Advertisement
Published Dec 12, 2024
Dr. Green and White Against All Odds: Breaking Down the Playoffs
circle avatar
Paul Fanson  •  Spartans Illustrated
Staff Writer
Twitter
@PaulFanson

I have spent a lot of time over the last 15 weeks of college football trying to predict how each conference will shake out and how that will ultimately impact the first ever 12-team college football playoff.

After 872 regular season and conference championship games, and after a last-second field goal lifted Clemson over Southern Methodist in the ACC Championship on Saturday night, we finally have an answer to all of these questions. Champions were crowned, bowl spots have been filled, and the bracket is now set.

Did the committee get it right? Who will win it all? Did my computer provide any good advice last week?

Let's explore all these questions, starting with the last one.

Advertisement

Championship Week Bad Betting Results

Let's take a quick look at the overview summary shown below in Figure 1 which compares the actual result of all nine games in Championship Week to the opening Vegas spreads.

The figure suggests that four of the games had a result that was close to expectation based on the opening spread. The other five games saw significant deviation, including Jacksonville State's blowout win over Western Kentucky to win the Conference USA title.

Five of the nine games ended in an upset, including Georgia over Texas in the SEC and Arizona State over Iowa State in the Big 12. Table 1 below summarizes the upsets for the weekend and compares the results to the computers' predictions.

Ironically, the computers both picked upsets in the Big Ten and ACC title games but not in the SEC or the Big 12. The only correct pick was my machine's bet on Ohio in the MAC Championship Game.

My computer's record of 1-2 (33%) brings the year-to-date tally down to 43-49 (47%). ESPN's Football Power Index went 0-2 to bring its record on upsets to 28-30 (48%) for the year.

There were no recommended bets against the spread this week, but overall my computer went 5-4 (56%) while the FPI went just 2-7 (22%). This allowed my computer to close the overall gap with the FPI to just four games. My machine now has a year-to-date record of 383-366 (51.1%) while the FPI is sitting at 387-362 (51.7%).

There was one recommended point-total bet and Army and Tulane obliged by just topping 47.5 points. This correct "lock" pick elevated the total lock record to 25-22 (53%) for the season. The complete set of recommended point-total bets has a year-to-date record of 97-69 (58%).

Did the Committee Get it Right?

Yes, they did.

The relevant conference title games played out according to my "second most likely scenario" in last week's Bad Betting Advice column. The final bracket that I projected matched the actual bracket with one exception: the inclusion of SMU over Alabama in the final spot.

But after watching SMU lose on a last second field goal, I also changed my vote prior to the release of the bracket on Sunday afternoon.

I could quibble with some details of the seeding. For example, my final college football playoff poll, which combines a predictive and a result-based metric, had both Ohio State and Indiana ranked slightly higher than the committee ranks them.

My result would have placed No. 7 Ohio State versus No. 10 Tennessee with the winner poised to take on No. 2 Georgia. No. 8 Notre Dame would still have played No. 9 Indiana, but the winner would have drawn No. 1 Oregon. I believe that my result would have resulted in a slightly more balanced bracket.

I would have swapped Arizona State and Boise State in the No. 3 and No. 4 seed positions. My data suggests Boise State is ranked too high.

I also have data that supports the argument of taking Alabama over SMU. The Crimson Tide finished No. 9 in my playoff poll. SMU finished at No. 14, behind No. 10 South Carolina, No. 11 Arizona State, No. 12 BYU, and No. 13 LSU.

But at the end of the day, the committee placed teams where they needed to go based on the poll that they issued the week prior to Championship Week. They were handcuffed by being forced to issue this poll weekly. The obviously superior model is the one used by the NCAA basketball tournament, where the bracket is simply released on Selection Sunday with no preview. But that is not what the powers that be at ESPN want, even if it leads to an inferior product.

As for Alabama's case, it is hard to feel sorry for a team last lost three regular season games, two of which were to teams that went .500 (Oklahoma and Vanderbilt). It is even more difficult to feel bad for them considering that they were selected for eight of the first 10 playoffs and they won National Titles in three of those eight appearances.

Despite the position in my final playoff poll, I would much rather give a chance to the team from a power conference that finished the regular season with a record of 11-1. The committee got it right.

College Football Playoff Analysis

Let's now dig into the bracket in more detail. There has been a lot of recent discussion that awarding first round byes to conference champions creates an imbalance in the overall bracket. I pointed out this fact in my preseason preview. Figure 2 below shows the results of a simulation that I conducted back in August.

We knew months ago that the teams seeded No. 5 and No. 6 were likely to be stronger than the teams seeded No. 3 or No. 4. We can also see from Figure 2 the advantage that the top seeded teams should receive if the teams were seeded according to their actual strength (in green).

In a "perfect bracket," the top four teams would be expected to have between a 14% and 21% chance to win the title. Getting a first-round bye is also the equivalent to about a five-percentage-point boost in odds.

As for the actual bracket in 2024, Figure 3 below gives my computer's odds for each team to win the title. In this case I used my final power ranking to estimate point spreads and win probabilities for all possible matchups. As a reference, the figure also shows the odds in the situation where the teams are reseeded to match my power rankings.

The figure shows an interesting pattern. Surprisingly, neither No. 1 Oregon nor No. 2 Georgia are are the favorites to win the National Title. Those teams have the fifth and sixth best odds, and that is because my power rankings have the Ducks and Dawgs ranked No. 5 and No. 6, respectively.

According to my computer, No. 5 seed Texas has the best odds (28%), followed by No. 8 Ohio State (22%), No. 6 Penn State (15%), and No. 7 Notre Dame (14%). This makes sense as these four teams occupy the top four in my computer's final power poll. The bottom six teams in the bracket cumulatively have just a 5.5% chance to win the title.

The other, blue-hashed bars in Figure 3 provide slightly different information. These bars represent the odds for each team to win the National Title in a reconfigured bracket where the same teams are seeded based on my power rankings. This gives an indication as to whether the actual seeding is an advantage or a disadvantage.

For example, both Ohio State (No. 1 seed according to my rankings) and Notre Dame (No. 3) have odds which are lower than expected. This makes sense, as both teams must play in the first round when the power rankings suggest that they should get a bye. If the Buckeyes and Irish win, they then draw one of the other strong teams (Oregon or Georgia) in the quarterfinals.

The seeds for Tennessee and Indiana are in line with the power rankings, but both teams are drawing more difficult opponents (Ohio State and Notre Dame) than they should.

By contrast, Oregon, Georgia, Texas (No. 2 in my rankings), and especially Penn State (No. 4) all have better odds than expected. This is surprising for the Longhorns and Nittany Lions considering that both teams would have received a first-round bye if they had been seeded based on my rankings.

The explanation is that the first round and quarterfinal opponents for both teams are the weakest four teams in the tournament. Texas' path to the semifinals goes through Clemson (No. 11) and Arizona State (No. 9) while Penn State's path goes though SMU (No. 10) and Boise State (No. 12).

In this case, playing two weaker teams in a row is mathematically better than getting a first-round bye and then playing a team such as Indiana (No. 7) or Oregon (No. 5) in the quarterfinals.

Oregon and Georgia both have slightly easier than expected path because neither team would have received a bye if the tournament had been seeded based on my power rankings.

Figure 4 below shows a slightly different way to quantify the difficulty of each team's path to a National Title.

In this analysis, I calculate the National Title odds for a strong reference team if that team were placed in any of the 12 positions on the bracket. Figure 4 shows the odds using the actual bracket (in green) and the bracket if the teams were seeded by actual strength (in hashed blue)

The reference data set (in hashed blue) shows the true impact of seeding in this general 12-team format. The reference team would have about a 30% chance to win the National Title as a No. 1 seed, but only a 12% chance if placed as the No. 12 seed. There is a clear disadvantage of being seeded No. 9 or lower and there is a steady decline in the odds for placement anywhere from No. 1 to No. 8.

The data represented by the green bars shows the true difficulty of each team's path. In this year's tournament, No. 6 Penn State (32%) has the easiest path, followed closely by No. 5 Texas (31%).

After that, the next seven teams have paths of similar difficulty. No. 2 Georgia (25%) and No. 3 Boise State (25%) are on the top of this last. No. 7 Notre Dame (23%) and No. 8 Ohio State (21%) are just a few percentage points behind, with No. 1 Oregon (20%), No. 4 Arizona State (20%), and No. 11 SMU (19%) all checking in with paths of roughly equal difficulty.

No. 10 Indiana (14%), No. 12 Clemson (13%), and No. 9 Tennessee (10%) have the most difficult paths.

In this year's tournament, my computer believes that Ohio State is the strongest team overall. However, the second-best team (Texas) has such a significant advantage in path that the Longhorns have a six-percentage point advantage in overall odds.

That all said, this analysis is predicated on the idea that my projected power rankings and point spreads are accurate. The current Vegas point spread and money line for the playoffs give slightly different values, as shown below in Figure 5.

The bookmakers in Vegas give the best National Title odds to No. 1 Oregon (+360, 22%) and No. 5 Texas (+360, 22%), with No. 2 Georgia (+450, 18%) and No. 8 Ohio State (+475, 17%) just behind.

Vegas has more confidence in Oregon and Georgia, while my model is more confident in Texas, Ohio State, and Notre Dame.

My years of studying tournaments always lead me to the same conclusion. While some paths can be slightly harder or slightly easier than others, the strongest team will usually have the best odds to win a given tournament.

However, teams like Texas and Penn State this year may have a strong enough edge in path difficulty to overcome a disadvantage in strength.

We cannot agree and likely still do not know which team is the strongest, but four rounds of exciting playoff action is the best way to find out.

Which team will win it all? My computer forecasts that the top-seeded home teams will all survive the first round. In the quarterfinals, it expects No. 8 Ohio State to exact revenge on No. 1 Oregon and for No. 5 Texas to easily dispatch with No. 4 Arizona State. In the bottom of the bracket, it expects both No. 7 Notre Dame and No. 6 Penn State to advance to the semifinals.

In the Final Four, it projects wins by No. 7 Notre Dame and No. 8 Ohio State which then leads to the Buckeyes claiming the National Title.

Against all odds, that completes the lesson for today. Enjoy bowl season and the playoffs.

Join the discussion on this article in our premium forums by clicking here.

You can also follow us on Twitter, Facebook, Threads, TikTok, Instagram, and Bluesky.

For video content, including our Red Cedar Radar and WE GOT IT podcasts, find us on YouTube and consider subscribing.