Game Predictions Week 13

Game probabilities for week 13 NFL games are listed below. The probabilities are based on an efficiency win model explained here and here. The model considers offensive and defensive efficiency stats including running, passing, sacks, turnover rates, and penalty rates. Team stats are adjusted for previous opponent strength.

Vprob	Visitor	Home	Hprob
0.23	GB	DAL	0.77
0.59	ATL	STL	0.41
0.27	BUF	WAS*	0.73
0.22	DET	MIN	0.78
0.40	HOU	TEN	0.60
0.19	JAX	IND	0.81
0.37	NYJ	MIA	0.63
0.61	SD	KC	0.39
0.44	SEA	PHI	0.56
0.27	SF	CAR	0.73
0.81	TB	NO	0.19
0.49	CLE	ARI	0.51
0.72	DEN	OAK	0.28
0.77	NYG	CHI	0.23
0.16	CIN	PIT	0.84
0.93	NE	BAL	0.07

NFL Home Field Advantage by Climate 2

In a previous post, I looked at how home field advantage is affected by weather. Each NFL city was categorized by its December climate--dome, warm, moderate, and cold. By comparing the winning percentage of home teams in games in the early season (weeks 1-12) and the late season (13-17) the effect of cold weather was estimated.

In this post, the effect of cold weather on home field advantage will be measured more precisely. By using logistic regression, relative team strength is accounted for. In addition, the statistical significance of the observed weather effect indicates if the effect is real and systematic, or just a result of luck and small sample size.

The last post left off with this table. The left most column describes the visiting team's climate, and the top row describes the home team's climate. For each combination of visiting and home climate, the change in home team winning percentage from early to late in the season is listed. Positive numbers indicate that cold weather may favor the home team. For example, when dome teams play at warm cities, the home team winning percentage was 20% higher late in the season than early in the season. But when dome teams play at moderate cities, the home team winning percentage appeared to be 9% lower late in the season.

Difference	Dome	Warm	Mod	Cold
Dome	0%	+20	-9	+35
Warm	-5	+14	-7	-3
Mod	-8	-24	-6	-13
Cold	-8	+2	-15	+8

Several match-ups indicate that the cold and wind of late-season outdoor football has an effect on HFA. Consistent with other research, it appears that dome teams suffer when playing in cold climates.

Another remarkable combination is moderate weather teams playing in warm cities (-24%). But it's not clear why moderate teams would have an easier time playing in the balmy breezes of Florida or San Diego in December. Other notable match-ups are dome teams playing at warm cities (+20%), and cold teams playing at moderate cities (-15%).

To determine if the observed differences in HFA between early and late season games is really due to the change of weather, and not due to relative team strength or luck, several logistic regression models were run. The models were based on every regular season game from the 2002 through 2006 seasons (n=1280). For each game, each team was designated either Team A or Team B. The general model specification was the following:

Dependent variable:
Team A won

Independent variables:
Team A season efficiency stats
Team B season efficiency stats
AHome
[Weather dummy variable]

The team efficiency stats include offensive and defensive passing and running efficiency, turnover rates, and penalty rates. AHome is a dummy variable that is 0 when Team A is away and 1 when Team A is home. The [Weather dummy variable] is 1 when the particular climate match-up of interest is present for the game, and 0 when otherwise.

A general HFA variable (AHome) was included to isolate the effect of weather from the general home field advantage due to travel, psychology, officiating, or other effects.

Several models were run for each climate match-up of interest. The table below lists the statistical significance of the 'weather variable' and the resulting home field advantage calculated from the regression results. Note that the overall HFA rate is 57.5%.

Visitor	Home	p-value	HFA (%)
Dome	Cold	0.03*	88
Dome	Warm	0.28	75
Warm	Warm	0.58	71
Warm	Cold	0.74	76
Mod	Warm	0.11	40
Mod	Cold	0.95	62
Cold	Mod	0.13	49

Accounting for relative team strength, the only truly significant result is for dome teams in cold weather (p=0.03), with an expected home team winning percentage of 88%. This result is confirmed by the actual 86% home team winning percentage when dome teams play at cold cities late in the year. The regression's estimate of 88% suggests that the cold city home teams may have been slightly weaker compared to their dome opponents over the past 5 years.

Two other types of match-ups might be considered marginally significant--moderate teams at warm cities (p=0.11), and cold teams at moderate cities (p=0.13). Both results indicate a reduced HFA later in the season for those match-up types. But since there were 16 combinations of climate match-ups, chances were that we could see one or two type-I errors, i.e. see significance when none is truly there. Because there is no a priori theoretical reason why to expect those results, we shouldn't deem them significant.

Finally, one last model specification was run to make sure no other climate match-up types were significant. The final model included all late-season match-ups types together. No additional types were significant.

It's clear that the situation of a dome team playing playing in a cold city late in the season creates a much stronger HFA than normal. But a larger data set is needed to conclusively analyze other weather match-ups. Dividing 1280 games into 16 types of match-ups and 2 weather periods creates small sub-samples.

A prediction model's accuracy may benefit from enhancing the weight of HFA in certain situations. Looking ahead, there are only three 'dome at cold' match-ups in 2007. STL travels to CIN in week 14, and NO visits CHI and DET visits GB in week 17.

Season Win Projections Week 11

Season win totals and division standing projections are listed below. As before, projections are based on each team's opponent-adjusted generic win probability (GWP). The projections account for future opponent strength, and total wins account for current and projected wins. Methodology is described here.

Team	Rank	Proj GWP	Fut Opp	Proj W
AFC E
NE	1	0.90	0.55	15.4
BUF	19	0.44	0.51	7.7
NYJ	27	0.28	0.57	3.7
MIA	23	0.34	0.54	2.0
AFC N
PIT	5	0.74	0.47	11.4
CLE	17	0.60	0.42	9.6
CIN	20	0.51	0.43	6.1
BAL	24	0.22	0.63	5.3
AFC S
IND	3	0.87	0.52	13.2
JAX	8	0.59	0.52	10.5
TEN	14	0.50	0.52	9.0
HOU	16	0.38	0.55	7.3
AFC W
SD	10	0.65	0.53	8.9
DEN	13	0.62	0.44	8.7
KC	21	0.50	0.52	7.0
OAK	31	0.15	0.52	2.9
Team	Rank	Proj GWP	Fut Opp	Proj W
NFC E
DAL	2	0.87	0.54	14.2
NYG	9	0.60	0.51	10.6
WAS	11	0.51	0.58	8.0
PHI	12	0.43	0.57	7.6
NFC N
GB	7	0.73	0.43	13.4
DET	26	0.26	0.51	7.6
MIN	15	0.59	0.47	7.5
CHI	30	0.24	0.50	5.4
NFC S
TB	4	0.87	0.40	11.2
NO	28	0.35	0.42	6.1
CAR	25	0.30	0.55	5.8
ATL	22	0.32	0.51	4.9
NFC W
SEA	6	0.71	0.38	10.3
ARI	18	0.56	0.41	8.4
STL	29	0.22	0.44	3.3
SF	32	0.21	0.44	3.3

Luckiest NFL Teams After Week 11

Based on opponent-adjusted generic win probability (GWP), the number of expected wins can be estimated for each team. Teams that have won more games than expected can be considered lucky, while teams with fewer wins than expected can be considered unlucky.

The list of NFL teams sorted from luckiest (positive numbers) to unluckiest is posted below. We would expect most teams to be within +/- 1.0 wins. So teams outside that margin can be deemed significantly lucky or unlucky.

Team	GWP	Actual Wins	Expected	Luck
GB	0.64	9	6.4	2.6
DET	0.35	6	3.5	2.5
CHI	0.24	4	2.4	1.6
CLE	0.48	6	4.8	1.2
TEN	0.50	6	5.0	1.0
NYG	0.61	7	6.1	0.9
NE	0.91	10	9.1	0.9
JAX	0.61	7	6.1	0.9
NO	0.33	4	3.3	0.7
DAL	0.85	9	8.5	0.5
BUF	0.45	5	4.5	0.5
ARI	0.45	5	4.5	0.5
CAR	0.37	4	3.7	0.3
BAL	0.38	4	3.8	0.2
HOU	0.48	5	4.8	0.2
SF	0.20	2	2.0	0.0
OAK	0.20	2	2.0	0.0
KC	0.41	4	4.1	-0.1
DEN	0.52	5	5.2	-0.2
PIT	0.74	7	7.4	-0.4
SS	0.65	6	6.5	-0.5
IND	0.85	8	8.5	-0.5
PHI	0.56	5	5.6	-0.6
WAS	0.56	5	5.6	-0.6
SD	0.57	5	5.7	-0.7
STL	0.27	2	2.7	-0.7
ATL	0.38	3	3.8	-0.8
MIN	0.49	4	4.9	-0.9
CIN	0.43	3	4.3	-1.3
NYJ	0.35	2	3.5	-1.5
TB	0.80	6	8.0	-2.0
MIA	0.38	0	3.8	-3.8

Poor Miami. They've played well enough to win about 4 games, but are still waiting for their first victory.

NFL Home Field Advantage by Climate

In this post I'll begin an analysis of home field advantage in the NFL and its relationship to climate. Others have examined the connection between weather and HFA previously, but here I'll attempt to present the data with a clear and novel approach. This post represents the 'clear' part. In following posts, the novel part will use logistic regression to account for team strength and determine the significance of each particular type of climate match-up.

I began by dividing each home city into four categories: dome, cold, moderate, and warm based on a combination of each city's average December high temperature and wind speed. The dome cities include STL, NO, MIN, DET, IND, and ATL. The cold cities are GB, BUF, CLE, CHI, KC, DEN, NYG, NYJ, NE, PHI, PIT, and CIN. Moderate cities include BAL, WAS, CAR, SEA, OAK, SF, TEN, and DAL. The warm cities are MIA, TB, JAX, HOU, SD, and ARI.

Based on all NFL regular season games from the 2002 through 2006 seasons, the winning percentage of the home team was calculated for each type of climate match-up. I divided the season into "early" and "late." The early season is defined as weeks 1 through 12 and the late season is defined as weeks 13 through 17. For example, the winning percentage of the home team in match-ups of cold teams at moderate cities in the early season is 57%, with n=77 examples of such cases.

The home winning percentage of all types of weather match-ups are presented below in a series of pairs of tables. Each table is presented the same way, with the visiting team climate on the left and the home climate on the top. The first table in each pair is for the early season (pre-December), and the second table is for the late season (December games).

Sample size is usually an issue when populations are divided up among several classes. For that reason, the first pair of tables lists the number of cases of each type. For example, the top right cell of the first table lists the number of games featuring dome teams playing at cold cities in the early season. The same cell in the second table lists the same type of match-up in the late season.

Wk 1-12	Dome	Warm	Mod	Cold
Dome	31	29	44	62
Warm	34	20	50	62
Mod	40	49	51	78
Cold	59	66	77	128

Wk 13-17	Dome	Warm	Mod	Cold
Dome	15	17	21	21
Warm	16	11	17	30
Mod	25	20	17	40
Cold	20	28	43	59

The second pair of tables simply lists the straight-up winning percentage of the home team in each type of match-up. For example, the bottom left cell of the first table lists the home team winning percentage when cold teams play at domes in the early season. The same cell in the second table lists home winning percentage of cold teams at domes in the late season.

Wk 1-12	Dome	Warm	Mod	Cold
Dome	61%	45	61	51
Warm	68	50	66	66
Mod	60	59	59	76
Cold	53	52	57	50

Wk 13-17	Dome	Warm	Mod	Cold
Dome	61%	65	52	86
Warm	63	64	59	63
Mod	52	35	53	63
Cold	45	54	42	58

What immediately stands out is the very high winning percentage of cold teams hosting dome teams late in the season. The most remarkable result, however, may be the 35% home winning percentage of warm teams hosting moderate teams. Note that there are only about 20 cases of each type of match-up in the past 5 years, so these results could be due to luck or due to general team strengths of the according type of teams. Perhaps a couple moderate teams have been relatively dominant over warm weather division rivals between '02 and '06.

The final table lists the difference in home winning percentage between late season and early season match-ups. Simply put, it is late season winning percentage minus early season winning percentage. A high positive number indicates cold weather may give an advantage to the home team. A negative number or near-zero number suggests otherwise. We'd expect to see a zero for dome teams at dome teams, because outdoor weather is obviously not a factor. This method begins to account for relative team strengths over the period studied.

Difference	Dome	Warm	Mod	Cold
Dome	0%	20	-9	35
Warm	-5	14	-7	-3
Mod	-8	-24	-6	-13
Cold	-8	2	-15	8

Take dome teams for example. By reading across, we see that dome teams seem to have no greater HFA in late season than the early season against other dome teams--as we'd expect. We also see that they are at a 20% disadvantage playing at warm cities but, for some reason, have a 9% better advantage playing at moderate cities late rather than early. Lastly, we see that dome teams appear to be at a severe disadvantage playing in cold cities late in the season, apparently giving up 35% advantage to the cold.

As mentioned above, some of the observed differences in HFA due to weather may be due to luck and relative team strengths among the weather-classes. The final table is a simple way of accounting for team strength, but it does not address the possibility that the differences are primarily due to luck. The final part of this article will use logistic regression to more powerfully account for team strength and test for statistical significance.

Game Predictions Week 12

Game probabilities for week 11 NFL games are listed below. The probabilities are based on an efficiency win model explained here and here. The model considers offensive and defensive efficiency stats including running, passing, sacks, turnover rates, and penalty rates. Team stats are adjusted for previous opponent strength. Games in which the model disagrees with consensus favorites are highlighted in red.

Vprob	Visitor	Home	Hprob
0.70	GB	DET	0.30
0.06	NYJ	DAL	0.94
0.86	IND	ATL	0.14
0.27	BUF	JAX	0.73
0.71	DEN	CHI	0.29
0.42	HOU	CLE	0.58
0.31	MIN	NYG	0.69
0.37	NO	CAR	0.63
0.21	OAK	KC	0.79
0.78	SEA	STL	0.22
0.48	TEN	CIN	0.52
0.19	WAS	TB	0.81
0.18	SF	ARI	0.82
0.25	BAL	SD	0.75
0.09	PHI	NE	0.91
0.14	MIA	PIT	0.86

I wouldn't put too much stock in the NO-CAR probability. The QB/WR injury situation in CAR needs to be watched, and NO has been so inconsistent.

Week 11 Efficiency Rankings

NFL team efficiency rankings are listed below in terms of generic winning probability. The GWP is the probability a team would beat the league average team at a neutral site. Each team's opponent's average GWP is also listed, which can be considered to-date strength of schedule. GWP modifies the generic win probability to reflect the strength of past opponents. Offensive ranking (O Rank) is based on each team's offensive GWP, i.e. it's the team's GWP assuming it had a league-average defense. D Rank is vice-versa. Rankings are based on a logistic regression model applied to data through week 11. A full explanation of the methodology can be found here.

RANK	TEAM	Last Wk	GWP	Opp GWP	O Rank	D Rank
1	NE	1	0.91	0.49	1	2
2	DAL	3	0.85	0.49	2	5
3	IND	2	0.85	0.54	4	1
4	TB	5	0.80	0.48	3	6
5	PIT	4	0.74	0.43	12	3
6	SEA	6	0.65	0.41	9	16
7	GB	7	0.64	0.47	8	14
8	JAX	9	0.61	0.53	5	22
9	NYG	8	0.61	0.51	14	8
10	SD	13	0.57	0.54	16	10
11	WAS	12	0.56	0.57	19	7
12	PHI	10	0.56	0.49	7	20
13	DEN	15	0.52	0.53	6	29
14	TEN	11	0.50	0.52	28	4
15	MIN	14	0.49	0.50	13	21
16	HOU	18	0.48	0.45	17	25
17	CLE	20	0.48	0.54	10	27
18	ARI	21	0.45	0.49	22	11
19	BUF	17	0.45	0.58	18	15
20	CIN	16	0.43	0.53	11	30
21	KC	26	0.41	0.50	27	9
22	ATL	19	0.38	0.47	23	18
23	MIA	27	0.38	0.53	20	23
24	BAL	22	0.38	0.41	29	13
25	CAR	24	0.37	0.52	24	19
26	DET	25	0.35	0.46	25	12
27	NYJ	28	0.35	0.53	21	31
28	NO	23	0.33	0.52	15	32
29	STL	29	0.27	0.48	26	24
30	CHI	30	0.24	0.53	30	26
31	OAK	31	0.20	0.43	31	28
32	SF	32	0.20	0.51	32	17

Updated NE Undefeated Watch

Pittsburgh and the Giants are obviously the biggest obstacles remaining for the Patriots. Here is the updated probability NE will finish the regular season undefeated.

Vprob	Visitor	Home	Hprob
0.88	NE	BUF	0.12
0.09	PHI	NE	0.91
0.90	NE	BAL	0.10
0.21	PIT	NE	0.79
0.03	NYJ	NE	0.97
0.03	MIA	NE	0.97
0.81	NE	NYG	0.19

0.88*0.91*.90*0.79*0.97*0.97*0.19 = 0.45