National Championship Win Probability

Updating for game results and injuries, I’ve run my model to generate a win probability for the National Championship game on Monday between Alabama and Clemson:

Projected winner Opponent  Win Prob.
Alabama Clemson 68.37%

This equates to a line of Alabama -7 (as of this writing the Las Vegas consensus line is Alabama -6.5).

The advantage for Alabama is largely from their roster value; coaching efficiency slightly favors Clemson in a weighted average over the past 4 seasons, despite an edge to Alabama for the 2016 season (coaching efficiency measures how well teams fare compared to expectations based on roster value differences. The higher mark for Clemson here implies that Clemson has performed better than expected against their opponents over the last 4 years when isolated to roster values compared to Alabama’s expectations. The edge in 2016 for Alabama is largely due to Clemson’s loss to a less talented roster compared to no losses for Alabama). Alabama’s season performance rates a little higher than Clemson’s (again, due to Clemson’s loss), and there is no tangible home-field advantage for either team.

What does this all mean?

Alabama is more likely to win, but it would not be shocking to see Clemson come out on top. For reference, my model had the line at Alabama – 16.5 against Washington (Alabama won by 17, covering by 0.5; nice), and Clemson +3 against Ohio State (Clemson won by 31, covering 34; not so nice).

Throughout the season Alabama rarely covered the spread my model set for them (only 4 out of 14), but the average spread was about -34 so that doesn’t really imply under-performance. In fact, this is the first spread my model has for Alabama that is not double digits (the next smallest was -14 against LSU).

Clemson followed a strikingly similar path, also covering only 4 of 14 with an average spread of about -31. My model had Clemson favored by double digits in all but two games: -7 against Florida State and +3 last week against Ohio State.

A few factors to consider in this game that are not accounted for in my model:

  • Lane Kiffin is no longer running the offense for Alabama. This will be Steve Sarkisian’s first time running the show for Bama; quite a stage for a debut. Perhaps Sark has been more involved behind the scenes all year? It’s hard to imagine Saban wasn’t prepared for this scenario.
  • Will Alabama be a little complacent? Clemson fell just short in this game last year, and the revenge-seeking emotional edge could be enough to overcome the slight talent disparity.
  • Going 15-0 is really hard to do. Clemson entered this game last year at 14-0 while Alabama was 13-1. In 2014, Florida State entered the playoff at 13-0 and was obliterated by a 1-loss Oregon team. Sometimes the experience of a loss helps later in the season. Small sample size caveats aside, is this Alabama team led by a freshman QB really going to be the first to go 15-0?

16 Team Playoff

The number of bowl games has grown to the point where most of them are irrelevant. So much so that players are deciding to skip the lesser games to prepare for their post-college playing careers. Bowl games are supposed to be a reward for a great season; instead they have been watered down to merely an exhibition game that coaches can point to as an achievement to justify bonuses and contract extensions. All of this could be fixed with the implementation of an expanded playoff. I briefly touched on this in an off-season feature, but let’s now take a look at what a 16-team playoff would’ve looked like this year with the following criteria:

  1. All 10 conference champions receive an auto-bid; 6 at-large
  2. Home sites for the first 2 rounds, hosted by the higher seed

Here is what this field would look like:

Auto bids

1 Alabama

2 Clemson

4 Washington

5 Penn State

7 Oklahoma

12 Western Michigan

13 Temple

14 Western Kentucky

15 San Diego State

16 Appalachian State


3 Ohio State

6 Michigan

8 Wisconsin


10 Colorado

11 Florida State

And the bracket:

Round 1

16 Appalachian State at 1 Alabama

9 USC at 8 Wisconsin

12 Western Michigan at 5 Penn State

13 Temple at 4 Washington

11 Florida State at 6 Michigan

14 Western Kentucky at 3 Ohio State

10 Colorado at 7 Oklahoma

15 San Diego State at 2 Clemson

Yes, please! Every game is meaningful, and even in round 1 there are some fantastic match-ups. If the higher seeds advance, Round 2 looks even better:

8 Wisconsin at 1 Alabama

5 Penn State at 4 Washington

6 Michigan at 3 Ohio State

7 Oklahoma at 2 Clemson

We do get a re-match of a regular season game, but it was one of the best games of 2016 and as a traditional rivalry would have massive appeal. If the higher seeds advance in Round 2 we’d have the current playoff bracket. This format gives 15 excellent games, and could completely replace an antiquated bowl system where 38 of the 41 games are essentially exhibitions with only a handful of interesting match-ups.

A look back on the 2016 Preview

I posted a season preview in early August which included projections on conference champions and NY6/CFP participants. In the interest of maintaining accountability, let’s reflect on how those projections turned out.

Conference Champion Runner-up Conference
Florida State Clemson North Carolina VA Tech ACC
Ohio State Penn State Nebraska Wisconsin Big Ten
Oklahoma Oklahoma State Big XII
UCLA Washington Oregon Colorado PAC12
Alabama Georgia Florida SEC
South Florida Temple Houston Navy AAC
Western Kentucky Southern Miss LA Tech CUSA
Northern Illinois W. Michigan Ohio MAC
San Diego State Boise State Wyoming MtnWest
Arkansas State (co-champs) Appalachian State SunBelt

Yeesh, not great. Had a good pulse on the Big XII and Sun Belt, but a lot of swings and misses elsewhere.


Final Four picks   Bowl
Alabama Oklahoma Peach
Florida State Ohio State Fiesta
Other New Years 6 picks   Bowl
Clemson San Diego State Orange
Michigan UCLA Rose
Tennessee Notre Dame Sugar
Oregon Georgia Cotton

Picked 2 of 4 playoff teams correct, and had a 3rd in the NY6. The two other playoff picks weren’t far off, both landing in other NY6 games. Total of 6 out of 12 NY6 participants correct, but with most of the misses being comically bad.

Title Game   Location
Alabama Florida State Tampa


National Champion

Alabama is still very much in play to win the national title, and I wouldn’t change my pick now.

Win Probabilities Bowl Games

National championship game probability will be posted after teams are determined.

Projected winner Opponent  Win Prob.
Utah Indiana 90.14%
Alabama Washington 85.68%
LSU Louisville 84.93%
Mississippi State Miami OH 83.93%
Oklahoma Auburn 80.05%
Boise State Baylor 77.06%
Air Force South Alabama 75.24%
Tulsa Central Michigan 74.55%
Oklahoma State Colorado 74.38%
Pitt Northwestern 73.89%
Western Kentucky Memphis 73.49%
Florida Iowa 72.37%
Temple Wake Forest 72.02%
Old Dominion Eastern Michigan 71.19%
Tennessee Nebraska 70.50%
Washington State Minnesota 70.01%
Arkansas State Central Florida 69.29%
South Florida South Carolina 69.06%
Troy Ohio 67.71%
North Texas Army West Point 66.53%
USC Penn State 65.15%
Stanford North Carolina 63.71%
Michigan Florida State 63.63%
Middle Tennessee Hawaii 63.31%
Miami FL West Virginia 62.39%
BYU Wyoming 62.39%
Texas A&M Kansas State 60.24%
Maryland Boston College 58.91%
Ohio State Clemson 58.25%
Appalachian State Toledo 57.99%
Southern Miss Louisiana-Lafayette 57.36%
Colorado State Idaho 57.13%
Houston San Diego State 56.44%
Wisconsin Western Michigan 54.46%
New Mexico UTSA 54.03%
Kentucky Georgia Tech 53.36%
TCU Georgia 52.67%
Navy Louisiana Tech 52.65%
Virginia Tech Arkansas 51.52%
North Carolina State Vanderbilt 50.55%

BoW projected wins results

Prior to the start of the season I’d listed expected wins for all 128 FBS teams. Now that the season is over it is time to take a look at how those projections fared (all win totals referring to regular season only; excludes conference championship games):

Correct win total: 22

Off by 1 game: 33

Off by 2 games: 39

Off by 3 games: 15

Off by 4 games: 8

Off by 5 games: 6

Off by 6 games: 5

Over/Under correct picks: 61/128

Essentially, these picks are distributed close to normal:


Though I would’ve liked to see a higher number in the (-1,1) range, I am pleased that the distribution errs on the narrower side of the distribution, which implies the results are better than a normal distribution would expect to show. I’ll spend the off-season investigating why there is a spike in differences of +/- 2 games.

The most interesting aspect of large deviations from expected wins is the impact it has on head coaches. Here are the head coaching changes made so far this year, compared to where their teams fell related to expectations:

Fired coaches

Cincinnati: 3 wins below expectation (expected 7; actual 4)

Florida Atlantic: 2 wins below (5; 3)

Florida International: As expected (4; 4)

Fresno State: 3 wins below (4; 1)

Georgia State: 2 wins below (5; 3)

Indiana: As expected (6; 6) *Fired due to off-field situation

LSU: 2 wins below (9; 7)

Nevada: 1 win below (6; 5)

Oregon: 5 wins below (9; 4)

Purdue: 1 win below (4; 3)

San Jose State: 2 wins below (6; 4)

Texas: 1 win below (6; 5)

The only fired coaches who met on-field expectations were Florida International (whose low expectations are an indictment on the coach’s lack of acquisition/development of talent) and Indiana (who fired their coach for reasons independent of wins and losses). Even though some coaches were barely below expectations this year, they have been falling short for several years prior to this season as well (e.g. Purdue and Texas).

On the other end, these head coaches were hired-up due to exceeding expectations:

Tom Herman, Houston: 1 win above (8; 9)

Willie Taggart, South Florida: 3 wins above (7; 10)

Matt Rhule, Temple: 2 wins above (7; 9)

Jeff Brohm, Western Kentucky: As expected (9; 9)

It should be noted that Tom Herman was 4 wins above expected in 2015, Matt Rhule was 3 above in 2015, and Jeff Brohm was 4 above in 2015. These all contributed to elevated expectations that each coach was still able to meet or exceed in 2016.

How can this be used going forward?

The data from 2016 can be used to project who will be hot commodities and who will be on the hot seat in 2017. Coaches who found themselves on the extreme ends of this curve who remain with their 2016 teams are as follows:

Way Below Expectations (4 or more games; excluding first year coaches)

UCLA (-6) Jim Mora

Michigan State (-6) Mark Dantonio

Northern Illinois (-5) Rod Carey

Notre Dame (-5) Brian Kelly

Rice (-4) David Bailiff

Arizona (-4) Rich Rodriguez

Buffalo (-4) Lance Leipold


Way Above Expectations (4 or more games)

Colorado (+6) Mike MacIntyre

Western Michigan (+6) P. J. Fleck

Tulsa (+5) Philip Montgomery

Troy (+5) Neal Brown

Wisconsin (+5) Paul Chryst

West Virginia (+4) Dana Holgorsen

Wyoming (+4) Craig Bohl

Army (+4) Jeff Monken

Eastern Michigan (+4) Chris Creighton







Ticket Watch NY6, 12/6/16

Get-in prices for all of the NY6 bowls will updated in the Google spreadsheet every week until the week of the games. Tickets to the Cotton Bowl became comically cheap once the teams were announced; you can get in for $6 and can get a really good seat for under $50 easily. Tickets for the national championship are steadily rising, and will jump even further up if Alabama and Ohio State get there; prices will plummet if it is Washington and Clemson, and will stay about where they are if only one of Alabama or Ohio State get there.

The Rose Bowl get-in price shot up once USC was announced there, and the Orange Bowl had a slight uptick with the announcement of Michigan and Florida State. The Fiesta Bowl is very inexpensive for a playoff game, but this can be explained by two things: location and teams involved. First, unlike the Peach Bowl for Alabama, the Fiesta Bowl is too far away to drive to for both Clemson and Ohio State. Additionally, both teams’ fan bases are likely holding out for a trip the following week to Tampa. Each of these teams has played in the title game in the past two years; a semi-final is nothing special to them.

CFP Instant Reaction

The field is set:

#1 Alabama vs. #4 Washington at Peach Bowl

#2 Clemson vs. #3 Ohio State at Fiesta Bowl

Close, but out: Penn State, Michigan, Oklahoma

What went wrong?

Penn State lost a non-conference game to Pitt. Had they scheduled a super weak slate like Washington and gone 3-0 in non-conference, the rest of their schedule was strong enough to leap Washington. Lesson: Schedule non-conference games that you will win. The Big Ten is strong enough that winning the conference (or not!) is good enough to get in, so long as you don’t lose out of conference.

Michigan lost two conference games in November. Margin of loss was irrelevant; two losses is too many when 4 teams have fewer than two losses. Lesson: Schedule night games at home every year. Michigan needs to re-think their ban on night games against Michigan State and Ohio State. Both are at home next year, and at least one should be played at night. The atmosphere provides a massive home-field advantage which can make the difference in a super close game.

Oklahoma lost two non-conference games. The Sooners went 9-0 in the Big XII, which was plenty good enough to earn a bid as Alabama was the only other P5 team to go unbeaten in conference play. Lesson: Ease up a bit on the non-conference scheduling. A win against a lesser mid-major instead of a loss at Houston may have been enough to jump Washington. The non-conference slate becomes even less relevant once the Big XII adds a championship game.

It is unfortunate for college football fans as a whole that Washington is rewarded for going unbeaten in a super weak non-conference schedule. Expect to see even more cupcake games among contenders going forward.