It's a little silly to consider these the final rankings, particularly for the top teams. After week 17's meaningless or almost-meaningless games, many of which were blowouts, it makes more sense to consider last week's rankings as truer estimates of team strength. Even for the teams that played full strength games, many of their previous opponents did not, which affects their strength of schedule and ultimately, their own final rankings. For the record, however, here are the 2009 regular season team efficiency rankings.
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
| DAL
| 5
| 0.77
| 0.54
| 2
| 11
|
2
| SD
| 2
| 0.75
| 0.48
| 3
| 18
|
3
| NO
| 3
| 0.74
| 0.49
| 1
| 23
|
4
| PHI
| 4
| 0.73
| 0.54
| 11
| 2
|
5
| IND
| 1
| 0.72
| 0.48
| 8
| 10
|
6
| NE
| 6
| 0.70
| 0.53
| 4
| 21
|
7
| GB
| 9
| 0.68
| 0.41
| 7
| 3
|
8
| PIT
| 10
| 0.64
| 0.44
| 9
| 13
|
9
| NYG
| 7
| 0.63
| 0.57
| 10
| 14
|
10
| NYJ
| 11
| 0.63
| 0.51
| 24
| 1
|
11
| MIN
| 14
| 0.63
| 0.42
| 5
| 19
|
12
| HOU
| 13
| 0.62
| 0.47
| 6
| 20
|
13
| DEN
| 8
| 0.62
| 0.53
| 17
| 4
|
14
| BAL
| 12
| 0.59
| 0.48
| 12
| 5
|
15
| CAR
| 18
| 0.54
| 0.56
| 21
| 8
|
16
| WAS
| 17
| 0.53
| 0.53
| 18
| 12
|
17
| TEN
| 15
| 0.51
| 0.52
| 13
| 17
|
18
| CIN
| 16
| 0.50
| 0.48
| 19
| 7
|
19
| ATL
| 20
| 0.48
| 0.55
| 16
| 24
|
20
| ARI
| 19
| 0.47
| 0.43
| 15
| 16
|
21
| BUF
| 24
| 0.44
| 0.51
| 27
| 6
|
22
| SF
| 22
| 0.43
| 0.44
| 23
| 9
|
23
| JAC
| 21
| 0.41
| 0.47
| 14
| 30
|
24
| MIA
| 23
| 0.39
| 0.58
| 20
| 26
|
25
| CHI
| 25
| 0.35
| 0.46
| 22
| 15
|
26
| TB
| 26
| 0.32
| 0.58
| 28
| 22
|
27
| SEA
| 27
| 0.27
| 0.45
| 25
| 25
|
28
| KC
| 28
| 0.26
| 0.54
| 26
| 28
|
29
| OAK
| 29
| 0.21
| 0.57
| 29
| 27
|
30
| CLE
| 31
| 0.16
| 0.50
| 32
| 29
|
31
| DET
| 32
| 0.15
| 0.49
| 30
| 32
|
32
| STL
| 30
| 0.14
| 0.48
| 31
| 31
|
TEAM
| OPASS
| ORUN
| OINT%
| OFUM%
| DPASS
| DRUN
| DINT%
| PENRATE
|
ARI
| 6.5
| 4.1
| 3.0
| 1.5
| 5.9
| 4.5
| 3.5
| 0.44
|
ATL
| 6.0
| 4.2
| 3.0
| 0.6
| 6.9
| 4.0
| 2.8
| 0.32
|
BAL
| 6.3
| 4.7
| 2.5
| 0.7
| 6.0
| 3.4
| 4.2
| 0.55
|
BUF
| 5.2
| 4.4
| 4.3
| 0.8
| 5.4
| 4.7
| 5.4
| 0.43
|
CAR
| 5.6
| 4.8
| 4.3
| 1.3
| 5.8
| 4.4
| 4.4
| 0.35
|
CHI
| 5.8
| 4.0
| 4.8
| 1.3
| 6.0
| 4.3
| 2.4
| 0.42
|
CIN
| 5.7
| 4.1
| 2.7
| 1.0
| 5.6
| 3.9
| 3.5
| 0.43
|
CLE
| 4.4
| 4.2
| 4.1
| 0.8
| 6.9
| 4.6
| 1.9
| 0.33
|
DAL
| 7.3
| 4.8
| 1.6
| 0.7
| 5.9
| 4.0
| 1.9
| 0.45
|
DEN
| 6.1
| 4.2
| 2.3
| 0.7
| 5.4
| 4.5
| 3.3
| 0.39
|
DET
| 5.0
| 4.0
| 5.5
| 0.9
| 7.4
| 4.4
| 1.7
| 0.37
|
GB
| 6.9
| 4.3
| 1.4
| 0.4
| 5.6
| 3.6
| 5.6
| 0.53
|
HOU
| 7.5
| 3.5
| 2.9
| 1.3
| 6.0
| 4.3
| 2.6
| 0.41
|
IND
| 7.4
| 3.5
| 3.2
| 0.5
| 5.5
| 4.3
| 2.7
| 0.26
|
JAC
| 6.0
| 4.5
| 1.9
| 0.9
| 7.2
| 4.1
| 2.9
| 0.27
|
KC
| 5.0
| 4.4
| 3.2
| 1.8
| 7.0
| 4.7
| 2.9
| 0.32
|
MIA
| 5.5
| 4.4
| 3.5
| 1.3
| 7.0
| 4.2
| 3.1
| 0.31
|
MIN
| 7.1
| 4.1
| 1.3
| 0.9
| 6.0
| 3.9
| 2.1
| 0.38
|
NE
| 7.3
| 4.1
| 2.2
| 0.8
| 6.2
| 4.4
| 3.5
| 0.37
|
NO
| 7.7
| 4.5
| 2.2
| 0.7
| 6.2
| 4.5
| 4.5
| 0.38
|
NYG
| 7.0
| 4.1
| 2.6
| 1.5
| 6.5
| 4.2
| 2.6
| 0.41
|
NYJ
| 5.6
| 4.5
| 5.3
| 1.1
| 4.6
| 3.8
| 3.4
| 0.34
|
OAK
| 4.8
| 4.1
| 3.7
| 1.1
| 7.0
| 4.5
| 1.8
| 0.47
|
PHI
| 6.9
| 4.3
| 2.4
| 0.9
| 5.5
| 4.1
| 4.3
| 0.47
|
PIT
| 7.1
| 4.2
| 2.6
| 1.0
| 5.8
| 3.9
| 2.2
| 0.39
|
SD
| 8.0
| 3.3
| 1.9
| 0.3
| 5.9
| 4.5
| 2.6
| 0.29
|
SF
| 5.4
| 4.3
| 2.7
| 0.7
| 5.9
| 3.6
| 3.1
| 0.38
|
SEA
| 5.4
| 4.0
| 3.1
| 1.4
| 6.6
| 4.2
| 2.3
| 0.41
|
STL
| 4.6
| 4.3
| 3.9
| 0.9
| 7.3
| 4.4
| 1.6
| 0.41
|
TB
| 5.3
| 4.0
| 5.5
| 1.7
| 6.5
| 4.8
| 3.9
| 0.35
|
TEN
| 6.2
| 5.2
| 3.2
| 1.8
| 6.5
| 4.3
| 3.3
| 0.40
|
WAS
| 6.0
| 3.9
| 3.0
| 0.8
| 6.0
| 4.0
| 2.2
| 0.34
|
Avg
| 6.1
| 4.2
| 3.1
| 1.0
| 6.2
| 4.2
| 3.1
| 0.39
|
I hadn't noticed this before, but Minnesota is by far the weakest of the teams with a bye. They have less than a 4% chance of winning it all, according to last weeks efficiency ratings. Intuitively, that seems too low.
It does seem low. It's mostly due to their very weak strength of schedule--2nd weakest in the league. Their W-L record might be a little deceptive.
I'll definitely post the tournament odds on my website later today (although they can easily be calculated by hand).
If I get time later this week, I might re-simulate the whole season using the final power ratings. I think those kinds of analysis are often interesting, in terms of illustrating the stochastic nature of a relatively short 16 game schedule. That type of simulation would allow us to see just how much of an outlier the Vikings season was in terms of their playoff seeding given their "true" team efficiency.
That would be very cool. I'd suggest using the previous ratings (going into wk 17).
I wouldn't take week 17 completely out of the equation.
So losing by 23 to the Bills is costly to the Colts' ranking, eh?
I'd cut off the rankings after 14 games -- in their 15th the Colts threw in the towel for the rest of the regular season while still winning against the Jets, the Saints were in the midst of losing three straight after starting 13-0, etc., the good teams were already tanki^h^h, er, giving their subs valuable opportunities to develop. While the bad and collapsing teams were already mailing it in (see: Giants -- not that we'd want the Postal Service to sue over describing the Gints' season closers using these words).
There's a lot of talk these days about going to an 18-game season. Just imagine.
If they do that, the whole last month of the season will be like this.
Really, there would be two sessions of exhibition games each year. The first would be the two pre-season games where the coaches all use generic game plans to keep from giving secrets away, and only play their subs and depth players to keep their top players from being hurt
... the next would be the last two-to-four games of the regular season, the pre-post-season games, in which the coaches of all the playoff teams use generic game plans to keep from giving secrets away, and only play their subs and depth players to keep their top players from being hurt.
Between the top teams throwing in the towel on the rest of the regular season and the bottom teams who have nothing left to play for throwing
in the towel on everything, for two more weeks, the whole last month of the season would be a mess.
The season should be long enough to reliably determine the division winners, no longer.
If you add a bunch more games that don't give the top (and bottom) teams anything to play for, that's how they'll play them.
16 games is enough, IMHO, maybe too many already.
Hi. Just discovered this site a few days ago, very neat.
If I refer back to last week's efficiency rankings and look at the upcoming playoff matchups, looks like there are some pretty obvious favorites. NYJ, NE, and GB all look pretty clear-cut based on a crude analysis using your numbers. DAL/PHI seems a bit subtler.
Does this sound right to you or am I ignoring some of the important specfics that come with each individual matchup (e.g., who's at home, strength of opp def, strength of opp off, etc)?
Let me know!
whats the chances of doing the raiders before and after russell seeing as they seem to be lower than they should be considering thier strength of schedule.
Brian, these are still based on team cumulative stats, right? Now that you have play-by-play data, do you think the efficiency ratings would benefit at all from using play-by-play stats so you could adjust for score, field position, down/distance, etc., i.e. would that make the ratings any more predictive?
why not throw out meaningless games from the ratings altogether?
Jim,
I hope you are kidding right....
Week 16 was meaningless for only one team in the league....the Colts. Everybody else had something to play for. You could argue it was meaningless for a lot of the teams that were already eliminated, but realistically most of those players knew that they were playing for their own jobs. If you got paid millions of dollars to play for Cleveland this year and football was your only real job skill, would you really feel like you had enough job security to not bother with the last two weeks...
As for week 17, yes it happens on the last week of the year. If you were to extend the season, week 17 would be important again for everybody. Week 18 would be important for almost everybody. Week 19 would replace week 17, but you would get 2 more meaningful weeks. Arguably, several more teams could take off week 18, but there wouldn't be many.
The NFL has a lot of parity. In my opinion, Bucs/Saints proved that. I don't think the Saints were tanking that game. I think they were trying to win it. Home field advantage was still on the line. They didn't rest anybody important. The Saints just had a bad day and the Bucs had a good one. The quality of the NFL is such that even the really bad teams are still made up of elite professional athletes and can easily win with a couple of lucky breaks.
If you think that the playoff position for all teams will be locked in with nearly 25% of the season left to play....
I am curious if it would make sense to develop a ranking by GWP vs a team with the best offense and the best defense rather than the average offense and defense.
After all, the ultimate purpose is to win the Super Bowl, which means beating the best teams, not average teams.
Tournament odds are up on my website.
Colts getting down last week. not happy with week 17 rankings
If these rankings are intended to be predictive, shouldn't you count recent data more heavily than early season data? That's probably one advantage weighted DVOA has over TER.
Not really. I've tested it both ways, and over-weighting recent games doesn't add any predictive power. I know Doug Drinen and some others have found the same result. In some cases it helps, and in others it hurts. I do think it helps quantitative ratings match most people's gut sense of which teams are good and bad, but that's not the goal. On net it's a wash.
I've been doing NBA analysis for more than a year and I can tell you that in ALMOST all cases, recency-weighted data does not have any enhanced predictive capabilities. However, I've also found that this is not a universal statement. In some isolated cases I can actually show where recency-weighted data is more significant (at least I can for the NBA).
For example, I recently created an algorithm that calculates a "Velocity Rating". My Velocity ratings are, currently, the only data that I compile that is weighted towards more recent results. When I tested the rating against 6+ years of historical data, I found a very odd result. The Velocity rating is more accurate than the Vegas spread when predicting the straight-up HOME winners. Curiously, the same rating fails against the Vegas spread when predicting straight-up ROAD winners. I also found that the same rating has a 56% win rate against the spread when the rating difference is significant between the two teams. Strangely, none of my other ratings have the same predictive power against the spread.
Chris,
It would seem that your tournament odds skew the results slightly in favor of the by teams. The reason I think this is because the wild card teams have to win to have a chance to face the bye teams. However, a winning result would probably be the result of strong play in week 18, which in turn would likely improve their efficiency ratings. In other words, whomever the Colts or Chargers face will probably have better efficiency stats after week 18 than they have now.
Am I wrong?
Anon, don't you know that these ratings are something like 70% against the spread over the past eight years? What nerve you have suggesting ideas for improvements! Next, you'll probably ask for the exact formula so you can retire early ^H^H^H^H^H^H^H^H^H^H^H^H^Hvalidate such claims!
Jonathan-You're probably right, but this would have a very small effect in most cases. Only if the underdog really stomped all over their opponents would the probabilities move significantly.
The ranking is not good as i was expecting.the cowboys leading and Colts and Eangles going down:(