Thursday, August 22, 2019

College Football Playoff Predictor - Year 5

Saturday's Florida-Miami game will kick off the sixth season of the Playoff era in college football, putting us at the halfway point of the original 12-year contract.  What we've learned from the first five years with regards to the playoff committee's logic is that it's good to be an undefeated or one-loss power conference team, and....that's about it.  The two scenarios most likely to cause chaos (a title week upset or a really good two loss team) have yet to impact the landscape five years in.  There hasn't been a major title game upset since 2013 Michigan State made Urban eat sad pizza the year before the Playoff, and the first two-loss team with a strong strength of schedule in years (2018 Georgia) was drowned out by all the undefeated teams.  So even though I like to think I know what's going to happen should we get a little weirdness in 2019, I still don't know.  So here's hoping for some excitement.

My model for predicting the participants in the CFP is mostly the same as what I described last year, with one minor change.  I mentioned in my last post from 2018 that I was mildly annoyed by the two-loss stragglers having slightly higher odds than were realistic.  It's not unprecedented for a team with the resume of 2018 Michigan to make the end-of-year top four, but there clearly was not an 8% chance of that happening.  So I added another parameter to my logistic regressions which reflects the average number of playoff-worthy 0 and 1-loss teams, with the 1-loss teams being weighted at 50% (since only about half of them typically make the top four).  So if my simulation spits out 0.8 undefeated teams and 2.4 1-loss teams on average, this factor would be 0.8+(0.5*2.4), which is 2.  The coefficient for this new parameter has a negative value, which means that a higher number will make each 1-loss and 2-loss team less likely to make the Playoff, giving them lower odds in my output.  Taking the 2018 Michigan example, their odds decrease from 7.9% at the end of the year to 6.6%.  So yeah this doesn't exactly change everything too wildly, but it does make it a little bit better, which was all I expected in the first place (nerd version: my overall brier score for the whole dataset went from .100 to .089.....woohoo).

A note before we get into the good stuff:  I know that 4 years of Alabama-Clemson dominance has given us some fun games, but it's also served to delineate the sport even further into princes and paupers.  This can make an exercise like projecting CFP odds a little stale, especially when you see (spoilers) the same two teams at the top again.  That said, I think there's still two really good reasons to do this.  One, I can use the season simulation I've set up to dig into the weeds a bit more and discuss things that aren't just the Playoff.  In past years, I've talked about conference title odds and such, but I am happy to add other items to mix.  I'll try and come up with some interesting ideas, but if there's anything you would like to see dear reader, let me know.  

Second and more importantly, the dam is going to break at some point.  The best teams stay the best teams over the long run, but there's always periodic chaos or changing of the guard and we're more than due for both of those.  By running everything through my model, I can gauge the likelihood of this happening and separate out what has a slight chance of happening (LSU in the Playoff?) from what isn't going to happen (Nebraska is...going to have to wait a year).  And if that makes us all 5% smarter college football fans, then I think it's worth it.

The Playoff Probabilities

I have included every team that finished at least one simulation with two or fewer losses (or undefeated if you're Group of Five).  One thing to note is that I do not have updated S&P+ numbers, so the aggregate rank relies on the post-recruiting release from February.  This shouldn't affect things much as 1) it's only 20% of the measure, and 2) the pre-season updates generally don't shift things too much.

Rank Team Agg Rank CFP Prob
1  Clemson 2 71.16%
2  Alabama 1 63.62%
3  Oklahoma 6 46.84%
4  Michigan 5 41.13%
5  Georgia 3 29.06%
6  Ohio State 7 26.43%
7  LSU 4 22.35%
8  Notre Dame 8 14.50%
9  Washington 14 14.33%
10  Penn State 12 12.04%
11  Oregon 13 10.97%
12  Florida 9 7.71%
13  Michigan State 16 4.59%
14  Utah 18 4.35%
15  Auburn 10 3.54%
16  Mississippi State 15 3.51%
17  Florida State 21 3.39%
18  Missouri 17 2.62%
19  Iowa 19 1.84%
20  Wisconsin 24 1.57%
21  Miami (FL) 25 1.55%
22  Texas A&M 11 1.26%
23  Minnesota 38 1.19%
24  Virginia Tech 34 0.95%
25  Nebraska 31 0.89%
26  Washington State 33 0.83%
27  Texas 23 0.79%
28  Oklahoma State 27 0.78%
29  Iowa State 28 0.75%
30  Syracuse 44 0.64%
31  North Carolina State 41 0.57%
32  Stanford 30 0.56%
33  TCU 32 0.55%
34  USC 26 0.51%
35  Tennessee 22 0.49%
36  Baylor 36 0.41%
37  UCLA 37 0.27%
38  Arizona State 39 0.23%
39  Virginia 42 0.20%
40  Kentucky 40 0.15%
41  Pitt 50 0.13%
42  Boise State 35 0.13%
43  Indiana 48 0.11%
44  Central Florida 29 0.10%
45  South Carolina 20 0.08%
46  Wake Forest 61 0.05%
47  Texas Tech 47 0.04%
48  Northwestern 49 0.03%
49  West Virginia 51 0.03%
50  Memphis 45 0.03%
51  Ole Miss 43 0.02%
52  Arizona 53 0.02%
53  Boston College 63 0.02%
54  Appalachian State 54 0.01%
55  Cincinnati 46 0.01%
56  Duke 58 0.01%
57  Vanderbilt 55 0.01%
58  Purdue 60 0.01%
59  North Carolina 57 0.01%
60  Kansas State 56 0.01%
61  BYU 52 0.01%
62  Marshall 76 0.00%
63  California 59 0.00%
64  Toledo 71 0.00%
65  Arkansas 62 0.00%
66  Ohio 79 0.00%
67  Louisville 75 0.00%
68  Fresno State 64 0.00%
69  Army 68 0.00%
70  Western Michigan 69 0.00%
71  Florida International 83 0.00%
72  North Texas 85 0.00%
73  Utah State 67 0.00%
74  Troy 82 0.00%
75  Temple 77 0.00%
76  San Diego State 70 0.00%
77  South Florida 72 0.00%
78  Florida Atlantic 87 0.00%
79  Louisiana Tech 90 0.00%
80  Air Force 74 0.00%
81  UAB 100 0.00%

Clemson and Alabama are about a touchdown clear of the field in the ratings, so it only makes sense that they're well ahead of the pack here.  Clemson's easier schedule gives them a leg up on the Tide, as the Tigers have a 43% chance of going undefeated (almost exactly double Bama's chances).  Oklahoma is third on account of the ideal scenario for sneaking into the CFP: A conference with a lot of a good teams but few great ones.  And Michigan takes fourth place on account of their home-friendly schedule making things slightly easier for them (ND, Iowa, Michigan State, and Ohio State all travel to Ann Arbor).

The other noteworthy item is the usual SEC cannibalization.  The conference has ten of the 25 best teams, per the ratings systems, but of course those teams have to play each other, so they can't all make the CFP.  The most dramatic discrepancies belong to the teams that also have to play Clemson in non-conference play: Texas A&M (11th in the ratings, 22nd in the CFP odds) and South Carolina (20th in the ratings and 45th (!) in the CFP odds).

Conference Title Races

The ACC is intriguing on both sides.  Clemson somehow feels underrated at 93%, but I can see juuust enough of a faint glimmer of hope from the three teams behind them to make that 7% chance of chaos realistic.  On the Coastal side the media picked Virginia to win but they're still a clear underdog.  I understand that there has been more upheaval elsewhere and it's fun to pick the one team that's never won the division, but come on.  Should still be fun and they still could win it!


The American is almost exactly as you would expect, with UCF and Memphis both favored to repeat.  But the recent history of the conference suggests that you shouldn't sleep on the challengers.


As I mentioned earlier, the Big 12 has one elite team and a bunch of perfectly fine teams (and Kansas).  There's a lot of buzz around Texas but they'll need to show more consistency and variety on offense before we can count on another Red River Rematch.  I'm not sure if we're all ready for Baylor to be back on a national stage, but you might have to be.


Urban Meyer's "retirement" gives Michigan the upper hand in the East for once.  And while their schedule is relatively manageable, there's still too many solid teams right behind them for the Wolverines to rest easy.  The West is once again a free-for-all.  The ratings think Wisconsin is the second-best team in the division, but their schedule features a brutal cross-division slate (Michigan, Michigan State, and Ohio State) and they have to travel to Nebraska and Minnesota.


The CUSA has a lot of fine teams but no particularly good teams which means we should get some fun races.  This does allow for a potential overachiever (a la 2017 FAU) to jump out and lap the field though.


The MAC looks like the MAC of the last several years.  The only notable change is Miami making some solid progress.  And Blogacz-favorite EMU continues to hang at about 5%, which is a perfect semi-realistic underdog chance (road games at Toledo and NIU will make it hard, but they do miss Ohio completely).


Utah is the most likely Pac-12 champ which says as much about the conference as it does the Utes.  My favorite power conference underdog would probably be Arizona, which is the most tantalizing 3% I've ever seen. 


Boise continues their reign of terror over the better division of the Mountain West.  The other side is catching up quickly, with a pair of established favorites and an improving set of challengers.


The numbers say there's a 50% chance someone other than Bama and Georgia make the SEC title game, but my sense of fatalism says that is too low.  Still, as I said earlier, everyone with at least a 2% chance to win their division is a top-25 team!  So there's more potential for chaos here than maybe anywhere else.


These odds makes the Sun Belt division look about even, but the West is weaker overall, meaning that App State is close to even odds to win the whole thing.  Texas State is the most random 6% of the year so maybe keep an eye on them.


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 (and my blog), I think it is appropriate.

Week 1 - 0.099 playoff teams lost (Ranked 14th of 14)

1. Auburn (55% win probability) vs. Oregon - 0.025 playoff teams lost
2. Florida (71%) vs. Miami - 0.011
3. Georgia (90%) at Vanderbilt - 0.010

Opening week is generally decent and sometimes amazing.  This year it is the least exciting weekend by my playoff-related metric, even when including the Week Zero Florida-Miami game.  The main reason for this placement is that most of the top contenders have easy victories.  But go a little further down the list and you'll find plenty of fun matchups like Boise-FSU and BYU-Utah.  So I don't think this is the actual worst weekend, but it's still suboptimal for fully quenching our thirst for college football.

Week 2 - 0.106 playoff teams lost (Ranked 13th)

1. Texas A&M at Clemson (85%) - 0.040
2. LSU (74%) at Texas - 0.021
3. Cincinnati at Ohio State (87%) - 0.012

This may be the actual worst weekend of the season.  I'm pumped for LSU to derail the Texas hype train, and I think USC-Stanford should be a good game, but there isn't much else to get excited about here.  Maybe Army will put a scare into Michigan?

Week 3 - 0.139 playoff teams lost (Ranked 10th)

1. Oklahoma (79%) at UCLA - 0.033
2. Alabama (88%) at South Carolina - 0.027
3. Clemson (92%) at Syracuse - 0.020
4. Ohio State (80%) at Indiana - 0.018

This is a weird weekend.  On one hand we get the best collection of road tests for top contenders of the year.  On the other hand, there's a decent chance they win all these games handedly.  On the other other hand, there's a decent selection of other stuff (Iowa-Iowa State, Kentucky-Florida, Virginia-FSU) that will make this the first solid weekend of the year.

Week 4 - 0.188 playoff teams lost (Ranked 7th)

1. Notre Dame at Georgia (72%) - 0.062
2. Michigan (74%) at Wisconsin - 0.040
3. Oregon (64%) at Stanford - 0.014
4. Washington (74%) at BYU - 0.012

If the previous week was the first solid weekend of the year, this is the first...solider? weekend of the year.  We get our first matchup of contenders in ND-UGA.  We get a slate of fun SEC stuff (Florida-Tennessee, A&M-Auburn, Missouri-USC, and Vandy hosts LSU).  We get a pivotal Utah-USC matchup.  And Texas plays Oklahoma State for the rights to be the main challenger to Oklahoma in the Big 12.  Good stuff.

Week 5 - 0.132 playoff teams lost (Ranked 12th)

1. Ohio State (71%) at Nebraska - 0.027
2. Texas Tech at Oklahoma (90%) - 0.015
3. USC at Washington (72%) - 0.015

If Week 2 isn't the worst week, then it's this one.  I have some personal affinity for the OSU-Nebraska matchup, but man, I just can't find much here to get the average football fan excited for.  Hence I designate this the annual "spend time outside instead of watching football" weekend.  Enjoy.

Week 6 -  0.158 playoff teams lost (Ranked 9th)

1. Michigan State at Ohio State (72%) - 0.036
2. Iowa at Michigan (82%) - 0.030
3. Georgia (77%) at Tennessee - 0.024

This is fine.  We also have Washington-Stanford and Auburn-Florida this weekend, and that's about it.  Also, have you noticed how none of the big matchups thus far have been 50-50 propositions?  Kind of weird!  Maybe makes you think this year will be super chalky or something.

Week 7 - 0.258 playoff teams lost (Ranked 2nd)

1. Alabama (79%) at Texas A&M - 0.047
2. Florida at LSU (69%) - 0.041
3. Oklahoma (78%) vs. Texas - 0.037
4. Florida State at Clemson (91%) - 0.031
5. Penn State (56%) at Iowa - 0.021

Here we go.  We get one of Bama's biggest road tests and the Red River Shootout and the weirdly fun PSU-Iowa rivalry and a bunch of other stuff.  And Washington has to go to Tucson.  Get excited.

Week 8 - 0.256 playoff teams lost (Ranked 3rd)

1. Michigan (61%) at Penn State - 0.078
2. Oregon at Washington (55%) - 0.042
3. LSU (64%) at Mississippi State - 0.034

Here we go again.  Not only do we get the biggest game of the year to date as well as solid depth, but our top three games are all pretty close to being coin flips.  October is a good month.

Week 9 - 0.235 playoff teams lost (Ranked 5th)

1. Notre Dame at Michigan (68%) - 0.077
2. Auburn at LSU (70%) - 0.031
3. Penn State (51%) at Michigan State - 0.027
4. Wisconsin at Ohio State (77%) - 0.024

A B1G weekend indeed.  Not a ton of depth here, but this quartet of games should be enough to close out October in style.

Week 10 - 0.138 playoff teams lost (Ranked 11th)

1. Georgia (66%) vs. Florida - 0.050
2. Utah at Washington (66%) - 0.026
3. Oregon (61%) at USC - 0.015

With the extra bye week this season, there's a lot of sparse weekends, and this is one of them.  Clemson and Bama both have a bye, so we're left with some good Pac 12 action and the Cocktail Party.  Utah might already have a leg up in the South at this point, so their game with the Huskies might mean a lot more for the North.

Week 11 - 0.232 playoff teams lost (Ranked 6th)

1. LSU at Alabama (78%) - 0.105
2. Iowa State at Oklahoma (85%) - 0.026
3. Missouri at Georgia (83%) - 0.024
4. Clemson (92%) at NC State - 0.021

The biggest game of the year buoys an otherwise decent week.  We also get a few important Big Ten matchups in PSU-Minnesota and Iowa-Wisconsin.

Week 12 - 0.244 playoff teams lost (Ranked 4th)

1. Alabama (83%) at Mississippi State - 0.047
2. Georgia (61%) at Auburn - 0.045
3. Michigan State at Michigan (79%) - 0.041
4. Oklahoma (79%) at Baylor - 0.034

So I know the Bulldogs haven't had good luck against the Tide recently, but this feels like a golden opportunity.  They get Bama at home coming off of what should be a brutal fight with LSU, and for the first time in years the expectations for MSU are far more modest than they should be.  We shall see.

Week 13 - 0.181 playoff teams lost (Ranked 8th)

1. Penn State at Ohio State (64%) - 0.057
2. Texas A&M at Georgia (75%) - 0.027
3. TCU at Oklahoma (86%) - 0.024
4. Michigan (86%) at Indiana - 0.020

The week before Thanksgiving isn't particularly deep, but it does have a few bangers.  Georgia and A&M play their first conference game against each other, which is their first regular season meeting since 1980.  In other news, Ohio State plays the Nittany Lions the week before the Big Game, which they have to be super excited about.

Week 14 - 0.342 playoff teams lost (Ranked 1st)

1. Ohio State at Michigan (66%) - 0.105
2. Alabama (78%) at LSU - 0.057
3. Oklahoma (76%) at Oklahoma State - 0.040
4. Clemson (85%) at South Carolina - 0.035
5. Texas A&M at LSU (72%) - 0.024
6. Notre Dame (69%) at Stanford - 0.016

Thanksgiving week has become the annual best week of the season.  The best programs of the past five years all happen to have a major rivalry game played then, so obviously it takes on a great deal of import for the Playoff.  And the above list doesn't even include the Apple Cup or the important Big Ten West games (Minnesota-Wisconsin and Iowa-Nebraska).  Should be good.

No comments:

Post a Comment