Horseshoes in Houston
The Houston Texans either have the worst luck in the world, or they need to fire their entire strength and conditioning team. Can T.J. Yates step in and succeed in his first NFL season as a fifth-round draft pick? Doubtful. Luckily for the Texans, they still have games against Carolina and Indianapolis remaining, so Yates will have some warmup time if Houston decides that he is the answer. There's also a chance that Kubiak and Ko. get the chance to bring in Sage Rosenfels to fill the void for the rest of the year, but that would involve the Miami Dolphins agreeing to release him and let him be on his merry way.
Tim Tebow Watch
The 6-5 Broncos are still ranked only 23rd, but that's up six spots from last week. The Denver Tebow's have scored more than 18 points just once in Tim's six starts, and while the Broncos defense has some good players on it, the offense keeping them off the field for long stretches just won't be enough to beat good teams. However, the Broncos only play one team in the top-half of the league the rest of the way, so expect the Tebow hype to continue all the way to the playoffs.
Clawless Jags
The Jacksonville Jaguars fired Jack Del Rio this morning, despite the team having the second best defense in the league. The team's decision to let go of David Garrard may very well have kept them from making the playoffs this year, but whether or not that was Del Rio's decision is unclear. Whoever takes over the Jags next season will need to find another weapon at Wide Receiver if they believe that Blaine Gabbert is their man.
And now, without any further ado, here are your full Week Thirteen rankings.
RANK | TEAM | LAST WK | GWP | Opp GWP | O RANK | D RANK |
1 | HOU | 1 | 0.82 | 0.48 | 3 | 3 |
2 | GB | 3 | 0.74 | 0.45 | 1 | 26 |
3 | PIT | 2 | 0.72 | 0.47 | 7 | 5 |
4 | NE | 5 | 0.69 | 0.53 | 2 | 27 |
5 | NO | 6 | 0.69 | 0.49 | 4 | 18 |
6 | DAL | 4 | 0.65 | 0.50 | 5 | 17 |
7 | PHI | 7 | 0.62 | 0.53 | 9 | 16 |
8 | NYG | 8 | 0.62 | 0.51 | 6 | 22 |
9 | DET | 9 | 0.61 | 0.48 | 18 | 1 |
10 | BAL | 12 | 0.59 | 0.52 | 13 | 7 |
11 | CHI | 11 | 0.57 | 0.53 | 20 | 4 |
12 | OAK | 13 | 0.55 | 0.48 | 10 | 10 |
13 | SF | 10 | 0.53 | 0.49 | 16 | 6 |
14 | ATL | 15 | 0.53 | 0.48 | 14 | 15 |
15 | NYJ | 14 | 0.52 | 0.53 | 23 | 8 |
16 | CIN | 18 | 0.50 | 0.44 | 12 | 23 |
17 | BUF | 17 | 0.48 | 0.53 | 15 | 24 |
18 | MIA | 20 | 0.48 | 0.52 | 21 | 19 |
19 | SD | 16 | 0.45 | 0.47 | 11 | 30 |
20 | WAS | 21 | 0.44 | 0.49 | 24 | 11 |
21 | TEN | 19 | 0.44 | 0.48 | 19 | 14 |
22 | JAC | 25 | 0.40 | 0.55 | 32 | 2 |
23 | DEN | 29 | 0.38 | 0.51 | 27 | 13 |
24 | CLE | 22 | 0.38 | 0.47 | 25 | 21 |
25 | CAR | 23 | 0.38 | 0.48 | 8 | 31 |
26 | TB | 30 | 0.36 | 0.56 | 17 | 29 |
27 | STL | 26 | 0.35 | 0.52 | 31 | 9 |
28 | SEA | 24 | 0.34 | 0.51 | 28 | 12 |
29 | MIN | 28 | 0.34 | 0.50 | 26 | 20 |
30 | ARI | 27 | 0.33 | 0.48 | 22 | 25 |
31 | KC | 31 | 0.27 | 0.49 | 30 | 28 |
32 | IND | 32 | 0.23 | 0.50 | 29 | 32 |
TEAM | OPASS | ORUNSR% | OINT% | OFUM% | DPASS | DRUNSR% | DINT% | PENRATE |
ARI | 5.6 | 43 | 4.4 | 1.4 | 6.5 | 57 | 2.2 | 0.52 |
ATL | 6.5 | 42 | 2.5 | 1.0 | 6.6 | 62 | 2.8 | 0.41 |
BAL | 6.0 | 36 | 1.9 | 1.9 | 5.5 | 60 | 3.3 | 0.39 |
BUF | 6.3 | 43 | 4.0 | 0.6 | 7.1 | 55 | 4.4 | 0.38 |
CAR | 6.9 | 44 | 3.6 | 0.9 | 7.5 | 54 | 2.8 | 0.60 |
CHI | 6.3 | 38 | 2.8 | 1.0 | 6.2 | 61 | 3.5 | 0.43 |
CIN | 6.3 | 40 | 3.2 | 0.4 | 5.9 | 60 | 1.6 | 0.40 |
CLE | 5.2 | 40 | 2.0 | 0.9 | 5.7 | 52 | 1.9 | 0.39 |
DAL | 7.3 | 41 | 2.8 | 1.1 | 6.2 | 56 | 3.7 | 0.45 |
DEN | 5.2 | 42 | 2.7 | 1.6 | 6.3 | 59 | 1.9 | 0.42 |
DET | 6.4 | 38 | 2.9 | 0.7 | 5.3 | 63 | 3.9 | 0.50 |
GB | 8.5 | 41 | 1.4 | 0.9 | 6.9 | 53 | 5.0 | 0.30 |
HOU | 7.5 | 42 | 1.9 | 0.7 | 4.9 | 60 | 4.1 | 0.42 |
IND | 5.2 | 39 | 2.8 | 2.4 | 7.5 | 52 | 1.5 | 0.29 |
JAC | 4.1 | 38 | 3.1 | 0.8 | 5.5 | 63 | 2.9 | 0.36 |
KC | 5.5 | 37 | 4.4 | 1.4 | 7.3 | 57 | 4.3 | 0.40 |
MIA | 6.2 | 37 | 2.7 | 1.1 | 6.5 | 57 | 2.0 | 0.35 |
MIN | 5.5 | 43 | 2.5 | 1.0 | 6.8 | 56 | 1.6 | 0.47 |
NE | 8.0 | 46 | 2.4 | 0.8 | 7.2 | 54 | 3.6 | 0.40 |
NO | 7.4 | 49 | 2.4 | 0.5 | 6.2 | 56 | 1.4 | 0.38 |
NYG | 7.6 | 36 | 2.5 | 1.3 | 6.5 | 51 | 3.6 | 0.38 |
NYJ | 5.9 | 41 | 2.9 | 2.1 | 6.1 | 59 | 3.7 | 0.40 |
OAK | 7.0 | 40 | 4.5 | 0.9 | 5.7 | 55 | 3.3 | 0.65 |
PHI | 6.8 | 50 | 4.5 | 1.2 | 6.6 | 59 | 3.1 | 0.41 |
PIT | 7.0 | 43 | 2.6 | 1.4 | 5.1 | 59 | 1.9 | 0.42 |
SD | 6.8 | 42 | 4.1 | 1.2 | 7.2 | 59 | 3.4 | 0.39 |
SF | 6.0 | 40 | 1.7 | 0.9 | 5.9 | 64 | 3.6 | 0.52 |
SEA | 5.4 | 33 | 3.6 | 1.1 | 6.4 | 66 | 3.0 | 0.57 |
STL | 4.7 | 39 | 1.5 | 2.3 | 6.0 | 57 | 2.9 | 0.46 |
TB | 6.0 | 43 | 3.8 | 1.3 | 7.5 | 54 | 3.1 | 0.51 |
TEN | 6.2 | 34 | 2.6 | 1.0 | 5.9 | 57 | 2.5 | 0.54 |
WAS | 6.0 | 40 | 4.5 | 1.3 | 6.2 | 60 | 2.3 | 0.43 |
Avg | 6.3 | 41 | 3.0 | 1.2 | 6.3 | 58 | 3.0 | 0.43 |
I find it rather amazing Dallas, Philadelphia, and New York are all ranked #6-8 considering how each's season has gone and the perception around them. Everyone's writing the Eagles off as if they're terrible, talk is swirling about a Giants second half swoon and Coughin potentially being fired, and the Cowboys are shooting up Power Rankings despite barely pulling out consecutive last second wins over two 3-7 teams.
Also, I'm amused that after early in the season when the NFC was nearly set and the AFC a mess that things have changed so much.
The AFC now has four playoff slots filled with Oakland heavy favorites for the 5th, but the seeding order is WIDE OPEN as four teams have roughly 20-30% shots of getting the #1 seed.
Meanwhile, the NFC only has three spots filled with Dallas equally heavy favorites for the fourth, but #1 and #2 seeding is nearly set and at the other end it's a dogfight for the Wild Cards with Detroit, Chicago, and Atlanta all having roughly equal chances of making the postseason.
I know it is easy to cherry pick examples, but it does seem like your model overvalues the Eagles and undervalues the Ravens. I realize that the Ravens are flawed and that losing games to Jacksonville and Seattle will cost them, but they have also beaten Pittsburgh and Houston (the third and first ranked teams) by a combined 87-41. How does the model deal with "bad games" and have you ever broken it down by home/away splits.
Great question!
With rare exceptions, tems are always just 1 game from even on home-away splits.
Also curious about Detroit's number one ranked defense, which doesn't really agree with the eye test.
I think whichever system Pro-Football-Reference uses has them as very slightly below average (which doesn't really agree with the eye test, either).
I have to agree with the first Anonymous poster on the model's treatment of the Ravens. The model seems to disregard Strength of Victory (SOV) entirely. SOV = the combined winning percentage of the opponents a team has beaten. SOV is a key NFL tie breaking procedure for playoff seeding and, possibly, the only one useful for ranking teams in different conferences against one another. If you run the calculations, the Ravens SOV is tied with the Packers for the best in the NFL at 43%. For comparison, Pittsburgh's SOV is 29%. Unless the model for ranking teams is only concerned with fantasy football stats, and not with winning, it seems seriously flawed as a basis for comparing the strength of one team versus another. Does anyone really believe the Eagles are a far better team at this point than the Ravens? Your model appears to treat stats in all games with equal dignity. Should the model be more concerned with the quality of opponents when ranking teams (at least ones with the same number of wins) since one team's won-loss record may be inflated due to wins against poor quality opponents and another's may be understated from wins against better quality teams? Isn't winning the ultimate stat?
Strength of schedule is fully accounted for. Bal's Sov comes mostly from 2 blowouts. They've laid some eggs against some very bad teams too. This model is tuned to predict future wins, not count previous ones.
Brian, if that were the case, HOU would not still own the top spot. Any of your top 10 teams would destroy them.
Of course. They're on their 3rd string QB. That's a completely separate issue from how the model incorporates past strength of opponent.
Anon, Houston wouldn't own the top spot because any top 10 team would destroy them? That's an odd thing to say considering Houston is 1-1 against the #3 Steelers and #5 Saints with 50 points scored and 50 points allowed in those games. Or does the definition of "destroy" now include "evenly matched"?
Other Anon, you are overweighting Strength of Victory while disregarding Weakness of Loss (WOL) entirely. The Ravens have lost three games to the #21 Titans, #22 Jaguars, and #28 Seahawks. Those teams are a combined 10-20 against the rest of the league. They are ranked properly according to how they've played this season.
Also, SOV is only a key tiebreaker IF the teams aren't in the same division AND the teams didn't play head-to-head AND they didn't have four common opponents, and even then it's second behind conference record.
To be honest, James, even with Schaub I think the other top ten teams would beat HOU. That's just my opinion though.
CIN's defense is ranked 8th inn both DPASS and DRUNSUCCESS%, yet is only ranked 20th overall. What explains that in the model? Opponent strength?
I have to agree with many others here Brian. Regardless if you think SOV, WOL, remaining schedule, etc. are properly accounted for, and by virtue the accuracy of this model, one glance at it and it's immediately discreded for the simple reason it makes no sense.
Dallas, Philly, NY Giants, all ranked ahead of Baltimore, then further down Chicago and Oakland both ahead of SF? Pardon? Look at the Ravens remaining schedule, it's a joke, but clearly the teams' remaining schedules are factored into your model no?
Don't get me wrong, I love your website and the concept of using "advanced" statistical methods to debunk common misconceptions and beliefs based on insuffecient rational/statistical grounds. I just think (key word lol) that these rankings need to be tweaked to reflect reality a little better. Whether that involes adjusting your weightings etc I dunno, but as others have pointed out, you can't honestly think Philly is a better team (or going to win more remaining games) than Baltimore or SF can you?
@James in the AFC if the Ravens and Pats win out (not that unlikely if you look at their schedules) SOV will be the deciding tiebreaker and the "10th ranked" Ravens will be the #1 seed.
Brian, you've just been "discreded" how does that make you feel?
"one glance at it and it's immediately discreded [sic] for the simple reason it makes no sense"
Future opponents are not factored into the model. For that, see the weekly playoff projections.
Yes, assuming PHI was able to continue at their to-date levels of efficiency, they would win more games. PHI has had some terrible luck and some very close losses. They're 4-7, but have a positive score differential against a tough slate of opponents.
Just look at the table of statistics above. That's all you need to do to understand why teams are ranked where they are. Look at the numbers. LATFN.
Why bother looking at anything other than win-loss records if you think that way?
Thanks for the advice. I'll get right on the task of ignoring the regression results and tweaking my numbers to fit your intuitive guesses of which teams are better. I should have something for you by...now. I'm done. Here's the list:
GB, SF, BAL, PIT, NE, NO, ATL, HOU, OAK...
If you'd like to see a team ranked higher in the list, go ask them to improve their pass efficiency. Don't ask me to monkey with the model.
My question regarding CIN was just that, a question. It seems a statistical curiosity is all. Obviously I am not questioning a properly conducted multivariate regression model. It is what it is. Questioning a mechanical process like that is like questioning the validity of 2 + 2 equaling four. We may quibble over the comprehensiveness of the independent variables in the model, but I'm sure you've included what you can identify at this time to make the best fit. CIN defense simply seems to be an interesting statistical case. thoughts?
I disagree with all the above posters, but as a huge Packer homer, I would like to point out that Packers are underrated by your model, but not through the fault of your model. I mean to say, I honestly believe the Packers let their foot off the gas when they are up, and have in general been able to keep games out of reach for the opponent. They have yet to even be tied with another team in the fourth quarter, and they've won almost every single game by double digits. This, in addition to my eyes, tells me that the Packers really are in control of these games to the extent that if they went all out every game they would have a much better score differential and much better defensive effeciency. Capers calls coverages that allow for consistent gains of yardage at the cost of eating up the clock once the Packers are in a big lead. A month or so back you even had a study that shows GB has been playing much better D in high leverage situations. Maybe I just want to believe the Packers are as good as I think they are given their talent on D, but I really think there is a basis for what I'm saying given the unit's past performance (including the 2010 season) and the unusual situations they often find themselves in.
Anonymous SOV guy here. That was my first post and this will be my last. I don't mean to be overly critical of your work so don't take it personally. I like your site and keep coming here because most "Power Rankings" are just so much bloviating and you try to tie performance to objectively measurable criteria. That's good. But I would urge you to continue to strive to improve your system, rather than discounting criticism and any system that ranks the Eagles as one of the best teams in the league and measurably better teams far behind them (Bal, SF) is seriously flawed. You might try to figure out where the flaw lies rather than discounting it. I work in finance and I see "quantitative arrogance" impede clear eyed thinking all the time. If reality doesn't conform to the model, there must be something wrong with reality. Um, no. That leads to projecting infinite house price appreciation, safe, high-yielding investments and other nonsense. If you were a pilot and your instruments told you you were flying at 30,000 feet and you looked out the front window into the side of a mountain, would you make a correction? It's your system. Decide, but don't demean.
Now, Brian, SOV (as applied by NFL tie-breaking procedures) has nothing to do with points scored, so BAL's SOV it is not based on early blowouts. BAL has won 4 of its last 5 against the league's toughest opponents. That's how they're playing now. If you're into fantasy football, PHI is a better team. If you're into who's a winner, you'd be better off with BAL. You might want to factor in some form of SOV if you want to calibrate your model closer to actual reality. Your model is powerful, and based on a lot of effort, but not necessarily as accurate as it could be. Here's another purely statistical model that does a better job conforming to actual results: http://www.footballoutsiders.com/dvoa-ratings/2011/week-12-dvoa-ratings What have they figured out that you have overlooked?
And James, you are correct. SOV is only the fifth tie breaker behing head-to-head, intraconference, common opponents, etc. But I'm recommending you look at it for an "off label" use (the relative ranking of teams in different conferences, particularly to break ties in ranking identical records). And you are also correct that it totally disregards "weakness of loss" (WOL). So what? If you want to discern relative strength isn't a winning team that beats other winning teams consistently preferable to one that loses to better teams consistently? What does losing to better teams say about a team's playoff prospects? (Not that I want to go down the WOL path, since WOL is not a league statistic, but 2 of NO's losses were against teams with a 6-16 combined record, and they are not penalized for WOL in this model either.) If you play better quality opponents AND do well, you should be recognized. This model igonores or discounts that inescapable fact.
Thank's for the web space to express my views. I will recede into the background as a lurker now.
Brian, I've brought this idea up before, and I'll try to make the question brief: Is it possible that certain factors in the 2nd table are overrated (conversely implying that others are underrated)? I know that weighting those variables differently will produce different results--and that your regression analysis have lead you to use these weights. A follow-up question would be--what historical data set did you use to determine the weights? [For a really bad example, one could use pbp data (if obtainable) from 1970-1977, and say that ypcarry & run success rate are the two most important factors.]
For a real example--over the course of a season, a .5 ypa difference between 2 teams means ~300 yds passing total, so <20 yds/game. Now, maybe that's a whole lot, but it could be a combination of more spikes (if they aren't removed from the stats), or one team's QB throws the ball away 3x/game instead of taking sacks, or one team gets those ~20 yds/game via running. It might also have to do with the team with the lower YPA starting with better field position--so their QB's "long TD passes" are 60 yds instead of 70, because he constantly had a shorter field to work with. Many factors could make those ypa differences basically null. Now, over 1 ypa, including sacks, plus bonuses for TD's and maybe first downs, with penalties for INT's--then adding weights--okay, now we've got something that will help us order teams well. Of course, that's just one component of the formula.
"Here's another purely statistical model that does a better job conforming to actual results: http://www.footballoutsiders.com/dvoa-ratings/2011/week-12-dvoa-ratings What have they figured out that you have overlooked?"
Right...
That's the thing you don't understand. FO conforms to past results. But we already know past results. We can watch them on tv or look them up in the paper. Until you understand the distinction, you're better off sticking with FO, and your job in finance.
Regarding the question about CIN, it's strength of opponent (oppOGWP).
DVOA is bunk. Jets were #1 a couple weeks ago. They're still top 5 or something. Brian had the Texans as #1 weeks before DVOA caught up.
These rankings aren't perfect, and Brian never claimed they were. But they're clearly the least imperfect rankings around.
I'll throw in another one. This site had the Cowboys ranked at or near the top weeks ago when they were 2-3. Now they've ripped off 5 out 6 and are on top of the nfc east.
brian,
Would you say that wins and losses are a function of high leverage situations? I hypothesize that certain teams are built for those situations and some are not. On a slightly different point some teams are built in such a manner that those situation are avoided.
With your rankings I see them as a list of teams I would least likely want to be in a tie game with late in the 4th qtr.
I think the main complaint from Packer backers is that the Packers are head and shoulders above everyone else for team I'm least likely to be in a tie with in the 4th qtr. At least that would be my guess.
This is a problem with regression. It sometimes answers too narrow of a question. Then people want to blame the output which is correct. What we should do instead is seek to broaden the scope of the question.
Ultimately, the Packers will be in one of those 4th qtr situations. This is proven by the 5th law of relativity which died with Albert.
It's amazing how people nitpick one or two unintuitive aspects of the power rankings and proceed to discredit the model as a whole.
I could just as easily point to the fact that this model correctly predicted the strength of Houston (and how underrated they were by FO and others) and the weakness of Tampa Bay (and how overrated they were by FO and others) 6 weeks ago. The former has won 5 straight games and the latter has lost 5 straight games.
As a longtime follower and lurker, I share Brian's sentiment (mostly frustration) when a new slew of posters comment on the Team Rankings. Do everyone a favor and read everything else on this website first. Even if you disagree with the philosophy and objective method behind the model, at least you will be informed and not sound completely ignorant. It's perfectly fine to disagree with the reasoning behind the model, but you can't effectively or seriously doing that be having a gut feeling that one team should be ranked higher than another.
Brian - maybe you can create a "grit and determination" variable so we can finally all breathe a sigh of relief as denver rightfully ascends in the rankings and philadelphia drops.
What kind of joke job in finance do you have where you can take time off in the middle of the day to type a page of nonsense? Probably some kind of a financial adviser or other joke role.
that screwed-up sentence should read: "It's perfectly fine to disagree with the reasoning behind the model, but you can't effectively or seriously do that by having a gut feeling that one team should be ranked higher than another.
wow ... what an incredible jackass anon finance is.
Brian has finally made it as a blogger--he has his own gaggle of trolls.
anon finance,
if you have a genuine intellectual curiosity, go back and read the comments from past power rankings and the posts about how the model was built.
every question/comment you made has already been asked, answered, and explained.
FWIW. I compared the weekly probabilities to game results, and came up with this:
> 80% | There were a total of 11 games in which the stronger team had a win probabilitiy of > 80%. Of those games, the stronger team won a total of 9 games (or 82% of the time).
70% - 79% | There were a total of 27 games in which the stronger team had a win probabilitiy of 70% - 79%. Of those games, the stronger team won a total of 21 games (or 78% of the time).
60% - 69% | There were a total of 40 games in which the stronger team had a win probabilitiy of 60% - 69%. Of those games, the stronger team won a total of 21 games (or 53% of the time).
50% - 59% | There were a total of 50 games in which the stronger team had a win probabilitiy of 50% - 59%. Of those games, the stronger team won a total of 28 games (or 56% of the time).
===================
For some reason, the 60% - 69% results seem to be off. I don't think I made a mistake, but I suppose I could have.
Brian, would it be possible to create something like this for the college game?
Anonymous (the one who posted results for each percentage bracket):
It's called statistical artifice.
I just wished the rankings would include special teams in some fashion. A significant part of the game is overlooked entirely.
Anon above me, the problem is one that has been well and truly tackled. There are few teams (maybe five percent of the league in any given year) that have a truly superior special teams unit, all other teams are essentially indistinguishable in terms of their special teams performance compared to random noise, so they are, for all intents and purposes, the same. Now, since the aim of these rankings is to represent teams in a predictive manner, it makes no sense to allow 95% of teams to be randomly adjusted by non-repeatable special teams performance, so that the other 5% can be a bit more accurate. In the end this would just make the entire set of rankings far less meaningful.
To posit a similar idea, should we also be biased towards those teams that win more coin-tosses than they lose over a season, because they are somehow better at calling a coin-toss?
In response to Packer homer, the Packers are #2 in the power ranking and not #1 because their defense sucks. Aaron Rodgers has been playing flawlessly for 11 straight games. Maybe he will keep it up. Or maybe he’ll have one off week or have to play in a foot of snow or get injured. Green Bay better hope it’s not in the playoffs, because their defense sure won’t be able to bail them out. The Packers’ may be more vulnerable than other top-tier teams because they are so dependent on one player having an all time great season.
"SOV Anon"
Besides your weak post (i doubt you ever read articles of this site or you just don´t understand it), the most ridiculous theme is your idea to overvalue SOV b/c you obviously like BAL. If going (more or solely) w/SOV, the 1-15 Cowboys of 1989 may would be around 15th in your Power Ranking, b/c after all they beat the 10-6 Redskins. Just this little example shows how silly it is to overestimate SOV.
I guess w/Brian´s model those Cowboys rightfully would have been dead last.
Karl, Germany
I fall in the camp of feeling that the rankings are a bit off. I recognize that one of the values of statistical analysis is to remove one's perceptual biases, but it's also true that statistical analysis has limitation, particularly in a sport with as many variables as football.
Brian, you've addressed this when you talk about luck, but I would like to see you post some end-of-season/play-off predictions to test your model. We're far enough into the season that there is enough data for projections.
Obviously for teams with major personal losses, the projections will suffer, but does the model hold up as a projection tool for the non-flukey teams? At what point is there enough data to test the model as longer-term projection tool?
Hey Anon, why don't you read this post on prediction accuracy: http://www.advancednflstats.com/2011/10/prediction-accuracy.html
If that's not enough for you, why don't you look at http://files.nfl-forecast.com/ which uses Brian's data to predict the playoffs. It has predictions starting in Week 3.
Division Winners predicted in Week 3: New England, Baltimore, Houston, Oakland, Dallas, Green Bay, New Orleans, San Francisco.
Current Division Leaders: New England, Baltimore, Houston, Oakland, Dallas, Green Bay, New Orleans, San Francisco.
Pretty damn good, no?
Technically Buffalo was favored in the division @ that point, but New England had a better chance of making of the playoffs. Still I agree...pretty dang good.
Also, I looked it up, and as much as I like Brian's model, Every single team that was being predicted as a division winner, currently was leading the division or tied for the lead. And there are a couple of teams that in retrospect are quite a bit off. (Atlanta, Cincy, Washington) Anyway, love Brian's work, but upon closer examination, not really that impressive the choice of division winners picked after week 3. I could have made almost the exact same predictions had I only looked at records and points differential as the tiebreakers. The most impressive call to me was Green Bay over Detroit, despite both being 3-0 but Detroit having an amazing +55 point differential and "just" +35 for Green Bay.
http://www.shrpsports.com/nfl/stand.php?link=Y&season=2011&divcnf=div&week=Week%203
Thanks, James. That is pretty damn good , even if my gut feeling that the Eagles are no where near the 7th best team is being validated as I type. I suppose that shows you where gut and model each have validity. The model shows you where your gut/pet theory needs to be challenged, and your gut shows where the model fails to factor something (like whatever is wrong with the Eagles, who rarely look better than their record suggests, IMO).