0.59	SD at NO*	0.41
0.28	ATL at PHI	0.72
0.17	KC at NYJ	0.83
0.44	STL at NE	0.56
0.91	WAS at DET	0.09
0.31	CIN at HOU	0.69
0.30	CLE at JAX	0.70
0.39	NYG at PIT	0.61
0.35	SEA at SF	0.65
0.33	IND at TEN	0.67

*See the comment below by Jeremiah regarding the London game. The probability above has been updated to reflect the neutral site location of the SD-NO game.

Week 7 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.

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.

Offensive rank (ORANK) is the ranking of 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.

RANK	TEAM	LAST WK	GWP	Opp GWP	ORANK	DRANK
1	WAS	1	0.76	0.55	1	3
2	PHI	3	0.72	0.57	4	4
3	SD	2	0.68	0.52	5	13
4	CAR	4	0.68	0.54	9	2
5	BUF	14	0.65	0.49	14	8
6	PIT	12	0.65	0.45	18	1
7	CHI	6	0.63	0.53	10	7
8	NYG	5	0.63	0.43	2	18
9	NO	8	0.62	0.54	6	16
10	DAL	7	0.59	0.52	11	15
11	ATL	11	0.59	0.46	3	23
12	MIA	9	0.59	0.50	7	27
13	ARI	10	0.58	0.57	8	20
14	TB	13	0.55	0.54	17	5
15	TEN	17	0.54	0.39	20	10
16	NYJ	15	0.50	0.50	28	9
17	MIN	18	0.49	0.51	21	11
18	GB	22	0.48	0.46	16	12
19	JAX	20	0.46	0.52	12	22
20	IND	19	0.46	0.49	13	19
21	DEN	16	0.45	0.48	15	29
22	OAK	23	0.44	0.52	24	21
23	BAL	24	0.42	0.48	27	6
24	HOU	21	0.41	0.48	22	31
25	SF	25	0.40	0.50	30	14
26	STL	30	0.40	0.61	19	26
27	NE	26	0.36	0.47	25	28
28	CLE	28	0.35	0.56	23	24
29	SEA	29	0.34	0.51	26	25
30	CIN	27	0.31	0.52	31	17
31	KC	31	0.23	0.52	32	30
32	DET	32	0.23	0.50	29	32

The to-date season efficiency stats are listed below.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.16	3.24	0.023	0.025	6.71	4.01	0.017	0.40
ATL	6.68	5.02	0.019	0.010	6.35	4.40	0.024	0.34
BAL	5.39	3.75	0.042	0.033	5.52	2.84	0.044	0.50
BUF	7.02	3.64	0.017	0.023	5.47	4.07	0.021	0.26
CAR	6.90	3.66	0.025	0.014	5.16	4.01	0.018	0.42
CHI	6.50	3.63	0.017	0.019	5.65	3.73	0.035	0.40
CIN	4.05	3.21	0.030	0.037	6.00	4.37	0.020	0.35
CLE	4.89	3.78	0.033	0.023	6.65	4.80	0.057	0.45
DAL	7.60	4.87	0.034	0.041	5.95	4.01	0.009	0.55
DEN	7.02	4.63	0.027	0.032	7.10	5.09	0.010	0.39
DET	5.16	4.28	0.036	0.038	8.19	4.79	0.006	0.41
GB	6.67	3.66	0.018	0.028	5.34	4.89	0.056	0.61
HOU	6.51	4.39	0.037	0.035	7.74	4.41	0.019	0.16
IND	6.30	3.28	0.031	0.011	5.98	4.41	0.030	0.47
JAX	5.76	4.11	0.022	0.016	6.75	4.46	0.031	0.40
KC	4.10	4.43	0.040	0.029	7.50	5.62	0.020	0.27
MIA	7.26	4.14	0.016	0.017	7.35	3.61	0.017	0.28
MIN	5.77	4.28	0.033	0.028	6.32	2.95	0.018	0.46
NE	5.21	4.38	0.022	0.015	7.14	4.59	0.037	0.37
NO	8.06	3.56	0.027	0.030	6.11	4.26	0.021	0.50
NYG	6.83	5.62	0.021	0.013	5.64	3.76	0.022	0.49
NYJ	5.70	4.46	0.045	0.026	5.97	3.15	0.024	0.33
OAK	5.30	4.52	0.012	0.035	6.38	4.56	0.036	0.47
PHI	6.87	3.68	0.018	0.017	5.55	3.54	0.033	0.30
PIT	6.15	3.86	0.018	0.031	4.14	2.94	0.029	0.43
SD	7.99	3.81	0.026	0.019	6.17	4.20	0.022	0.26
SF	6.07	4.43	0.051	0.039	6.23	3.88	0.035	0.47
SEA	4.18	4.87	0.041	0.012	7.27	3.85	0.005	0.33
STL	4.92	4.08	0.018	0.023	7.42	5.02	0.023	0.45
TB	5.78	4.50	0.031	0.016	5.92	3.73	0.053	0.48
TEN	6.11	4.50	0.032	0.017	4.77	3.63	0.048	0.39
WAS	6.30	4.73	0.000	0.013	5.45	3.85	0.021	0.32
Avg	6.13	4.16	0.027	0.024	6.25	4.11	0.027	0.40

Play of the Week

One neat thing we can do with a win probability model is identify a 'play of the week.' I realize its an old and overdone concept (I feel like I should get Old Spice to sponsor it), but now we really can point to the truly most important play among the several hundred each week of the season.

In week 7, the official Advanced NFL Stats play of the week came late in the Charger-Bills game. With Buffalo ahead 20-14 with 6:16 left in the 4th quarter, the San Diego offense drove deep into Bills territory. With a 1st down and goal from the Bills' 9 yard line, the Chargers actually had the upper hand with a WP of 0.60. Phillip Rivers dropped back and threw directly into the hands of Bills defender Kawika Mitchell who returned the ball 34 yards.

As a result, the Chargers dropped to a 0.14 WP, a catastrophic fall of 0.46. The Bills went on to put together a field goal drive that essentially put the game out of reach. Congratulations Kawika, you've won a lifetime premium membership to Advanced NFL Stats!

In-Game Win Probabilities 0.3

No new improvements to the in-game win-probability model this week as I was out of the country. There are still some bugs (mostly due to the data feed), but all games now get a near real-time graph, and there are separate pages for ongoing games and final games. It considers score, possession, field position, and time remaining. Down and distance adjustments are still in work. Check it out at wp.advancednflstats.com for this week's games. You can see last week's games here. Feedback is always more than welcome. It can be pretty mesmerizing. I'll keep it on-line through MNF.

Week 7 Game Probabilities

Win probabilities for week 7 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
0.71	SD at BUF	0.29
0.35	NO at CAR	0.65
0.24	MIN at CHI	0.76
0.73	PIT at CIN	0.27
0.84	TEN at KC	0.16
0.18	BAL at MIA	0.82
0.14	SF at NYG	0.86
0.82	DAL at STL	0.18
0.09	DET at HOU	0.91
0.47	IND at GB	0.53
0.54	NYJ at OAK	0.46
0.05	CLE at WAS	0.95
0.14	SEA at TB	0.86
0.63	DEN at NE	0.37

How Important is the Coin Flip in OT?

All of our favorite teams have been on the short end of the stick when it comes to sudden death overtime in the NFL. The opposing team wins the coin flip, gets a decent return, completes a couple passes, then kicks a game-winning 40+ yard field goal. Our team never even gets a chance to touch the ball. It's a painful end to an otherwise exciting game.

Everyone knows the coin toss can be decisive. The team that wins the toss instantly becomes favored to win the game, but just how heavily?

From the 2000 through 2007 regular seasons, there have been 124 overtime games. In every single game except one (I believe), the team that won the toss elected to receive. And those receiving teams won 60% of the time (and tied once). That's a relatively large advantage, particularly when compared to home field advantage.

Home teams have only won 51% of OT games. The weakness of HFA isn't too surprising given the way it diminishes throughout a game. It's strongest in the 1st quarter and then diminishes through subsequent quarters until it's almost non-existent in OT. Fans are presumably at their most involved at this point in a game, which suggests crowd involvement is not the primary source of HFA.

The dreaded 'lose-the-coin-toss-never-touch-the-ball' scenario happened in 37 out of the 124 OT periods, or about 30% of all overtime games. That's too often in my opinion. The NFL's current sudden death format can be exciting and lead to quick resolutions. But if almost 1 out of 3 games is over before the unlucky coin toss loser even touches the ball, a lot of teams and fans are going to be left with a bitter and empty feeling.

One suggestion is to go to the college format where each team gets alternating tries to score from the opponent's 25 yd line. It eliminates the never-touch-the-ball problem, but it has its own shortcoming. Namely, the team that gets its possession second has a distinct advantage because it knows exactly what type of score is needed to tie or win. For example, if the first team doesn't score at all, the second doesn't need to risk passing and can safely run 3 times before kicking an easy field goal. Or, if the first team scores a touchdown, the second team knows it must forego the field goal and go for the touchdown, even on 4th down if necessary.

The NCAA mitigates the advantage of going second by alternating the order on successive rounds. But the team that goes second in the first round will have the overall advantage because there are many more 1, 3, or 5 round overtimes than 2, 4, or 6 round overtimes. But the overall advantage is estimated to be small, at about 52%.

Although I like the NCAA format, I'd make a couple changes. First, no field goals. This has two effects. First, it eliminates the advantage of the team to go second. Both teams simply need a touchdown, period. Second, it puts the game solely in the hands of the offenses and defenses, and not in those of an individual place kicker.

Because removing field goals might prolong the game excessively, I'd add another requirement. Only 2-point conversions would be allowed. In the NCAA, teams are forced to go for 2-pt conversions if no team has won by the 3rd round. But I'd institute that rule beginning with the 1st round, maybe the second.

Unfortunately, a lot of fans find that the NCAA format is not "pure" football. And I sympathize with that opinion, so here are a few more suggestions.

One idea is to play a semi-sudden death format. The current system would be kept, except that a winner is declared only after one team is ahead after an equal number of possessions. So if the initial receiving team scored first, the other team would receive a kick-off and have an opportunity to tie or win. I like this idea, except that from the NFL's point of view, OT games would last longer and ties would be more common. It might also suffer from an advantage problem, because a team in the score-or-die situation would have the same advantage that the team to go second in the NCAA format has.

David Romer's suggestion is to move the kickoff line from the 30 to the 40 in overtime to help equalize the chance of either team scoring first. This would drastically increase touchbacks, which according to Romer would halve the receiving team's advantage. Starting at the 15 yd line is the theoretical neutral point in the NFL, where both teams have an equal chance of scoring next.

This guy makes a related observation. In 1994, the NFL moved the kickoff line from the 40 to the 30 to reduce touchbacks and increase scoring. But unwittingly, this change also increased the frequency of the never-touch-the-ball phenomenon in OT.

Another idea is to have dueling kickoffs. Both teams would return a kickoff, and the team with the furthest return gets possession at the start of the sudden-death period.

Perhaps the silliest but most original idea is the field position bid. Both teams would submit a secret bid of how far back they'd be willing to start with the ball. The team that bids the deepest in its own territory would get the ball there. A football version of Name That Tune, I suppose.

Week 6 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 based on offensive generic win probability 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	OGWP	DGWP
10	ARI	8	0.64	0.63	13	19
11	ATL	19	0.63	0.42	9	24
24	BAL	21	0.37	0.46	28	12
14	BUF	12	0.54	0.34	21	14
4	CAR	5	0.71	0.54	10	1
6	CHI	9	0.70	0.54	8	5
27	CIN	28	0.28	0.55	30	18
28	CLE	31	0.27	0.57	19	28
7	DAL	7	0.69	0.50	4	9
16	DEN	11	0.53	0.47	11	25
32	DET	32	0.12	0.56	31	32
22	GB	23	0.45	0.45	18	15
21	HOU	26	0.46	0.58	16	30
19	IND	20	0.52	0.47	12	17
20	JAX	25	0.50	0.58	14	22
31	KC	30	0.13	0.52	32	31
9	MIA	6	0.66	0.51	7	23
18	MIN	15	0.52	0.52	15	10
26	NE	22	0.31	0.51	25	29
8	NO	13	0.67	0.52	6	11
5	NYG	2	0.71	0.33	2	16
15	NYJ	10	0.54	0.50	26	4
23	OAK	16	0.41	0.57	24	20
3	PHI	4	0.78	0.56	5	8
12	PIT	14	0.61	0.43	20	3
2	SD	3	0.81	0.51	1	13
25	SS	27	0.36	0.43	29	21
29	SF	24	0.27	0.52	27	26
30	STL	29	0.25	0.66	22	27
13	TB	17	0.61	0.63	17	6
17	TEN	18	0.53	0.39	23	7
1	WAS	1	0.83	0.60	3	2

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.16	3.24	0.023	0.025	6.71	4.01	0.017	0.40
ATL	6.68	5.02	0.019	0.007	6.35	4.40	0.024	0.34
BAL	4.89	3.70	0.049	0.032	4.95	2.77	0.048	0.48
BUF	6.71	3.70	0.021	0.028	5.34	4.05	0.018	0.25
CAR	6.72	3.62	0.028	0.017	5.04	3.84	0.017	0.45
CHI	6.25	3.78	0.020	0.016	5.47	3.46	0.024	0.40
CIN	4.32	3.12	0.036	0.036	5.64	4.34	0.023	0.34
CLE	5.23	3.81	0.041	0.023	6.67	4.66	0.067	0.51
DAL	7.90	4.76	0.025	0.041	5.94	3.72	0.010	0.56
DEN	7.26	4.72	0.022	0.027	7.10	5.09	0.010	0.30
DET	4.58	4.32	0.041	0.034	8.35	4.86	0.007	0.44
GB	6.68	3.74	0.020	0.030	5.31	5.11	0.058	0.62
HOU	6.34	4.38	0.043	0.035	7.46	4.46	0.022	0.13
IND	6.49	3.30	0.027	0.009	5.86	4.63	0.036	0.38
JAX	5.76	4.11	0.022	0.016	6.75	4.46	0.031	0.40
KC	3.82	4.57	0.049	0.030	7.59	5.03	0.022	0.23
MIA	7.09	4.29	0.013	0.020	7.20	3.50	0.020	0.29
MIN	5.58	4.18	0.020	0.033	6.04	3.03	0.021	0.47
NE	5.26	3.77	0.025	0.019	7.14	4.59	0.037	0.30
NO	8.47	3.32	0.027	0.031	5.88	4.37	0.023	0.54
NYG	7.13	6.08	0.025	0.008	5.48	3.97	0.013	0.45
NYJ	6.03	3.66	0.043	0.020	5.97	2.88	0.028	0.31
OAK	5.16	4.64	0.015	0.039	6.88	3.94	0.032	0.44
PHI	6.87	3.68	0.018	0.017	5.55	3.54	0.033	0.30
PIT	5.76	3.71	0.022	0.029	4.48	2.78	0.036	0.49
SD	8.32	3.76	0.024	0.015	5.87	4.37	0.025	0.26
SF	6.02	4.67	0.047	0.030	6.38	3.94	0.040	0.36
SS	4.33	4.70	0.041	0.009	7.04	4.18	0.007	0.39
STL	4.75	3.75	0.020	0.029	7.80	4.94	0.007	0.48
TB	5.39	4.95	0.036	0.010	6.21	3.45	0.054	0.52
TEN	6.02	3.58	0.036	0.020	4.65	3.66	0.059	0.37
WAS	6.27	4.62	0.000	0.009	5.80	3.90	0.025	0.32
AVG	6.10	4.10	0.028	0.023	6.22	4.06	0.028	0.39

In-Game Win Probabilities Beta 0.3

Still more improvements to the in-game win-probability model. There are still some bugs (mostly due to the data feed), but all games now get a near real-time graph, and there are separate pages for ongoing games and final games. It considers score, possession, field position, and time remaining. Down and distance adjustments are still in work. Check it out at wp.advancednflstats.com and feedback is more than welcome. It can be pretty mesmerizing. I'll keep it on-line through MNF.

What's the Frequency, Kenneth?

I was interested in field position, not the theoretical point value of field position, or win probabilities, or anything else dense and cosmically incomprehensible. I was just curious how often the ball was snapped at each yard-line on the field. I was mostly interested because I wanted to see how field positions might bias the results of some of my other number crunching. So I plotted the frequency distribution of field position for every play from scrimmage from 2000 through 2007, and I found something odd.

Below is the distribution. The yard-lines are numbered according to the distance to the goal line. For example, the '70' would be an offense's own 30 yd-line.


The first thing that stands out is the large number of plays at the 20. Of course, this is due to the fact that touchbacks automatically put the ball here. No surprise there.

The next thing I noticed was the sawtooth pattern, a series of smaller spikes extending across the field. At first I thought this was random noise, but then I noticed how regular they were. The pattern wasn't completely apparent to me until I added the vertical grid lines. There is a noticeably higher number of snaps from every 5-yd increment on the field than at other yd-lines.

The spikes at the 25 and the 30, and even the 15 and 10, made sense to me. Since there were lots of plays from the 20, penalties would commonly put the ball at 5 and 10 yard increments from there. But the spikes continue down the entire length of the field, all the way to the end zone. I really doubt chains of penalties beginning at the 20 could account for that.

Could there be lots of series (aside from penalties) that tend to gain 5, 10, or 15 yards rather than 9, 11, 12, or 14? It seems unlikely in the extreme, although I can't disprove it for now.

I'm baffled. If I had to put money on it, I'd guess that because the 5-yd lines represent nice round numbers, and are boldly drawn from sideline to sideline, it's easy for refs to spot the ball there, either intentionally or subconsciously. This would be the case especially if the mark is somewhat ambiguous, such as when a ball carrier slides or a punt sails out of bounds.

There is one interesting exception, too. There's no spike at the defense's 10 yd line, but there is one at the 11 instead. Strange. My guess here is that because there are lots of plays on the goal line (probably due to lots of stuffed runs and/or pass interference calls in the end zone), 10-yard penalties such as holding would force the ball to be at the 11, instead of the 10.

I guess this belongs in the ever-growing category of 'things only I could possibly care about.'

Week 6 Game Probabilities

Win probabilities for week 5 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
0.57	CHI at ATL	0.43
0.37	BAL at IND	0.63
0.07	DET at MIN	0.93
0.35	OAK at NO	0.65
0.14	CIN at NYJ	0.86
0.66	CAR at TB	0.34
0.04	STL at WAS	0.96
0.76	MIA at HOU	0.24
0.24	JAX at DEN	0.76
0.43	DAL at ARI	0.57
0.76	PHI at SF	0.24
0.50	GB at SEA	0.50
0.13	NE at SD	0.87
0.95	NYG at CLE	0.05

Week 5 Efficiency Rankings

NFL team efficiency rankings are back for 2008. 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 based on offensive generic win probability 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	ORANK	DRANK
1	WAS	1	0.86	0.70	1	4
2	NYG	3	0.84	0.42	2	3
3	SD	2	0.79	0.64	4	13
4	PHI	5	0.76	0.61	6	7
5	CAR	4	0.75	0.54	5	1
6	MIA	18	0.74	0.61	8	20
7	DAL	6	0.70	0.49	3	11
8	ARI	8	0.68	0.64	10	17
9	CHI	12	0.67	0.52	9	2
10	NYJ	14	0.64	0.64	20	10
11	DEN	15	0.59	0.50	7	28
12	BUF	9	0.58	0.41	23	14
13	NO	7	0.58	0.58	13	15
14	PIT	13	0.57	0.42	17	5
15	MIN	10	0.56	0.55	14	12
16	OAK	16	0.52	0.54	22	16
17	TB	19	0.52	0.55	16	9
18	TEN	11	0.51	0.40	24	8
19	ATL	20	0.50	0.39	11	24
20	IND	21	0.47	0.49	12	22
21	BAL	23	0.44	0.39	25	6
22	NE	27	0.44	0.49	27	23
23	GB	26	0.42	0.48	18	19
24	SF	17	0.41	0.41	30	18
25	JAX	22	0.39	0.52	15	27
26	HOU	25	0.38	0.50	19	26
27	SEA	24	0.33	0.53	26	31
28	CIN	29	0.29	0.54	28	21
29	STL	28	0.27	0.63	21	30
30	KC	30	0.20	0.57	32	29
31	CLE	31	0.16	0.50	29	25
32	DET	32	0.13	0.51	31	32

To give everyone an insight into why the rankings are what they are, here are the team efficiency stats.

TEAM	OPASS	ORUN	OINTRATE	OFUMRATE	DPASS	DRUN	DINTRATE	PENRATE
ARI	7.1	3.3	0.022	0.025	6.5	4.1	0.021	0.36
ATL	5.9	5.5	0.023	0.004	6.3	4.6	0.030	0.34
BAL	4.9	3.8	0.038	0.022	4.1	2.8	0.060	0.55
BUF	6.7	3.7	0.021	0.028	5.3	4.0	0.018	0.25
CAR	6.9	3.8	0.014	0.020	4.6	3.8	0.019	0.50
CHI	6.2	3.8	0.026	0.018	4.9	3.7	0.028	0.39
CIN	4.5	3.3	0.043	0.035	5.9	4.5	0.014	0.37
CLE	4.0	3.5	0.051	0.024	6.7	4.1	0.057	0.52
DAL	8.0	5.0	0.031	0.032	5.7	3.9	0.006	0.51
DEN	7.6	4.5	0.021	0.023	7.1	5.2	0.012	0.28
DET	4.7	4.4	0.047	0.024	8.9	5.0	0.000	0.35
GB	6.8	4.0	0.024	0.034	5.7	5.1	0.054	0.66
HOU	5.7	4.4	0.042	0.039	6.8	4.5	0.018	0.15
IND	6.1	3.6	0.032	0.006	6.2	4.9	0.020	0.31
JAX	5.4	4.0	0.027	0.008	7.1	4.2	0.031	0.41
KC	3.8	4.6	0.049	0.030	7.6	5.0	0.022	0.23
MIA	6.5	4.3	0.008	0.019	6.7	3.3	0.009	0.30
MIN	5.3	4.1	0.018	0.024	6.3	2.8	0.023	0.41
NE	5.5	3.7	0.025	0.018	6.1	4.9	0.047	0.22
NO	8.2	3.2	0.031	0.038	6.3	4.5	0.022	0.60
NYG	7.2	5.8	0.008	0.009	4.4	3.7	0.016	0.48
NYJ	6.4	3.8	0.039	0.021	6.6	3.1	0.034	0.34
OAK	5.6	4.8	0.010	0.041	6.1	4.0	0.039	0.50
PHI	6.7	3.5	0.016	0.020	5.6	3.3	0.026	0.34
PIT	5.8	3.7	0.022	0.029	4.5	2.8	0.036	0.49
SD	7.8	3.8	0.029	0.014	6.1	4.4	0.025	0.27
SF	6.2	4.6	0.043	0.032	6.1	3.8	0.043	0.34
SEA	4.7	4.7	0.032	0.010	7.2	4.6	0.008	0.38
STL	4.8	4.0	0.025	0.030	8.2	4.7	0.009	0.43
TB	5.1	5.3	0.039	0.008	6.2	3.7	0.048	0.49
TEN	6.0	3.6	0.036	0.020	4.6	3.7	0.059	0.37
WAS	6.3	4.4	0.000	0.000	6.0	4.1	0.029	0.30
AVG	6.0	4.1	0.028	0.022	6.1	4.1	0.028	0.39

Reggie Bush vs. Gus Frerotte

Wow, another cardiac arrest of a game Monday night. Here's how it unfolded in terms of win probability. Reggie Bush almost single-handedly won the game for the Saints, responsible for two touchdown returns and another return into Minnesota territory leading to a field goal. At one point New Orleans was up 27-20 with less than 10 minutes remaining in the game. But Gus Frerotte was able to make some incredible passes down the stretch. Gramatica's missed field goal in the 4th quarter was the real back-breaker. Not only did it waste a solid New Orleans drive, but it gave the ball to the Vikings in good field position with just enough time to score and not enough time for the Saints to respond. But the 42-yard pass interference by New Orleans defender Kaesviharn in Minnesota's game-winning drive sure didn't help.

PIT-JAX in Win Probabilities

As I'm testing out my new win probability toy, there are lots of fun things to do. Here is a timeline of Sunday night's game between the Steelers and Jaguars. The Steelers got out of town with the W, but late in the 4th quarter it was Jacksonville's game to lose. The wild swings in win probability made for a fun game to watch. Check in tonight for the MIN-NO game's WP graph to be updated live.

In-Game Win Probabilities Beta 0.2

I'm testing a real-time win probability site again this week. The current state of each ongoing game will be reported along with the probability of winning for each team. The probabilities are based on over 2000 games from the past 8 regular seasons. Last week, the model was fairly basic and simply considered score, time, and possession. This week modifications for field position and seconds are included. You can think of each probability as "if the team with the ball had a first down at their current field position, this would be their chance of winning."

Next week, I hope to include down and distance to go. I'm also planning to include a timeline graph of each game's win probabilities.

Readers of this site are invited to check out the beta version during the games today. Be warned there will certainly be hiccups, bugs, and other problems throughout the day. So if it's not working one minute, it may be back up the next. It goes live shortly after the 1 o'clock kickoffs today. Comments and suggestions are more than welcome!

The link is wp.advancednflstats.com.

Week 5 Game Probabilities

Win probabilities for week 5 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.

Last week the predictions were 10-3. But if you are a new reader, don't expect the model to be that accurate every week. On average it will be between 70-75% correct, and some weeks it can go 6-8. Over the past 2 seasons however, it's been the most accurate prediction system I can find, but only slightly more accurate at picking winners than the spread.

P WIN	GAME	P WIN
0.15	KC at CAR	0.85
0.79	CHI at DET	0.21
0.53	ATL at GB	0.47
0.68	SD at MIA	0.32
0.15	SEA at NYG	0.85
0.56	WAS at PHI	0.44
0.59	TEN at BAL	0.41
0.54	IND at HOU	0.46
0.31	TB at DEN	0.69
0.42	BUF at ARI	0.58
0.28	NE at SF	0.72
0.12	CIN at DAL	0.88
0.47	PIT at JAX	0.53
0.30	MIN at NO	0.70