This week's stories are all about the shuffling on the top of our rankings.
- While the Texans were able to dispose of the Jaguars and hold on to the number one ranking, there was a ton of movement in the top-12 that cannot go unnoticed.
- While they are still off by .03 GWP, Green Bay is threatening the Texans coming off their bye week. Could a convincing win on the road against the Chargers push them on top of the ladder? Probably not, unless the Texans stumble against the lowly Browns. And that's not going to happen. It just won't.
- Are the Eagles back? Is Andy Reid off the hot seat? The rankings suggest that to be the case. Philadelphia ran all over the Cowboys, and it looks like Mike Vick may finally be learning how to use Brent Celek and the rest of his receiving core. On the defensive side, Jason Babin has been the disruptive force on the defensive line and if the linebacking core can come together, The Dream Team could end up making a playoff run.
- Somehow, the Saints stayed inside the top-5 despite losing to the Rams. I know Brian's formula probably doesn't have this statement in it, so I'm going to make a suggestion: =IF(team loses to Rams, -15 spots in the rankings).
- Despite a big time game that could tell us a lot, New England and Pittsburgh only moved one spot each. So, apparently, it didn't tell us much we didn't already know.
- It may not have anything to do with the top, but Tim Tebow's magic is apparently moving his team down seven spots in the rankings.
So, with all of that in mind, get a load of these rankings:
RANK | TEAM | LAST WK | GWP | Opp GWP | O RANK | D RANK |
1 | HOU | 1 | 0.69 | 0.50 | 3 | 5 |
2 | GB | 3 | 0.66 | 0.47 | 1 | 25 |
3 | PHI | 9 | 0.63 | 0.53 | 4 | 12 |
4 | PIT | 5 | 0.63 | 0.49 | 9 | 8 |
5 | NO | 2 | 0.63 | 0.49 | 7 | 13 |
6 | DAL | 4 | 0.62 | 0.55 | 8 | 11 |
7 | NE | 6 | 0.61 | 0.54 | 2 | 28 |
8 | BUF | 12 | 0.59 | 0.52 | 10 | 14 |
9 | NYG | 8 | 0.59 | 0.47 | 6 | 21 |
10 | DET | 10 | 0.59 | 0.45 | 15 | 1 |
11 | BAL | 7 | 0.58 | 0.52 | 16 | 2 |
12 | SF | 16 | 0.53 | 0.48 | 13 | 10 |
13 | SD | 11 | 0.53 | 0.45 | 11 | 17 |
14 | NYJ | 13 | 0.52 | 0.52 | 22 | 4 |
15 | CHI | 15 | 0.52 | 0.50 | 20 | 7 |
16 | TEN | 17 | 0.49 | 0.49 | 12 | 23 |
17 | WAS | 14 | 0.48 | 0.54 | 27 | 9 |
18 | OAK | 20 | 0.47 | 0.51 | 14 | 15 |
19 | STL | 22 | 0.46 | 0.60 | 28 | 6 |
20 | CAR | 18 | 0.46 | 0.49 | 5 | 31 |
21 | JAC | 21 | 0.45 | 0.55 | 32 | 3 |
22 | CIN | 19 | 0.43 | 0.44 | 18 | 26 |
23 | KC | 28 | 0.42 | 0.50 | 26 | 18 |
24 | ATL | 26 | 0.41 | 0.51 | 25 | 19 |
25 | MIN | 24 | 0.41 | 0.49 | 23 | 22 |
26 | MIA | 25 | 0.39 | 0.53 | 24 | 29 |
27 | ARI | 27 | 0.39 | 0.50 | 17 | 27 |
28 | CLE | 30 | 0.38 | 0.43 | 29 | 20 |
29 | SEA | 32 | 0.37 | 0.48 | 31 | 16 |
30 | DEN | 23 | 0.37 | 0.51 | 30 | 24 |
31 | IND | 29 | 0.35 | 0.50 | 21 | 32 |
32 | TB | 31 | 0.35 | 0.49 | 19 | 30 |
TEAM | OPASS | ORUNSR% | OINT% | OFUM% | DPASS | DRUNSR% | DINT% | PENRATE |
ARI | 6.0 | 45 | 3.9 | 1.2 | 7.0 | 54 | 1.9 | 0.57 |
ATL | 5.7 | 39 | 3.2 | 1.1 | 7.2 | 57 | 3.7 | 0.44 |
BAL | 5.9 | 39 | 2.3 | 2.1 | 4.8 | 61 | 3.1 | 0.45 |
BUF | 7.0 | 44 | 3.5 | 0.3 | 6.9 | 53 | 5.5 | 0.35 |
CAR | 7.5 | 45 | 3.1 | 0.7 | 7.4 | 53 | 1.8 | 0.57 |
CHI | 6.2 | 40 | 2.6 | 0.6 | 6.3 | 59 | 2.8 | 0.45 |
CIN | 6.1 | 40 | 3.0 | 0.3 | 5.8 | 58 | 1.3 | 0.41 |
CLE | 5.0 | 41 | 1.7 | 0.8 | 5.7 | 55 | 2.1 | 0.40 |
DAL | 7.1 | 41 | 3.4 | 1.4 | 6.1 | 57 | 2.8 | 0.38 |
DEN | 5.0 | 41 | 3.5 | 2.4 | 7.0 | 61 | 1.3 | 0.36 |
DET | 6.6 | 36 | 1.3 | 0.3 | 4.8 | 62 | 3.7 | 0.50 |
GB | 8.8 | 41 | 1.7 | 1.1 | 7.0 | 53 | 4.8 | 0.32 |
HOU | 7.5 | 40 | 2.0 | 0.7 | 5.3 | 57 | 3.8 | 0.42 |
IND | 5.4 | 43 | 1.8 | 2.7 | 7.6 | 52 | 1.1 | 0.31 |
JAC | 4.1 | 38 | 3.7 | 1.1 | 5.7 | 62 | 2.7 | 0.33 |
KC | 6.0 | 37 | 4.4 | 2.0 | 7.2 | 55 | 5.5 | 0.44 |
MIA | 5.6 | 41 | 3.5 | 1.1 | 7.3 | 56 | 0.8 | 0.30 |
MIN | 5.7 | 43 | 1.7 | 0.8 | 6.8 | 54 | 2.0 | 0.47 |
NE | 7.9 | 49 | 2.9 | 0.5 | 7.5 | 54 | 3.2 | 0.42 |
NO | 7.2 | 48 | 2.9 | 0.6 | 5.7 | 54 | 1.4 | 0.29 |
NYG | 7.9 | 37 | 2.1 | 1.1 | 6.0 | 53 | 3.4 | 0.41 |
NYJ | 5.9 | 37 | 2.6 | 2.2 | 5.8 | 57 | 5.0 | 0.45 |
OAK | 6.5 | 43 | 4.7 | 0.9 | 6.0 | 55 | 2.4 | 0.65 |
PHI | 7.3 | 50 | 4.5 | 1.6 | 6.2 | 56 | 3.7 | 0.39 |
PIT | 7.0 | 44 | 2.4 | 1.4 | 4.9 | 55 | 0.8 | 0.43 |
SD | 7.2 | 45 | 4.2 | 1.3 | 6.2 | 59 | 4.1 | 0.45 |
SF | 5.9 | 42 | 1.1 | 1.1 | 6.0 | 62 | 3.2 | 0.50 |
SEA | 5.1 | 34 | 2.8 | 1.9 | 6.7 | 66 | 3.4 | 0.48 |
STL | 4.7 | 40 | 1.1 | 2.1 | 6.3 | 53 | 2.5 | 0.46 |
TB | 5.9 | 45 | 3.6 | 0.6 | 7.5 | 56 | 2.9 | 0.55 |
TEN | 6.5 | 32 | 2.4 | 0.6 | 5.9 | 53 | 3.0 | 0.49 |
WAS | 5.8 | 44 | 4.8 | 1.2 | 6.3 | 58 | 2.7 | 0.37 |
Avg | 6.3 | 41 | 2.9 | 2.1 | 6.3 | 57 | 2.9 | 0.43 |
These rankings fail to comport with my pre-conceived notions. Please adjust accordingly. Thank you.
" =IF(team loses to Rams, -15 spots in the rankings)."
LOL
Still hard to get used to Houston being number 1.
Maybe I missed this before, but I noticed you had St. Louis with a 41% chance to win last week and had them ranked above 10 other teams. How did you know they weren't as bad as their winless record suggested?
Matt -- They aren't as bad as other people think. Not at all.
My comment was mostly a joke.
So if the Cardinals lose to the Rams, they should be #42?
I question the methodology of a ranking system that places Houston above Green Bay and Tampa Bay below Miami, Denver, and Indianapolis. The numbers aren't necessarily lying but they certainly don't appear to be a valid indicator of strength in the league. I guess we'll have to wait and see...
Any consideration to including 2010 numbers? It increases data and is often relevant due to the same cast of important players.
I have a really, really difficult time buying the notion that Tampa Bay is the worst team in the NFL and Houston is the best. Especially the Tampa Bay bit.
If you've found this page from a co-worker and just want to comment about how you don't believe the validity of the ranking system. Do everyone a favor and read this first: http://www.advancednflstats.com/2011/10/prediction-accuracy.html
I absolutely believe in the effectiveness of this model, but I don't understand how the model is putting TB below IND. They are just ranked one place below, but TB has better offensive and defensive rankings than IND and is placed below them. Is it their slightly lower strength of schedule which accounts for this?
Just curious. Love your site Brian.
Thanks Patrick,
every week an "Anonymous" comes over from Espn etc. to unload his "trash". Worst part of it is that they think they are real clever. It´s annoying. I am wondering why Brian does not delete this nonsense posts.
It´s kind of funny that SF is rated that low. It just shows how weak the NFC West every year is. It´s really a lottery win for just average teams to play there. Hopefully SF makes an early playoff exit like ATL did last year. B/C it´s really not justified that they get a bye week. I don´t like soccer, but in our country the standings are much more fair since "we" don´t have divisions.
Karl, Germany
Another interesting thing is that all the Martz haters which rove over the internet are absolutely wrong when they call for his head. Actually Y/PP (15th) is better than RUNSR (21st). Yet those guys demand more running plays, even tough Martz is as conservative as he can get since he found out his OL can´t pass block and the QB has happy feet. No need to mention that his offense performs better than his predecessors, as i expected.... Omg, those guys will miss him when CHI needs another 20 years for the next SB-Run.
Karl, Germany
I realize that the model is probably as accurate as it can be and all, and that good stats should surprise you 20% of the time, but as someone new to the site, I'd like to see a post about why the Bucs are dead last. That just seems quite odd to me.
Nate, I think that one big reason why TB is lower than IND in the rankings is because TB's penalty rate is so much higher than Indy's. Penalty rate isn't factored into off/def ranking.
Karl from Germany, the craziest thing about the niners is they haven't even had to play their division yet. They have 5 division games left. Despite that, nfl-forecast (which uses Brian's numbers) predicts the niners to finish 5-4. It also says that the niners are probably not going to get the 2nd seed. Their biggest competitor is New Orleans.
TB is last due to very poor defensive pass efficiency, a below average offense and a very high penalty rate. They're really bad at the things that both matter a lot and tend to be consistent. They're decent in the things that aren't good predictors or don't tend to be consistent.
Even though they are 4-3, they have a -38 point differential. The model is saying they are a true .350 team, which would put them at 2.5 wins right now. Their actual record isn't that far off.
That said, I also doubt they are the worst team in the league. But I also believe they are not nearly as good as they are given credit.
Brian (or other qualifed ANFLS writer), could you comment on Cincinnati being ranked 26th for defense?
They are eighth in yards/pass allowed and third in yards/rush allowed, third in yards/play allowed, and 3rd in scoring defense (I am aware that of these stats only yards/pass allowed is included in your model).
It is strange that they are second in the league in yards/rush allowed (3.3) but are near the bottom in success rate against the run.
It appears I was misinterpreting the success rate against the run (ie I thought it represented their opponent's success rate of running against them).
Have you ever explained how to transform these numbers into a given Game Probability? Is it a straightforward comparison with a few percentage points given for home field?
Speaking of which, why no Game Probabilities this week?
Karl from Germany, the division SF plays in is pretty much irrelevant at this point, since they've only played one game in the division.
The 49ers and also the Jets are ranked low by this system because it doesn't account for special teams at all, and those two teams have the league's two best special teams units.
What would be interesting to see is how each component of the second table contributes to the team's GWP (relative to a default of 0.50). In other words, a "+/-WPA" for each efficiency stat.
For example, Brian's coefficient for penalty rate is -1.587.
So, the "WPA" for Tampa Bay's Penalty Rate is:
1/(1 + exp(-(0.55-0.43)*(-1.587))) - 0.5
= -0.05
So, an otherwise average team with Tampa Bay's penalty rate would have a GWP of 0.45.
I know the WPA's wouldn't add exactly to the final GWP due to both the arithmetic of logistic regression as well as the strength of schedule adjustment. But it would add a bit more clarity as to exactly what is driving each team's ranking.
Anonymous: I think this should duplicate the GWP calc:
GWPH = GWP of home team
GWPA = GWP of away team
GWP of home team = 1/(1 + exp( ln(GWPA/(1-GWPA)) - ln(GWPH/(1-GWPH)) - 0.46))
where "ln()" is the natural logarithm and
"exp()" is the natural exponent
So, Indy should have a GWP of 0.55 over Atlanta.
Brian, have you considered estimating a "margin for error" on the scale (maybe you've done this in the past) that might add some context to rankings people question (like Houston and TB this week, and other teams on other weeks)?
Maybe it's .05, so that you could say something like "I doubt Tampa Bay is the worst team in the league, but I'm confident they're among the worst seven teams because the margin for error on this scale is .05"
If it were .05, that would still lock in Houston in the top two. What level of confidence do you have that Houston is among the top two? How confident are you that San Francisco isn't among the top 10 (the Niners are .06 behind Detroit in the 10th spot).
Maybe 0.10 is more accurate...so that one could be confident that Houston is among the top 10, but not really certain with a high level of confidence where within that 10 they'd rank. That would also give San Francisco a shot to be within the top 10 based on their current level of performance.
Is this something you've played with in the past? What's your gut estimate as to where the right margin for error would be? Might add some context to your article on predictions... because we're dealing with a meaningful margin of error when trying to make predictions. Does .05 sound right, or is that too tight? Would you be more comfortable at .10? Is there a way to factor that into the presentation to better portray the reality of the NFL?
Zach -- I wasn't referring to your comment -- I'm genuinely interested in how you knew before last week that the Rams weren't so bad.
Anonymous: there are several posts which describe it scattered through the website. Here are the ones I've got saved:
NFL Win Prediction Methodology (2007 model) - Background
Game Model Coefficients (2007 model) - If you want the numbers to work examples using old model
How the model works: A Detailed Example Part 1 (2007 model)
How the model works: A Detailed Example Part 2 (2007 model)
Instability Compensation
2011 Update to the model (run success rate replaces run yards/attempt) - actual title is "Team Rankings Week 4"
Coefficients for 2011 model - actual title is "Weekly Game Probabilities"
"Jared Doom said...
It appears I was misinterpreting the success rate against the run (ie I thought it represented their opponent's success rate of running against them)."
I made this mistake as well (if it is a mistake).
Is there some place these terms are defined, they are not in the glossary. That'd be helpful, thanks.
hey mike, sam, thanks for this. So, what do y'all think: Are the Colts really more likely than not to win this week?
Matt - I think it comes down to strength of schedule. If you look at the Week 8 team rankings, STL's Opponent GWP was the highest in the NFL at 0.58. That means that although STL put up poor efficiency stats (and a poor W/L record), they did that against strong teams. Their 0-6 record included games against Green Bay, Philly, NY Giants, Baltimore and Dallas. All of those teams were in the top 10 in the week 8 efficiency rankings.
I'm confused by the Bengals having the 26th best defense. They are 4th in the league in yards per game given up, so I assumed they had a good defense. Looking at the breakdown, their DPass is 8th best and Drun SR is 10th best.
Is it their strength of schedule? I just looked it up, and they have played a bunch of very weak offenses (Cle, Den, SF, Buf, Jax, Ind, Sea). I'm just surprised that makes the difference between 4th and 26th best defense.