As usual, my role in all of this is to attempt to add some semblance of order to the chaos. That takes the form of my College Football Playoff prediction model. The basic structure of this model is the same as last year, with a couple minor tweaks and an additional year of data added. I'll rehash the basics of the model in the next paragraph, but please read last year's post if you want more detail.
The model simulates the season using an aggregate team strength rating that combines ESPN's FPI, Jeff Sagarin's Pure Points metric, and Bill Connelly's S&P+. Once that simulation is complete, I sum up the number of times each power conference team finishes with 0, 1, and 2 losses and the number of times each group of five team finishes undefeated (this represents the four "buckets" of teams that have made the top four in the BCS/playoff era). I also calculate an average strength of schedule for each team, using the NCAA's basic definition of SOS. Once I have this information, I determine how likely a team is to make the playoff for each bucket, and then sum those up. For the power conference 0-loss bucket, I assign a 100% chance for each instance, as no eligible team has ever missed the top four. For the other three buckets I assign the probability based on a logistic regression that uses the past 19 years of BCS*/playoff outcomes as data and the SOS calculation as an input. Finally, I adjust these playoff odds so that the sum of all teams' likelihoods equals to 4. Most of the time this means increasing the odds for all teams**, which I do by adjusting the coefficients of the logistic regression using my old buddy Solver. The results of this are my playoff odds.
*Once we have more playoff data, I will either limit or heavily weight the model towards that. For the time being though, I think BCS data is a fairly good proxy of the committee's logic.
**I'm not precisely sure why the sum of initial playoff odds generally comes up short of 4, but I'm guessing it's either 1) that pre-season projections inherently clump teams together or 2) my simulation assuming complete independence of game results when that's not quite the case
The Rankings
Below are the initial playoff odds for 2017. I have only included the teams that finished at least one of my 10,000 simulations with two or fewer losses (or zero losses for group of five teams). This means that the best team left out is either Cal or Iowa State. I think we're safe.
Rank | Team | Agg Rank | POFF Prob |
1 | Alabama | 1 | 70.31% |
2 | Ohio State | 2 | 67.45% |
3 | Florida State | 3 | 35.88% |
4 | Oklahoma | 4 | 29.54% |
5 | Wisconsin | 9 | 21.17% |
6 | Clemson | 6 | 20.90% |
7 | LSU | 7 | 20.87% |
8 | Auburn | 5 | 20.71% |
9 | Penn State | 10 | 19.03% |
10 | Washington | 8 | 18.37% |
11 | USC | 11 | 15.00% |
12 | Stanford | 12 | 11.93% |
13 | Miami (FL) | 15 | 8.15% |
14 | Notre Dame | 18 | 7.88% |
15 | Michigan | 13 | 6.46% |
16 | Georgia | 14 | 3.76% |
17 | Louisville | 16 | 3.08% |
18 | Oklahoma State | 19 | 2.54% |
19 | Florida | 17 | 2.31% |
20 | TCU | 20 | 1.56% |
21 | Oregon | 25 | 1.54% |
22 | Washington State | 27 | 1.38% |
23 | North Carolina State | 24 | 1.16% |
24 | Texas A&M | 21 | 0.99% |
25 | Virginia Tech | 28 | 0.87% |
26 | Northwestern | 31 | 0.85% |
27 | Memphis | 39 | 0.73% |
28 | Tennessee | 22 | 0.60% |
29 | UCLA | 26 | 0.56% |
30 | Kansas State | 29 | 0.46% |
31 | Texas | 23 | 0.43% |
32 | Toledo | 55 | 0.40% |
33 | North Carolina | 34 | 0.34% |
34 | Houston | 48 | 0.31% |
35 | Boise State | 46 | 0.28% |
36 | Arkansas | 30 | 0.27% |
37 | Baylor | 36 | 0.25% |
38 | Mississippi State | 32 | 0.24% |
39 | Georgia Tech | 33 | 0.20% |
40 | South Florida | 49 | 0.18% |
41 | Ole Miss | 35 | 0.14% |
42 | Iowa | 40 | 0.11% |
43 | Nebraska | 44 | 0.10% |
44 | Pitt | 38 | 0.09% |
45 | Michigan State | 50 | 0.08% |
46 | South Carolina | 37 | 0.08% |
47 | Utah | 42 | 0.07% |
48 | Western Kentucky | 69 | 0.07% |
49 | Kentucky | 41 | 0.06% |
50 | Colorado | 43 | 0.04% |
51 | San Diego State | 61 | 0.04% |
52 | BYU | 47 | 0.03% |
53 | Tulsa | 72 | 0.02% |
54 | West Virginia | 45 | 0.02% |
55 | Western Michigan | 70 | 0.02% |
56 | Missouri | 52 | 0.01% |
57 | Appalachian State | 66 | 0.01% |
58 | Syracuse | 56 | 0.01% |
59 | Troy | 77 | 0.01% |
60 | Central Florida | 68 | 0.01% |
61 | Minnesota | 65 | 0.01% |
62 | Indiana | 57 | 0.01% |
63 | Duke | 54 | 0.01% |
64 | Arizona State | 51 | 0.01% |
65 | Vanderbilt | 53 | 0.00% |
66 | Colorado State | 63 | 0.00% |
67 | Arizona | 62 | 0.00% |
68 | Wake Forest | 60 | 0.00% |
69 | Louisiana Tech | 79 | 0.00% |
70 | Texas Tech | 59 | 0.00% |
71 | Oregon State | 58 | 0.00% |
72 | Arkansas State | 83 | 0.00% |
73 | Middle Tennessee | 86 | 0.00% |
The top four teams in the aggregate computer ratings are also the top four in playoff odds, with Alabama and Ohio State pulling away from everyone else. Ohio State has by far the best chance of finishing undefeated (25% to Bama's 18%), but Bama's SOS means that a hypothetical one-loss Tide is almost certain to make the field, which gives the teams roughly equal odds of a playoff berth. Florida State is closer to the top two in the aggregate ratings than the rest of the field, but suffers from a likely loss to Bama to start the season. Oklahoma takes the spot that most pundits have reserved for USC because 1) the Sooners will probably be better than USC in spite of losing their coach and their best skill position players, and 2) the Big 12's new title game scheme makes it really likely (87%) that the Sooners will get to play an additional game that they should be favored in.
The rest of the rankings are largely as you might expect. Notre Dame may be surprising as they aren't ranked in most publications, but there's too much talent in South Bend for the team to go 4-8 again (I hope). Breaking everything down by conference shows that the SEC and Big Ten have a leg up on everyone, as you might expect when you have half the top 15 to yourselves.
Conference | Exp Playoff Teams |
SEC | 1.204 |
B10 | 1.153 |
ACC | 0.707 |
P12 | 0.489 |
B12 | 0.348 |
IND | 0.079 |
AMER | 0.012 |
MWC | 0.006 |
MAC | 0.004 |
CUSA | 0.001 |
SB | 0.000 |
Conference Title Races
When I did this last year, Colorado and Wyoming had the lowest odds in their respective divisions, and then went on to win the damn things. So what I'm saying is that you should take teams with 2% odds at least 2% more seriously than you previously were.
Ohio State will be very good, but I can't help but feel that this overstates their odds a bit. In related news Michigan currently sits at 9% to win the division, but that's also basically where a little team called Penn State was a year ago. On the other side, Wisconsin's schedule is absolute cake. If they take care of Northwestern at home on the last day of September, I don't think anyone else catches them.
The new-fangled Big 12 title game is good in that it scraps divisions and just pairs up the top two teams. This should allow us to avoid some of the less attractive matchups we've seen elsewhere in recent years. That said, should Oklahoma really be as strong as the numbers say, this will have the unfortunate side effect of giving them the toughest possible matchup to close out the season. In other news, I'm not really sure why West Virginia is ranked to start the season. While I sure do love Dana Holgorsen on a personal level, he hasn't yet proven he can turn a rebuilding year (which 2017 most definitely is) into something special.
The numbers like a rematch of last year's title game, but there's enough turnover (with Jeff Brohm leaving WKU for Purdue) that I wouldn't be shocked if something weirder happened. Also, UAB BACK!
Toledo hasn't had a losing record since 2009, but they also haven't been to the MAC title game since 2004. Existing at the same time as the NIU dynasty and PJ Fleck's brief insurgence has made life tough for the Rockets, but this should be their year.
The Aztecs lose FBS' all-time leading rusher, but should continue to march on unabated in large part because that remains the worst division in FBS. The top of the Mountain division should once again be a lot of fun, especially if Wyoming is able to ride Josh Allen's draft hype to an upset or two.
The South is a runaway for many reasons. The Trojans will be good and they avoid North favorite Washington. UCLA is in turmoil. And everyone else doesn't quite have the talent to eclipse a fully operational USC. Fun note: Utah is just above Colorado in the computer ratings, but suffers here due to the hardest schedule in the conference, which features the top four teams from the North and a road date with USC. The Washington-Stanford game will probably decide the North, but Oregon and Washington State are still elite in enough areas to potentially become contenders (and neither of those two has to go to USC).
Auburn emerges as the main challenger to Bama because 1) they're probably a little better than LSU, 2) their cross-division games (@Mizzou, Georgia) are a little easier than LSU (@Tennessee, @Florida), and 3) they get Bama at home while LSU does not. On the other side, it's not clear that the East will be particularly great, but the boatloads of returning talent should raise the floor as well as increase the possibility of a breakout.
You made it all the way to the end of this section, so here is your reward: a Sun Belt preview. Appalachian State isn't that much better than Troy and Arkansas State, but they don't play either of them, so the road to a title (and maybe a sneaky major bowl bid) is as easy as it can be. At the bottom, Texas State was so bad last year that they find themselves behind two teams about to be kicked out of the conference (Idaho, NM State) and a team that just joined FBS (Coastal).
Weekends, Ranked
Now for my favorite part: previewing every weekend of the season, and telling you what's going to be great about them. As I've done previously, I will use a rather crude formula called "Playoff Teams Lost" that simply multiplies the odds of each team losing by 1/3 of their playoff odds. This means that middling games involving a top team will often be ranked ahead of a close-matched game between lower-ranked teams. You may disagree with that methodology, but since the Playoff is so central to the game at this point, I think it is appropriate. I have left out Week 0, as well as Championship week (there are still Sun Belt games then) and the Army-Navy week, as none of those weeks feature anything currently scheduled that will have playoff implications.
Week 1 - 0.224 playoff teams lost (Rank: 5th of 13)
1. Alabama (65% win probability) vs. Florida State - 0.160 playoff teams lost
2. Michigan (54%) vs. Florida - 0.014
3. Ohio State (95%) at Indiana - 0.012
After last year's "Best Opening Weekend Ever," this can't help but feel like a letdown. Sure, we have the most impactful game of the year in Atlanta, but no other contender faces a particularly stiff challenge to start the season. There are some nice matchups here and there (VT-WVU, UCLA-A&M, Tennessee-GT), but if you're thinking about the Playoff, there is just one game to care about.
Week 2 - 0.307 playoff teams lost (Rank: 2nd)
1. Oklahoma at Ohio State (70%) - 0.136
2. Auburn at Clemson (54%) - 0.069
3. Stanford at USC (58%) - 0.044
4. Georgia at Notre Dame (54%) - 0.019
A fine but lackluster Week 1 gives way to what I think is the best Saturday of the year. All four of these games will be great and they all run concurrently at night. Add in Nebraska's test against Oregon, a super-fun Wazzu-Boise game, the Penn State-Pitt rematch, and a half-dozen other great games, and you have a full Saturday of awesome.
Week 3 - 0.169 playoff teams lost (Rank: 10th)
1. Miami at Florida State (80%) - 0.046
2. Clemson (62%) at Louisville - 0.033
3. LSU (75%) at Mississippi State - 0.018
A couple of weeks with marquee games give way to a deeper slate without the fanfare at the top. Still, a weekend with these games as well as Texas-USC, BYU-Wisconsin, and Stanford at San Diego State should have enough intrigue to go around.
Week 4 - 0.150 playoff teams lost (Rank: 11th)
1. NC State at Florida State (86%) - 0.020
2. Oklahoma (83%) at Baylor - 0.017
3. Penn State (79%) at Iowa - 0.014
Since conferences tend to backload their biggest matchups these days, we get another subpar week as conference play gets rolling. But if the Wolfpack are going to make a move, it starts here against a Seminole team that might be beat up from playing in the featured game 2 of the first 3 weeks. So that's exciting enough.
Week 5 - 0.146 playoff teams lost (Rank: 12th)
1. Clemson (73%) at Virginia Tech - 0.021
2. USC (65%) at Washington State - 0.020
3. Northwestern at Wisconsin (80%) - 0.017
I am completely here for USC traveling to Pullman on a Friday night. I am also here for one of the few matchups that may actually challenge Wisconsin. Other than that, this is probably the weakest Saturday of the season. So maybe plan something else exciting for the last day of September.
Week 6 - 0.184 playoff teams lost (Rank: 9th)
1. Alabama (86%) at Texas A&M - 0.035
2. LSU (59%) at Florida - 0.033
3. Penn State (71%) at Northwestern - 0.021
Northwestern makes the list again! While the Wildcats should be legitimately good, their involvement in the third best game of the week can only indicate a subpar week. In NC State Watch 2017 news, the Wolfpack and their amazing defensive line host the defending Heisman winner and his terrible offensive line, so that should be fun.
Week 7 - 0.214 playoff teams lost (Rank: 6th)
1. Auburn at LSU (50.0%) - 0.069
2. Texas vs Oklahoma (80%) - 0.021
3. Ohio State (93%) at Nebraska - 0.017
Week 7 is similar to the previous weeks in its mediocre depth, but has a great feature game at the top. LSU-Auburn is not only an underrated rivalry and essentially a coin flip, but it will also likely determine who will be the main challenger to Alabama.
Week 8 - 0.214 playoff teams lost (Rank: 7th)
1. Michigan at Penn State (63%) - 0.037
2. USC (55%) at Notre Dame - 0.037
3. Louisville at Florida State (80%) - 0.032
I'm a big fan of this weekend. Michigan should have all their new places in piece by the time they face Penn State, USC faces their biggest challenge in the back half of the season, and Florida State tries to avenge last year's absolute drubbing at the hands of the Cardinals. Beyond the top games, Oklahoma travels to Manhattan, Houston and Memphis face off in a pivotal game in the American, and Boise takes on the Josh Allen hype train.
Week 9 - 0.206 playoff teams lost (Rank: 8th)
1. Penn State at Ohio State (81%) - 0.095
2. UCLA at Washington (78%) - 0.015
3. NC State at Notre Dame (65%) - 0.012
The last week before things get crazy is a nice palate cleanser. You get the one big matchup (OSU-PSU), a bunch of decent home tests for potential contenders (besides the ones above, you also get Clemson hosting GT and Oklahoma hosting Texas Tech), and the Cocktail Party, which should be the most pivotal game in the SEC East.
Week 10 - 0.302 playoff teams lost (Rank: 4th)
1. LSU at Alabama (81%) - 0.102
2. Oklahoma (72%) at Oklahoma State - 0.034
3 (tie). Auburn (71%) at Texas A&M - 0.022
3 (tie). Clemson (72%) at NC State - 0.022
An already great week becomes one of the best with two of the most impactful rivalries of the year leading the way. Add in Pac-12 North intrigue (UW-Oregon and Wazzu-Stanford), a weird road trip to Iowa for the Buckeyes, and another fun game for NC State (who will surely be first in the playoff rankings at this point), and November should begin in a glorious fashion.
Week 11 - 0.306 playoff teams lost (Rank: 3rd)
1. Florida State (55%) at Clemson - 0.092
2. Washington (50.4%) at Stanford - 0.050
3. Georgia at Auburn (74%) - 0.027
That is a stunning top three games. Divisions will be decided, and one or more legitimate playoff contender will almost certainly lose. And I also bet a dollar that all three of those games will be in different time slots, as Washington-Stanford is on Friday. We also have Oklahoma hosting TCU, Miami hosting Notre Dame, and the Tide traveling to StarkVegas, so the week is as deep as it is top-heavy.
Week 12 - 0.132 playoff teams lost (Rank: 13th)
1. Michigan at Wisconsin (63%) - 0.040
2. LSU (67%) at Tennessee - 0.024
3. UCLA at USC (75%) - 0.014
You probably can't tell from these top three games, but this is the weakest weekend of the year. Most of the best teams have non-threatening assignments such as a bye, an FCS team, or a home game with Illinois. Still, the top three plus a couple of other games that might be important if these season gets a bit weird (Okie State-KSU? Georgia-Kentucky?) should make for good enough viewing for the so-called worst weekend.
Week 13 - 0.367 playoff teams lost (Rank: 1st)
1. Alabama (67%) at Auburn - 0.124
2. Ohio State (77%) at Michigan - 0.067
3. Florida State (71%) at Florida - 0.040
After a relative down year, Rivalry Week is back with a vengeance. If the top three teams are still the top three teams at this point, they will all face road games against their most hated rivals, who should all be pretty good at the very least. Stanford-Notre Dame and the Apple Cup also happen, and NC State should be wrapping up a 12-0 regular season against the Tar Heels. What's not to love?