The team rankings below are 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, defensive interception rates, and team penalty rates. If you're scratching your head wondering why a team is ranked where it is, just scroll down to the second table to see the stats of all 32 teams.
Click on the table headers to sort:
RANK | TEAM | LAST WK | GWP | Opp GWP | O RANK | D RANK |
1 | NO | 1 | 0.80 | 0.45 | 1 | 16 |
2 | IND | 2 | 0.78 | 0.51 | 5 | 5 |
3 | SD | 3 | 0.78 | 0.48 | 2 | 19 |
4 | PHI | 4 | 0.73 | 0.51 | 9 | 3 |
5 | DEN | 6 | 0.72 | 0.56 | 16 | 2 |
6 | NYG | 7 | 0.71 | 0.56 | 6 | 9 |
7 | NE | 5 | 0.71 | 0.54 | 4 | 22 |
8 | DAL | 8 | 0.71 | 0.51 | 3 | 17 |
9 | GB | 10 | 0.64 | 0.38 | 11 | 4 |
10 | PIT | 9 | 0.63 | 0.42 | 10 | 14 |
11 | NYJ | 12 | 0.61 | 0.49 | 23 | 1 |
12 | BAL | 14 | 0.60 | 0.50 | 12 | 7 |
13 | MIN | 11 | 0.59 | 0.40 | 7 | 20 |
14 | HOU | 13 | 0.59 | 0.51 | 8 | 21 |
15 | TEN | 18 | 0.59 | 0.55 | 13 | 13 |
16 | WAS | 15 | 0.55 | 0.48 | 18 | 10 |
17 | CIN | 16 | 0.51 | 0.46 | 19 | 6 |
18 | ARI | 17 | 0.50 | 0.47 | 15 | 18 |
19 | JAC | 19 | 0.50 | 0.47 | 14 | 26 |
20 | SF | 20 | 0.42 | 0.48 | 21 | 8 |
21 | ATL | 25 | 0.41 | 0.55 | 17 | 29 |
22 | BUF | 21 | 0.40 | 0.48 | 26 | 11 |
23 | MIA | 24 | 0.39 | 0.57 | 20 | 25 |
24 | CAR | 23 | 0.39 | 0.52 | 25 | 12 |
25 | CHI | 22 | 0.34 | 0.45 | 22 | 15 |
26 | SEA | 26 | 0.31 | 0.45 | 24 | 23 |
27 | TB | 27 | 0.26 | 0.57 | 29 | 24 |
28 | KC | 29 | 0.22 | 0.58 | 31 | 27 |
29 | OAK | 30 | 0.19 | 0.60 | 30 | 28 |
30 | STL | 28 | 0.18 | 0.50 | 27 | 31 |
31 | CLE | 32 | 0.15 | 0.54 | 32 | 30 |
32 | DET | 31 | 0.13 | 0.50 | 28 | 32 |
And here are the sortable raw team efficiency stats. Passing, running, and penalties are in yards per relevant play. Fumbles and interception stats are in turnovers per relevant play.
TEAM | OPASS | ORUN | OINT% | OFUM% | DPASS | DRUN | DINT% | PENRATE |
ARI | 6.6 | 4.2 | 2.9 | 1.7 | 6.0 | 4.5 | 3.0 | 0.44 |
ATL | 6.1 | 4.2 | 3.2 | 0.8 | 7.2 | 4.3 | 1.8 | 0.33 |
BAL | 6.4 | 4.6 | 2.5 | 0.8 | 6.1 | 3.5 | 3.8 | 0.56 |
BUF | 5.2 | 4.3 | 4.4 | 0.9 | 5.4 | 5.0 | 5.9 | 0.38 |
CAR | 5.3 | 4.8 | 5.1 | 1.6 | 6.0 | 4.6 | 4.5 | 0.37 |
CHI | 5.8 | 3.9 | 4.8 | 1.3 | 5.8 | 4.4 | 2.6 | 0.47 |
CIN | 5.9 | 4.0 | 2.5 | 1.2 | 5.6 | 3.8 | 3.3 | 0.45 |
CLE | 4.2 | 3.8 | 3.8 | 0.9 | 7.1 | 4.4 | 1.5 | 0.35 |
DAL | 7.2 | 4.9 | 1.6 | 0.9 | 6.2 | 4.0 | 2.0 | 0.46 |
DEN | 6.1 | 4.3 | 2.1 | 0.9 | 5.2 | 3.9 | 3.1 | 0.36 |
DET | 5.0 | 3.8 | 5.3 | 1.0 | 7.6 | 4.5 | 1.8 | 0.40 |
GB | 6.7 | 4.3 | 1.6 | 0.6 | 5.3 | 3.6 | 5.4 | 0.58 |
HOU | 7.3 | 3.4 | 3.1 | 1.3 | 6.2 | 4.4 | 2.6 | 0.46 |
IND | 7.5 | 3.8 | 3.0 | 0.5 | 5.6 | 4.1 | 3.0 | 0.28 |
JAC | 6.1 | 4.6 | 1.4 | 0.9 | 6.8 | 4.0 | 2.9 | 0.30 |
KC | 4.6 | 3.9 | 3.0 | 1.9 | 7.2 | 4.5 | 2.2 | 0.31 |
MIA | 5.3 | 4.5 | 2.9 | 1.3 | 6.9 | 4.1 | 3.0 | 0.30 |
MIN | 6.9 | 4.2 | 1.3 | 1.0 | 5.9 | 4.1 | 2.1 | 0.38 |
NE | 7.3 | 4.1 | 2.2 | 0.7 | 6.2 | 4.4 | 3.3 | 0.38 |
NO | 8.3 | 4.5 | 2.3 | 0.8 | 6.0 | 4.4 | 4.9 | 0.42 |
NYG | 7.2 | 4.3 | 2.5 | 1.5 | 6.2 | 4.0 | 2.5 | 0.41 |
NYJ | 5.7 | 4.6 | 5.5 | 1.2 | 4.7 | 3.9 | 3.6 | 0.33 |
OAK | 4.6 | 3.9 | 3.5 | 1.2 | 7.2 | 4.6 | 2.2 | 0.44 |
PHI | 6.9 | 4.4 | 2.2 | 0.6 | 5.6 | 3.8 | 4.3 | 0.47 |
PIT | 6.9 | 4.2 | 2.8 | 0.9 | 5.6 | 3.7 | 1.8 | 0.39 |
SD | 8.1 | 3.2 | 1.7 | 0.2 | 5.8 | 4.3 | 2.6 | 0.30 |
SF | 5.3 | 4.4 | 2.5 | 0.7 | 6.1 | 3.7 | 2.7 | 0.39 |
SEA | 5.5 | 3.8 | 2.0 | 1.5 | 6.4 | 4.2 | 2.3 | 0.40 |
STL | 4.9 | 4.5 | 3.7 | 0.8 | 7.2 | 4.6 | 2.1 | 0.46 |
TB | 5.0 | 4.1 | 5.5 | 1.9 | 6.7 | 4.7 | 3.6 | 0.35 |
TEN | 6.2 | 5.3 | 2.8 | 1.7 | 6.4 | 4.2 | 3.2 | 0.38 |
WAS | 6.2 | 3.9 | 2.9 | 0.8 | 5.6 | 4.1 | 2.3 | 0.36 |
Avg | 6.1 | 4.2 | 3.0 | 1.1 | 6.2 | 4.2 | 3.0 | 0.40 |
As a complete shot in the dark first impression it looks to me like the whole NFCE is about 3 spots too high. PIT looks a bit high to me too.
I am surprised MIN and GB are quite that low, but they have played easy schedules.
Anyway interesting stuff as always. If I have some more down time at work I might try to dig into the numbers a bit to examine why some of them are so far off my intuitions.
In the case of MIN DRANK I might point to the large amount of time they have spent in a semi-prevent when ahead.
Is there any way to get from GWP to a specific game probability? I use your weekly game probablities for betting purposes when they come out, but Thursday is kind of late in the week. Any correlation between GWP and game probablity?
Sure. You can do it precisely with some math, but the quick way is to just compare GWP. Throw in home field, which is worth about 0.09 for even match-ups and worth 0.05 for mismatches.
Alright, so lets take the DAL @ NO game as an example. NO has a .80 GWP, and DAL has .71, and the game's in NO. So that's a .09 advantage based on GWP and a .09 GWP for NO being at home.
Would that make it a .59 chance NO wins and .41 that DAL wins, as an approximation?
Is the math to get it precisely somewhere on this site?
Nate, you can use the log5 method to find estimated win probability. http://www.sonicscentral.com/apbrmetrics/viewtopic.php?t=14&highlight=log5
So the Saints at .80, facing the Cowboys at .71, with a .59 HFA yields a 70% chance the Saints win. As an approximation, for close matchups you can simply do:
GWP_team1 - GWP_team2 + .500 + HFA
So the WP for the Saints:
.80 - .71 + .5 + .09 = 68%.
Thanks, Zach. That's just about on the money! It'll be .71 I think when the numbers come out.
Thanks guys, appreciate it.
How heavily to do you weight, if at all, recent games over earlier games?
Pittsburgh is not playing like a top 10 team recently, but they looked very good early in the year.
No over-weighting of recent games in the model this year. I've compared both methods and there was no detectable improvement in prediction accuracy when over-weighting. PIT started slow, got really good, and is now falling back down to earth.
Brian, have you ever toyed with using data from past seasons to see if it increases correlation? For example, doing some sort of linear combination of 2009 passing efficiency with 2008 passing efficiency, etc. I feel that model would be less succeptible to regression to the mean since most NFL data (from season to season) is relevant to the current team. At what point do you think increasing the dataset (in the NFL) compromises the process of predicting the current teams' future statistics? I ask because predicting MLB team success includes a significant percentage of previous (before that season) data.
Regarding NFC East and AFC west teams being ranked 3,4,5,6, and 8. Those teams have performed incredibly poorly when playing other top teams
The NFC East teams listed in the top 8 are 0-9 and negative 102 pts when playing teams listed in your top 15. I really doubt that if they were the 4th, 6th and 8th best teams in the NFL that they would produce such weak results against other top 1/2 teams.
Additionally SD and Den - the top teams the NFC East has lost most of the 9 games I mentioned (5) have preformed very poorly against your top 15 teams other than beating up on the NFC East.
Their record against your top 15 when not playing each other or the NFC East 1-5 - negative 65 pts.
Hi Brian,
I have been following you blog for a while now and as a big NFL fan. I must say that i have found your site a great help with my NFL betting. Just wanted to say Many Thanks for your insite.
Many Thanks
Mr Sport UK