website statistics
With just one more Sunday left before the postseason, one of the most popular features on this site has become the live playoff probabilities.  If you've read this article regularly this season, you've probably noticed how the efficiency rankings would seemingly differ significantly from what the postseason odds would suggest.  That largely stems from the fact that those odds are based on Vegas spreads, which obviously caters to the mainstream perception.

Nevertheless, it's still interesting to examine the disparity between those two AFA models.  To change things up in this abridged look at our penultimate rankings, let's highlight some of the biggest differences between the current odds and season-long model.  Both are predictive models, of course, so a deeper dive into a few case studies might reveal which one appears likelier to fulfill its purpose.

Efficiency Darlings

These three teams are ranked highly by the efficiency rankings, but their odds don't quite match up.

- Though they officially fell out of contention for the top seed on Monday night, the Denver Broncos have been the prohibitive top-ranked team in the AFA rankings all season.  Even with the loss, their Gross Win Probability (GWP) sits at 0.75, five percent better than second-place Seattle (who finally supplanted Miami).

However, the Playoff Probabilities see Denver as 36 percent favorites to reach the Super Bowl and just 20 percent possibilities to win it all, well behind New England and Seattle.  That's actually a bit more favorable than Elo, another metric that incorporates Vegas spreads into their model, which pegged the Broncos with a 17 percent chance of winning the Super Bowl headed into Week 16. 

Much has been made of Denver's shift into a run-heavy offense.  The numbers do bear this out; since C.J. Anderson took over for an injured Ronnie Hillman/Montee Ball duo in Week 10, the Broncos have run the ball on 44 percent of their offensive plays, compared to just 39.2 percent of plays over the first nine weeks.  But it's not clear that Denver has actually been a better offense; after averaging 6.21 yards per play in that earlier stretch, second-best in the league, the Broncos are down to 5.84 yards per play since Week 10, which ranks sixth in that span.

Obviously that latter mark is far from disastrous, but it also shouldn't be surprising.  There's an incalculable amount of data clearly illustrating how much less efficient running the ball is in comparison to passing, and that balance is likely even more skewed when considering that the Broncos happen to employ one of the three to five greatest quarterbacks of all-time.  It's not surprising that Denver's offensive Expected Points Added (EPA) has most hovered around zero since Week 10, pending the numbers from the Cincy game:

Clearly Peyton Manning's health is at the forefront of Denver's Super Bowl odds.  Twenty percent likely undersells Denver's championship odds with a healthy Manning, especially when considering the AFC's weakness.  Virtually every model has either the Patriots or Broncos as the AFC Super Bowl rep in 50 to 75 percent of simulations, so even while Denver would probably be underdogs at Foxboro, they would only need one home win to reach Gillette Stadium.

Of course, Manning has been dealing with a thigh injury, and some of those fluttering picks from his four-interception night this week have some wondering about potential dead-arm syndrome.  If there is indeed something wrong with Peyton, then it suddenly becomes hard to envision Denver's offense being championship-caliber behind a relatively inefficient running game, a statement that would have sounded shocking one month ago.

- The Miami Dolphins were officially eliminated from the playoff race on Sunday despite their thrilling comeback against the Vikings.  At 8-7, the Fins will make for an interesting case study when I delve deeper into some rationale for these articles in a postseason article.

Regardless, while the mainstream public seems to have formed a consensus "Big 4," the AFA rankings have similarly constructed a Rushmorian hegemony at the top.  However, the Dolphins remain solidly ahead of the AFC East-leading Patriots, with 0.65 GWP compared to 0.61 for the Brady Bunch.

That mark would suggest a true average record of 10.4-5.6 over a 16-game schedule, something that seemingly overestimates the Dolphins by as much as a full win.  With one game remaining, Pro-Football-Reference's Pythagorean win expectancy pegs the Dolphins with 8.2 expected wins, right in line with their 8-7 record. Other models like DVOA are high on Miami, but not to the same extent as AFA has been all season.

A very cursory look suggests that a bizarrely high offensive success rate might be behind Miami's ensconced position at the top.  The overall efficiency soup portrays the Dolphins as the ninth-best offense overall, but based on success rate alone, only the Saints have eclipsed Miami this season:

The EPA doesn't exactly match because there aren't many high variance situations in which Miami has stood out.  The Dolphins are almost exactly middle-of-the-pack in both third-down conversion rate and turnover percentage.  Success rate applies to all downs, but on first and second down, Miami's offense ranks 12th in yards per play.  That's solid, but hardly something that would augur the second-best offense in the league by success rate.

Miami has arguably been AFA's biggest mystery this season, and there isn't enough space to flesh out all the reasons why the Dolphins are seen as not just underrated, but potentially elite.  The Dolphins' record obviously did not reflect that this season, so it will be instructive at some point to examine why.

- On the heels of their fourth win in five games, the Pittsburgh Steelers were this week's biggest riser, jumping up five spots to reach the top 10 for the first time this season.  The model still puts them behind the Ravens, but at 0.56 GWP, the model doesn't see the Steelers as significantly worse than most of the second-tier AFC contenders.

However, the probability model seems to portray the Steelers as the fourth wheel in the AFC, behind Denver, New England and Indy.  This doesn't make them significantly different from these rankings, but at the same time, Pittsburgh is being given just a five percent chance to reach the Super Bowl and a two percent chance to win it all. 

That would seemingly underestimate the chances for the No. 1 offense, as the Steelers finally wrested away the top ranking from Denver in that category.  Pittsburgh ranks second in total offensive EPA, just behind Green Bay, but also well ahead of third-ranked Dallas. 

There's a chance that we've never seen a Steelers time like this at any point in NFL history.  With an absurd 11.05 offensive EPA per game yet a well below-average 5.74 defensive EPA allowed per game, Pittsburgh is more offensive skewed than it has ever been.  Last year's squad came very close, but in the AFA era (since 1999), the Steelers have not had a team in the good offense/bad defense EPA quadrant:

There's evidence that such wildly imbalanced teams can make deep playoff runs (2013 Broncos, 2011 Patriots), but they're rare and usually come from the weaker conference.  With the NFC South champ and quarterback-deprived Cardinals in the NFC field, the playoff field will likely match the season-long evidence that the AFC has reclaimed the upper-hand in the power struggle between conferences.

It will be interesting to see if Pittsburgh's chances shift significantly at all if they beat Cincy in Week 17 and earn a home game during the first round.  The Steelers would also get the No. 3 seed with a win, lining them up for a trip to Denver rather than New England, which looks like a slightly more favorable (albeit still daunting) draw at this point.  Regardless, Pittsburgh is emerging as a sneaky third wheel behind the Pats-Broncos stranglehold, something the odds could reflect next week.

Playoff Darlings

These teams are ranked highly by the playoff odds, but their efficiency rankings don't quite match up.

- Most mainstream models see the New England Patriots as the clear Super Bowl favorite at the moment.  Elo, weighted DVOA and SRS have the Pats first overall, while the oddsmakers at Bovada have made them massive 4/5 favorites to reach Glendale this February.

However, as I've alluded to earlier, the Pats are seen as the third-best AFC team in the efficiency rankings.  Many will point to New England's wretched September as the reason for that, but while the Pats were the third-worst offense by EPA in the first month, they also had an above-average defense propping them up during that stretch.  It's not as if they were overcoming a Jacksonvillian hole.

It seems as though the difference might stem from New England's passing game, which the AFA model sees as much less effective than most people.  The model uses net yards per attempt (NY/A), in which Tom Brady ranks just 15th among qualified QBs.  Brady's downfield passing game has not been his strength ever since Randy Moss' departure, and the plethora of three-step passing concepts the Pats lean on prevent his NY/A from ever matching vertically based quarterbacks like Tony Romo or Andrew Luck.

However, stats that incorporate turnovers, like AY/A and ANY/A, portray Brady as a top 10 quarterback.  EPA does consider turnovers as well, and in part because of that, Brady stands as a top five quarterback in that category this year.  

In the end, New England's overall ranking is a bit depressed by some fluky poor performances in run defense (particularly from the first half of the season).  The formula for beating the Patriots remains the same as it has ever been: fluster Brady with a strong pass rush and trust the corners to cover a suspect wide receiver group, while committing plenty of resources to Rob Gronkowski.  The Vegas-influenced probabilities suggest that this is an unlikely task, but the efficiency rankings seem to suggest a more vulnerable Pats team than many might believe.

- Virtually every model, including AFA, sees the Detroit Lions as a likely wild card entrant.  The Lions need a victory at Lambeau Field, where they have a historic 23-year winless drought.  Otherwise, they will likely head to Arlington to face the Cowboys in the first round.

Nevertheless, the playoff probabilities still see Detroit as the third-likeliest NFC Super Bowl rep, behind Green Bay and Seattle.  In fairness, the latter two teams dominate the odds, with a combined 79 percent chance of reaching Super Bowl.  But expanding the odds to the entire league, the Lions are the only squad outside of the "Big 4" that wins it all in more than two percent of simulations.

The efficiency rankings do like the Lions, moving them back up to ninth after their fourth consecutive win.  Interestingly, this looks like a case where Detroit might be overrated based on both AFA's model and the Vegas probabilities, two measures that are usually on vastly different ends of the spectrum.  DVOA and SRS see the Lions as a middling team, and even the Vegas-incorporating Elo had them 11th last week.

Matthew Stafford's statistical decline has drawn much of the attention, and the questions about Detroit's offensive ceiling are probably legitimate.  The Lions will most likely play every playoff game away from Ford Field.  I'd have to do a more comprehensive study to quantify this, but it would hardly be surprising if Detroit experienced the most disparate offensive location splits of any team this season:

Another significant issue might be the defense's reliance on big plays to defend the pass.  Detroit ranks seventh in pass EPA/P, but just 18th in pass success rate, suggesting that their elite 3.4 percent interception rate probably has a lot to do with the former mark.  Astoundingly, the Lions have had just one game without a takeaway this season, tied with four other teams for the fewest in the league. 

Turnovers are highly variable, so it's probably folly to try and predict how the Lions' takeaway pattern will play out over a tiny postseason sample.  It's interesting to see the probability model favor Detroit, and one wonders if they'll remain the top favorite outside the Power 4 even if they extend their Lambeau losing streak to 24 next week.

- There's usually at least one team from the bottom third of the efficiency rankings that reaches the postseason every year, but the 27th-ranked Atlanta Falcons might set a new low if they hold serve at the Georgia Dome in the NFC South title game.  Obviously a lot of the Falcons' situation is circumstantial, given that the South champ will be just the second division champ with a losing record since the 2002 realignment.

Still, the probabilities see the Falcons as 60 percent favorites to reach the postseason, even though the Panthers have wriggled their way up to 16th in the efficiency rankings.  We shouldn't suggest that the odds like the Falcons too much—for reference, even though the Chargers only have a 38 percent shot at reaching the postseason, they are still given a four percent shot at reaching the conference championship round, compared to five percent for Atlanta.

However, the model strongly disagrees with making the Falcons 60 percent favorites.  I won't spoil the exact odds, since Brian usually reveals them over at Sports on Earth, but the gap between the two teams in the rankings tells you which team the model favors to reach the playoffs. 

The winner of the Atlanta-Carolina matchup may very well face a team starting Logan Thomas or Ryan Lindley at quarterback, which could pit two of the weakest postseason squads ever against each other.  If Arizona drops two more spots in the final rankings, we could have two bottom-10 teams facing each other on Wild Card weekend, something that will likely prompt studies wondering if this is indeed the worst playoff game in league history.

1 DEN 1 0.75 0.50 4 1
2 SEA 3 0.70 0.50 7 3
3 GB 4 0.66 0.48 2 14
4 MIA 2 0.65 0.50 9 6
5 NE 6 0.61 0.51 8 13
6 IND 5 0.60 0.51 5 22
7 DAL 13 0.59 0.48 3 23
8 BAL 7 0.57 0.49 11 9
9 DET 11 0.57 0.47 17 4
10 PIT 15 0.56 0.48 1 29
11 KC 8 0.56 0.51 21 8
12 CIN 14 0.55 0.51 14 12
13 NO 9 0.54 0.50 6 30
14 BUF 10 0.54 0.51 28 2
15 CLE 12 0.54 0.48 18 7
16 CAR 19 0.52 0.50 20 10
17 HOU 18 0.52 0.49 16 11
18 PHI 17 0.52 0.50 13 18
19 SF 16 0.51 0.50 26 5
20 NYG 21 0.47 0.49 15 20
21 ARI 20 0.44 0.51 25 19
22 TEN 24 0.43 0.49 23 16
23 SD 23 0.42 0.51 22 21
24 MIN 25 0.41 0.50 27 17
25 WAS 26 0.41 0.48 12 28
26 STL 22 0.41 0.48 24 15
27 ATL 28 0.39 0.49 10 32
28 CHI 27 0.39 0.52 19 31
29 NYJ 29 0.34 0.53 29 24
30 JAC 30 0.29 0.52 32 26
31 TB 31 0.27 0.50 30 25
32 OAK 32 0.26 0.53 31 27

ATL 7.1 36 2.2 1.3 7.6 58 2.9 0.41
ARI 6.3 33 1.7 1.5 6.8 63 3.3 0.37
BAL 6.6 41 2.3 1.2 6.2 66 1.8 0.42
BUF 5.8 38 2.3 1.7 5.5 62 3.6 0.51
CHI 5.9 46 3.3 1.6 7.2 57 2.5 0.50
CAR 6.0 44 2.3 2.3 6.2 61 2.3 0.37
CIN 6.6 40 3.2 2.0 6.1 54 3.2 0.39
CLE 6.6 37 3.2 1.5 5.7 53 3.8 0.44
DAL 7.4 43 2.3 2.2 6.9 61 3.1 0.42
DEN 7.5 42 2.6 1.2 5.4 65 2.8 0.52
DET 6.3 40 2.1 1.6 5.7 64 3.5 0.49
GB 7.4 42 1.2 1.7 6.1 53 3.4 0.38
HOU 6.6 38 2.7 1.9 6.1 56 3.4 0.39
IND 7.0 46 2.6 2.3 6.6 57 2.3 0.42
JAC 4.9 36 3.4 1.3 6.7 60 1.0 0.28
KC 6.0 41 1.3 2.1 5.5 54 0.8 0.34
MIA 5.9 51 2.2 2.3 5.8 59 2.7 0.31
MIN 5.6 42 3.5 0.9 6.3 56 2.6 0.43
NE 6.6 45 1.6 1.0 6.2 58 2.9 0.51
NO 6.9 45 2.3 1.7 7.1 58 2.1 0.33
NYG 6.6 38 2.3 1.8 6.8 60 3.3 0.41
NYJ 4.9 44 3.2 2.3 6.6 60 1.2 0.47
OAK 5.1 34 2.5 1.9 6.7 58 1.8 0.45
PHI 6.7 42 3.4 2.1 6.5 61 2.0 0.46
PIT 7.4 43 1.4 1.2 7.2 61 1.8 0.44
SD 6.8 33 3.0 1.3 6.3 54 1.4 0.48
SF 5.6 41 2.2 1.9 6.0 62 3.9 0.46
SEA 6.5 49 1.4 2.2 5.5 63 2.3 0.54
STL 6.1 39 2.9 2.5 6.6 66 2.3 0.59
TB 5.7 33 3.7 2.6 6.8 62 2.1 0.49
TEN 6.4 38 3.3 1.9 6.3 57 2.4 0.50
WAS 6.7 42 3.2 2.6 7.0 63 1.2 0.55
Avg 6.4 41 2.5 1.8 6.4 59 2.5 0.44