Greetings, good people of Advanced NFL Stats. This week we'll take our first look at the state of the NFL playoff races, in a post complete with numbered lists, bullet points, and made-up numbers. Note: The data below was all calculated using the NFL Forecast app developed by Chris Cox over at NFL-forecast.com, which uses the win probabilities generated by the team efficiency model to simulate the NFL season 5,000 times. And if you don't buy the game probabilities from Advanced NFL Stats, you can tweak them as much as you like to generate your own playoff projection. I encourage everyone to download the app and test out your own scenarios. Since this is the first week, it'd probably be good to review some things to keep in mind when interpreting the numbers:
- Simulating a complete season this far in advance necessarily puts a great deal of weight on our estimates of team strength. For instance, even if a team is undefeated, if we have them ranked last, they're not going to be expected to win too many more games and hence won't be making the playoffs that often.
- The model assumes that team strength will remain constant, while in reality we know that team strength can, and often does, shift markedly during the course of the season, whether due to injuries, changes in strategy, or a team just up and deciding to play better.
- The model assumes that our estimate of team strength will remain constant for the remainder of season. Again, we know this will likely not be the case. Even if a team's actual strength remains the same, as they play more games and produce more data for the team efficiency model to crunch, their team efficiency stats will change and so will our estimate of their team strength.
Right up there with the Saints are the two remaining undefeated teams, the Packers and Lions, along with the 49ers, who are blessed with a division whose highest-ranked team is the #26 Arizona Cardinals. Rounding out the mix are the three non-Dream teams in the East: the Cowboys, Redskins, and, lastly, the Giants, who dropped 18% with their stunning loss to Seattle but still make the postseason 31% of the time. In the AFC, Pittsburgh (+27%) and San Diego (+23%) were the big winners last week, with both increasing their playoff chances enough to move into the Top 6 in the conference. In total, the AFC remains considerably more wide open than its sister conference: there are eight teams in the AFC with a total playoff probability of greater than 50%, but not one higher than 80%.
You can think of the football season as the gradual movement from every team having an equal chance of making the playoffs to a situation where 12 teams are at 100% and the other 20 are at zero. Thinking of it this way, you can quantify just where a playoff race rests on the spectrum of "wide open" to "locked up"--a "lockedupedness ratio" if you like, ranging from zero to one. By this (arbitrary and unscientific) measure, there is indeed a wide disparity between the conferences, with the LR for the NFC at .66 while that for the AFC is only .40.
High Impact Game of the Week: Houston at Baltimore
This meeting between AFC playoff contenders has large playoff implications for both teams, with each seeing about a 20% swing in their respective playoff probabilities depending on the outcome. With the picture in the AFC as muddied as it is, keeping pace with the rest of the field will be key.
Fun Facts
- Losing games does not increase your chances of making the playoffs. But sometimes winning games doesn't help much either. I submit to you the curious case of the Oakland Raiders, whose playoff chances declined 6% despite winning on the road against the Texans. Compare this to the Cowboys, who managed to increase their playoff chances by 10% without even taking the field. Well done, boys. Well done, indeed.
- Despite starting the season in a bit of a hole at 1-4 in a difficult division, the model gives the Eagles a 9% chance of getting into the playoffs anyway. If (somehow) they manage to go 8-3 from here on out, they make the playoffs in about half of the simulations.
- Rams fans, take heart. Though your team may have yet to record a single win, they nevertheless make the playoffs slightly more often than the Falcons, who are 2-3. Hold onto this fact as you face the Packers, Cowboys, and Saints over the next three weeks.
AFC EAST | ||||
Team | 1st | 2nd | 3rd | 4th |
NE | 54 | 35 | 10 | 0 |
BUF | 37 | 41 | 19 | 2 |
NYJ | 8 | 22 | 58 | 12 |
MIA | 0 | 2 | 12 | 86 |
AFC NORTH | ||||
Team | 1st | 2nd | 3rd | 4th |
PIT | 42 | 36 | 18 | 5 |
BAL | 43 | 35 | 17 | 5 |
CIN | 13 | 23 | 46 | 18 |
CLE | 2 | 6 | 20 | 73 |
AFC SOUTH | ||||
Team | 1st | 2nd | 3rd | 4th |
HOU | 64 | 30 | 5 | 0 |
TEN | 33 | 56 | 10 | 1 |
JAC | 2 | 12 | 66 | 19 |
IND | 0 | 2 | 19 | 79 |
AFC WEST | ||||
Team | 1st | 2nd | 3rd | 4th |
SD | 61 | 31 | 7 | 1 |
OAK | 34 | 47 | 15 | 5 |
KC | 3 | 13 | 41 | 43 |
DEN | 2 | 10 | 37 | 51 |
NFC EAST | ||||
Team | 1st | 2nd | 3rd | 4th |
DAL | 61 | 26 | 10 | 3 |
WAS | 26 | 42 | 24 | 8 |
NYG | 10 | 24 | 46 | 20 |
PHI | 2 | 9 | 21 | 68 |
NFC NORTH | ||||
Team | 1st | 2nd | 3rd | 4th |
GB | 57 | 41 | 2 | 0 |
DET | 43 | 54 | 3 | 0 |
CHI | 0 | 4 | 63 | 33 |
MIN | 0 | 2 | 32 | 66 |
NFC SOUTH | ||||
Team | 1st | 2nd | 3rd | 4th |
NO | 97 | 3 | 0 | 0 |
CAR | 2 | 47 | 32 | 20 |
TB | 1 | 35 | 41 | 22 |
ATL | 0 | 15 | 27 | 58 |
NFC WEST | ||||
Team | 1st | 2nd | 3rd | 4th |
SF | 88 | 9 | 2 | 0 |
SEA | 6 | 44 | 32 | 18 |
ARI | 4 | 29 | 38 | 30 |
STL | 2 | 18 | 28 | 52 |
AFC Percent Probability Playoff Seeding | |||||||
Team | 1st | 2nd | 3rd | 4th | 5th | 6th | Total |
NE | 19 | 16 | 11 | 8 | 13 | 11 | 78 |
HOU | 16 | 16 | 17 | 16 | 6 | 6 | 78 |
PIT | 15 | 12 | 9 | 7 | 19 | 11 | 73 |
BAL | 16 | 12 | 10 | 6 | 15 | 11 | 69 |
SD | 10 | 12 | 16 | 21 | 4 | 6 | 69 |
BUF | 10 | 10 | 9 | 8 | 12 | 12 | 60 |
TEN | 6 | 9 | 9 | 9 | 11 | 12 | 55 |
OAK | 4 | 7 | 11 | 14 | 5 | 9 | 50 |
CIN | 3 | 3 | 3 | 3 | 8 | 10 | 30 |
NYJ | 1 | 2 | 3 | 3 | 4 | 7 | 20 |
CLE | 0 | 0 | 1 | 1 | 1 | 2 | 5 |
KC | 0 | 0 | 1 | 2 | 1 | 1 | 5 |
JAC | 0 | 0 | 1 | 1 | 1 | 1 | 4 |
DEN | 0 | 0 | 0 | 1 | 0 | 1 | 3 |
MIA | 0 | 0 | 0 | 0 | 0 | 0 | 1 |
IND | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
NFC Percent Probability Playoff Seeding | |||||||
Team | 1st | 2nd | 3rd | 4th | 5th | 6th | Total |
NO | 25 | 33 | 26 | 12 | 0 | 1 | 98 |
GB | 36 | 13 | 5 | 2 | 32 | 7 | 96 |
DET | 25 | 12 | 5 | 1 | 43 | 9 | 95 |
SF | 4 | 10 | 21 | 53 | 0 | 1 | 89 |
DAL | 6 | 20 | 25 | 9 | 5 | 16 | 82 |
WAS | 3 | 8 | 10 | 5 | 9 | 25 | 60 |
NYG | 0 | 2 | 4 | 3 | 5 | 16 | 31 |
PHI | 0 | 0 | 1 | 2 | 1 | 5 | 9 |
SEA | 0 | 0 | 1 | 5 | 0 | 2 | 8 |
CHI | 0 | 0 | 0 | 0 | 2 | 6 | 8 |
CAR | 0 | 0 | 0 | 1 | 1 | 4 | 7 |
TB | 0 | 0 | 0 | 1 | 1 | 3 | 5 |
ARI | 0 | 0 | 0 | 4 | 0 | 1 | 5 |
MIN | 0 | 0 | 0 | 0 | 0 | 3 | 3 |
STL | 0 | 0 | 0 | 2 | 0 | 0 | 2 |
ATL | 0 | 0 | 0 | 0 | 0 | 1 | 2 |
The division standings tables look beautiful, I think it would be great if you added the team's current records.
Do we really care about the teams' records? Ultimately we just want to know if they are going to make playoffs and in what order. Since the playoff probabilities factor record, tie breakers, and everything else in, team record doesn't give us any more information.
"OK, enough vegetables, on to the numbers..."
Who you calling a vegetable?
5000 simulations isn't even approaching enough. Try 500,000...
If 5000 simulations aren't enough for you, why don't you run the software 100 times and then average the results?
And whatever percentage point or two in increased accuracy 500,000 simulations would provide is completely overwhelmed by the inaccuracies of Brian's GWP, so I think we're good with 5,000 for now.
I don't like to see Monte Carlo simulations done improperly, and to get even close to the true % implied by Brian's ratings will take in the region of 50,000 at a minimum, but probably more. There is no reason to compound any error when it is avoidable.
There's not really much reason to do a monte carlo simulation, when at week 6, there's only about 2^27 possible win/loss permutations. Exact probabilities could probably be computed in a few seconds.
I think a big flaw in this model is that it assumes that GWP is constant throughout the season. GWP hasn't stabilized yet and by simple regression to the mean the top ranked teams are more likely to reduce GWP and bottom ranked teams are more likely to increase GWP.
Sam P, let's just consider the regular season games:
There are 179 remaining, in each game one of two teams can win, as such there are 2^179, plus playoff games. Where did you get 2^27?
Anyway, such is the case that a Monte Carlo simulation is the only sensible way.
True about the assumption GWP is constant. That's stated up front. But GWP is already heavily regressed to account for that.
What it doesn't account for is catastrophic 'black swan' events, such as a season-ending injury to critical player, say Drew Brees.
I agree that the Saints are the best in the NFC South, but how do the Bucs only have a 1% chance of winning the division? If they win today, they'll be tied for first with the tiebreaker.
I'd say it's because they have a 0.35 GWP to the Saints 0.65, and as such the Bucs have less than a 10% chance of being better than 8-8, according to this projection.
Gah, I added exponents instead of multiplied! (i.e. (2^16)^11 = 2^176, not 2^27)