Postseason Projections: Wild Card Round

And then there were twelve. After 17 weeks and 256 games, the field has been winnowed and the playoff bracket is set. For the postseason, we can skip the simulations and calculate each team's chances directly from the win probability model.

Most Likely Super Bowl: Patriots vs. Packers
This shouldn't be much of a surprise. The trifecta of a highly-ranked team, a bye week, and home-field advantage is not easy to overcome, and the model projects the two #1 seeds as the most likely teams to emerge from their respective conferences. That said, there is an 87% probability that at least one of these two will not make it to Super Bowl XLVI.

Longest of Long Shots: Denver Broncos
It seems even three consecutive losses is not enough to put a damper on Tebow-mania. Denver's odds of winning the Super Bowl are variously listed as anywhere from 50/1 to 120/1, but, regardless, the Broncos are still over-valued, with the model estimating their true probability of a Super Bowl win to be less than one in eight hundred.

Battle of the Conferences
The conferences are fairly evenly matched, with the AFC remaining the marginal (53%) favorite to win the Super Bowl. Adjusting Houston's GWP downward to account for their quarterback issues brings this number down—but only slightly, to just below 50%. (Keep in mind, the lower our estimate of Houston's team strength, the less likely it is they will make it through to Indianapolis.)

And now, the full postseason projections, with the tables below listing each team's percent probability of advancing to each successive round of the postseason. Enjoy.


AFC Percent Probability to Advance
TeamDivision RoundConference GameSuper BowlSup Bowl Champion
NE100583216
BAL10044188
HOU78452414
DEN19310
PIT81442414
CIN22621


NFC Percent Probability to Advance
TeamDivision RoundConference GameSuper BowlSup Bowl Champion
GB100634020
SF10039134
NO68432312
NYG6530147
ATL351462
DET321152


  • Spread The Love
  • Digg This Post
  • Tweet This Post
  • Stumble This Post
  • Submit This Post To Delicious
  • Submit This Post To Reddit
  • Submit This Post To Mixx

26 Responses to “Postseason Projections: Wild Card Round”

  1. Anonymous says:

    Do you account for the fact that the bracket is not fully fixed -- i.e., wildcard round teams are re-seeded next week?

  2. Josh Katz says:

    Yes, re-seeding is taken into account--you kind of have to for the numbers to make any sense.

  3. Anonymous says:

    What's the Texans injury adjusted probabilities?

  4. Keith Goldner says:

    So the Eagles can't win the Super Bowl?

  5. Anonymous says:

    They are not re-seeded. Highest seed always plays the lowest seed.

  6. Dale says:

    Does it take into account Tim Tebow's divine intervention intangible?

  7. Brian Burke says:

    Yes. Divine intervention is always accounted for.

    Anon-I believe that's what 're-seeded' means. In other words, The #1 seed won't always play the winner of the 4/5 game and the Ravens the winner of the 3/6 game. The league waits and see's who the lowest seed remaining is, and gives them to the #1 team.

  8. Anonymous says:

    If you believe these numbers, there are lots of good bets to be made in Vegas and the offshore world. Good luck!

  9. Josh Katz says:

    @Anonymous
    It depends what method you use to adjust for the injury. Worst case has Houston's probabilities at 61/22/7/3.

    But I'm not always the biggest fan of this kind of ad hoc adjustment, just because it can become somewhat arbitrary... eg, if we're making adjustments for the Texans, why not also do so for the Steelers, who lost Rashard Mendenhall? And what about Roethlisberger's gimpy ankle--shouldn't we also account for that? And Andy Dalton has the flu, the Saints will be without Lance Moore, etc. etc.

    It's worth keeping in mind that, at this point, the Texans already have six Schaub-less games baked into their stats. Their GWP right now may not be completely inaccurate. (And, though much of it came against subpar teams, TJ hasn't been that bad. His EPA/P is good enough for 12th in the league, on par with Roethlisberger and Mike Vick.)

    @Keith
    The Eagles can always win the Super Bowl. Always. But most likely they will take the lead and then blow it in the fourth quarter.

  10. Anonymous says:

    Do they have an "Intrade" equivalent for sports. Since the Broncos are overvalued, it would be cool if you could somehow "short" them since they seem overvalued by Vegas for example.

  11. Anonymous says:

    anon- Intrade used to be called tradesports.com and the political arm was spun out a ways back. Tradesports then folded. Don't know if there is any suitable replacement.

  12. Capn says:

    Brian,
    I asked you this a ways back and you implied there was not too much to it, but now that the Ravens are 8-0 at home and won their home games in a rather convincing way, don't you think there needs to be some adjustment for their home-road splits. I realize most teams have only a one game difference, but the Ravens were 8-0 at home this year beating 4 playoff teams in Pittsburgh, Houston, San Francisco, and Cincinnati by a combined 111-51. Meanwhile they were 4-4 on the road losing all 4 games to non-playoff teams (Jax, Seattle, Tenn, SD) by a combined 51-94.

    I don't know how you deal with this kind of split, but it seems like you would need to. It also appears that your model does not since it favors both Hou and Pitt over Bal to make it to the conf finals despite having to play the Ravens at home and despite the Ravens getting a bye. Under normal circumstances, this would seem like a stretch, but with the injury issues facing both Hou and Pitt, it seems almost comical. If I were a betting man, I would not bet against the Ravens at home against anyone. Comment?

  13. Anonymous says:

    Anon above - it makes no sense to take into account such dramatic home-away splits because there is no question they are due to randomness from a small sample size. Make the Ravens play 100 more games and see if it still stands. Remember the Falcons last year? Going into the NO game at the end of the season, all the talking heads were touting their (I believe undefeated) home record, saying there was no way they would lose at home..well regression to the mean bit them in the ass as the Saints came in and easily beat them, then ATL lost again at home in the playoffs.

  14. Capn says:

    @Anon

    I don’t really think it is small sample size at all. Since the 2008 season, the Ravens are 27-5 in home games. Their 5 losses were 13-9 to the Steelers in 2008 (who went on to win the Superbowl), 13-10 to the Titans in 2008 (went 13-3 and lost to the Ravens in the playoffs), 17-15 to the Colts in 2009 (the year they went 14-0 and lost in the Superbowl), 17-14 to the Bengals in 2009 (who went 10-6 and made the playoffs), and 13-10 to the Steelers in 2010 (when they were 12-4 and lost in the Superbowl). So in 4 seasons, the Ravens have lost 5 home games. All 5 teams that beat them went to the playoffs and 3 of the 5 went to the Superbowl. They lost by a total of 15 points in those 5 games. Overall, they have outscored their opponents 782-419 during that same 27-5 stretch. During that same stretch on the road, they have gone 21-18 including the playoffs.

  15. Anonymous says:

    Let us indulge in some math.

    The question is: given a generic home-field advantage, in logit space, of 0.13 (which I believe is the correction Brian uses), does the value of GWP for the Ravens in that time (averaging 0.613 for the three years I could find) provide a prediction statistically consistent with observed wins and losses?

    If the answer is no, then that means that there is evidence for "specialness" in the Raven's splits. If the answer is yes, it suggests the splits are perfectly consistent with how GWP and HFA is already handled.

    The numbers work out to be predictions of 24-8 and 19-21, home and away. If you treat the binomial errors as normal errors and treat it as a chi-squared with 2 dof (mathematically squishy but practically convenient and acceptable for the calculation at hand), you will find that splits like this occur 27% or more of the time.

    This is essentially a p-value of 0.27 , which in any statistical convention will not allow rejection of the null (namely, that GWP and HFA are adequately accounted for.)

    So the Raven's W-L records provide no compelling evidence of home-away specialness.

  16. capn says:

    Hey thanks. That is very interesting and compelling.

  17. Joseph says:

    IMO, HST definitely needs to be adjusted downward. They have the 2nd most chance in the AFC to get to and win the SB. I wouldn't give them much more than 25% to beat BAL, NE, or PIT--although maybe a 30% chance to beat PIT because they'd be at home for the AFCCG. Since they'd have to beat two of them after beating CIN, I think their probability should be something like 65/25/10/4--about what Josh Katz said. Anyway to show us a 2nd probability table for those odds?
    Also, I'm betting that most places are giving better than 50% odds for the NFC to win the SB--as GB & NO should be favored over whichever AFC team makes it to Indy.

    My playoff picks (if anyone cares)--NO over DET, NYG squeak it out over ATL, NO beats SF, GB whallops NYG, GB edges NO to break my heart. HST edges CIN, PIT takes care of DEN, BAL beats HST, NE beats beaten-up PIT, but BAL upsets NE only to lose the SB to GB.

  18. Anonymous says:

    Thanks Josh....I realize it's a slippery slope when you start adjusting one team and not another for key players missing/hurt. However, I think we can all agree that the QB is the most valuable player for each team when it comes to potential outcomes and it's apparent that Yates, while adequate, is a step down from Schaub's first 10 games. I think 61/22/7/3 is a good ad-hoc adjustment. Let's be honest, if we blindly followed the numbers, we all would be flying to Vegas to put down money on the Texans at 40-1 to win the Super Bowl when the numbers got them pegged at 7-1. I put a ton of faith in this site and these numbers, but to quote Meatloaf...I would do anything, but I won't do that!

  19. Anonymous says:

    I think you guys are underestimating Yates..

    In this post: http://fifthdown.blogs.nytimes.com/2012/01/05/n-f-l-probabilities-for-wild-card-round/?ref=football
    Brian says the win prob w/ Yates for the first game is still 72

  20. TMK says:

    Again, if you adjust for Yates, do you adjust Dalton for being sick, and if so how does that effect the adjusted probabilities? I think sticking with the original numbers is smart. If you start adjusting numbers down, then do you bring them back up if Yates does well? or is it a fluke? If HST beats BAL and NE then why woud they not be 14% to winn the SB?

  21. Anonymous says:

    TMK....Dalton being sick is quite different than Schaub being out, but I understand what you're getting at.

    This brings me to a larger point.....the numbers unadjusted for Houston don't come close to reality. The unadjusted numbers say Houston is 7-1 to win the Super Bowl. Vegas says it's 40-1.

    What does sticking with the original numbers get you when it's off by a factor of 6 from Vegas' numbers? Do we all like to think we're in on some super secret information unavailable to Vegas that the Texans are the 2nd most likely AFC team to win the SB even though Houston lost their last 3 games to Carolina, Indy, and Tennessee?

    Does anyone truly believe this? Anyone?

    After the playoffs, we have a sample size of exactly one iteration of games....in other words....ANY result will be one you can rationalize away because of limited sample size. The numbers unadjusted could say Houston has a 99% chance of winning this weekend and if they lose and we live with blinders on....we'd shrug our shoulders and say...well, that was the 1 in 100 chance.

    At the end of the day, this site by publishing expected outcomes is putting itself out there to be compared with Vegas. That's the standard. If it isn't better than Vegas, what's the point? I can get "different" anywhere....and if needed I can calculate game probabilities from Vegas' money lines myself.

    I'm not looking for this site to peg to Vegas...quite the contrary....the numbers need to speak for themselves....but we all should expect them to be in the area code and when they're not, I think it's a good idea to not blindly follow the numbers off the cliff.

  22. David Kravitz says:

    Reading comments I understand what you're saying about the Ravens, however there is absolutely no way to justify giving them less than 50% chance of making the conference game. They will be playing at home against likely Houston. They certainly have a 60% chance of beating Houston, and Houston has a 78% chance of coming to Baltimore, that exceeds 44% right there.

  23. Anonymous says:

    i applied this to nfl.com's playoff challenge to find what players to pick and i found a pretty good ratio through simple math for each player to earn the same amount through the doubling etc. for every ten pts a packer player scores a raven needs 12.3, texan 12.5, steeler/saint 13.3, and a patriot roughly 16 per game to create the same amount of points over the challenge

  24. Anonymous says:

    I love Anon's post three above mine (discussing blindly following ANS's projections and not critically thinking about what's really going on). More broadly, I see a concept that acts as the reciprocal of being results-oriented (to a fault). That is, rather than a results-oriented person illogically saying a predicition or decision is right/wrong based on outcome, I think we need to be careful not to be too far in the other direction and cite sample size or luck everytime we are wrong. Houston is great example of this mindset. Kind of hard to explain, but perhaps this idea has been discussed [in a more scientific manner] elsewhere. Thoughts?

    CR

  25. Anonymous says:

    Will there be another probability for the divisoinal round so we can se how the wins affected the playoff odds

  26. Josh Katz says:

    Done!

Leave a Reply