The games have all been played. The trophies have been handed out. The confetti cannons have all been fully discharged. The 2021-2022 college football season has come to an end.
On Jan. 10, the Georgia Bulldogs claimed the program’s first national title since 1980 by finally defeating Alabama by a score of 33-18. Over the previous month, 35 other teams also finished their seasons with a satisfying bowl victory. The Michigan State Spartans were one of those teams, due to the thrilling, come-from-behind victory over the Pittsburgh Panthers in the Peach Bowl.
Throughout the season and prior to bowl season, I provided a weekly preview and bits of advice on the games that were to come. Every week, I reported back on the action and on the accuracy of my advice. Today, it is time to provide one final recap of the 2021-2022 bowl season.
Figure 1 below shows the results of all 37 bowl games leading up to (but not including) the national title game on Monday night.
Relative to the spread, just three teams overachieved: Liberty in the LendingTree Bowl, Georgia State in the Camellia Bowl and Georgia in its Orange Bowl victory over some team from Washtenaw County. No team underachieved, yet still won. But, that doesn’t mean that no team underachieved.
On the contrary, there were a total of 18 upsets relative to the opening lines (i.e. the data points under the red line in Figure 1). The upsets, relative to the picks from the computers, are summarized below in Table 1.
In my bowl preview, my computer made a total of eight upset picks, while data that I extracted from ESPN’s FPI suggested a total of four upset picks. However, a pair of those picks were taken off the board due to COVID-related cancelations.
As a result, my computer wound up with five correct upset picks out of six, while the FPI data only got one upset pick correct out of three. All of these correct picks had spreads under three points, but 83 percent is notably good. This strong performance from my algorithm pushed the final upset tally up to 43-59 (42 percent) for the season while the FPI finished at just 29-46 (39 percent).
The high total number of upsets (18) is only slightly more than expected. My simulation suggested that we would see between 11 and 17 upsets.
Regarding picks against the spread for bowl season, none of the originally scheduled bowls met the criteria where I would issue a high confidence pick. But, for entertainment purposes, I did suggest a set of lower confidence picks. Those picks are summarized below.
Of the six recommendations that I made prior to bowl season, three of those bowls wound up being canceled, and the remaining three picks were all wrong. As you can see, there is a reason that these were not “high confidence” picks.
I considered all of these picks as merely an exhibition, so the final performance of the computers is the same as it was following championship week. But my machine ended up 60-35 (63 percent) for the year. The FPI was only 56-58 (49 percent) and the combined predictions of both computers went 97-82 (54 percent), which is consistent with the historical performance.
For the entire bowl season, my algorithm went 21-16 (57 percent) against the opening spread, while the FPI went only 16-21 (43 percent). Based on this result, my computer finished the 2021-2022 campaign with a final record of 381-388 (49.5 percent) against the opening lines. The FPI did very slightly worse at 380-389 (49.5 percent).
As for the national championship game itself, my computer was also vindicated. Georgia has been at the top of my power rankings for most of the year. I projected them as a 14-point favorite over Alabama, despite the result of the SEC Championship game. On Monday night, the Dawgs won by 15. Woof!
While the Bowl season is largely an exhibition, it does provide somewhat of a final exam for each team and, by extension, each conference. To this end, it is instructive to examine the final bowl record of each conference in comparison to the Las Vegas point spreads and to the predictions made by both my algorithm and ESPN’s FPI. Table 3 provides this comparison.
Table 1 is sorted based on the winning percentage of each conference. The rows in the “actual” section are shaded to highlight the conference with a generally better than expected performance (green), a worse than expected performance (red) or a neutral performance (yellow). The yellow shaded rows in the other parts of the table simply highlight correct overall predictions by each system.
As Table 1 shows, the Mountain West, Sun Belt and American Athletic Conferences had the highest overall winning percentages. The Mountain West was buoyed by two upset wins by San Diego State (over Texas-San Antonio) and Utah State (over Oregon State). The AAC scored two even more impressive upsets as Houston upset Auburn and UCF took out Florida.
The Sun Belt Conference also finished its bowl season with an impressive 0.750 win percentage, but Vegas and both computer systems picked the Sun Belt to run the table in bowl season. Appalachian State’s upset loss to Western Kentucky means that the Sun Belt Conference actually underachieved.
The remaining two Group of Five conferences (Conference USA and the Mid-American Conference) both overachieved with records of 3-5. C-USA can claim three total upset wins as Middle Tennessee (over Toledo) and UAB (over BYU) joined Western Kentucky as upset winners. In the MAC, both Central Michigan (over Washington State) and Western Michigan (over Nevada) won in surprising fashion.
As for the Power Five conferences, it was the Big 12 that claimed the highest win percentage with a record of 5-2. This was significantly better performance than what was predicted by either Vegas or the FPI, but my algorithm successfully predicted this gaudy record. Baylor (over Ole Miss), Kansas State (over LSU), Oklahoma State (over Notre Dame) and Texas Tech (over Mississippi State) all earned upset wins. In the final analysis, it appears that the Big 12 was a slightly stronger conference than everyone (except my computer) thought.
The Big Ten was the only other Power Five league to finish its bowl season with a winning record (6-4). This record was in agreement with the initial prediction from the Vegas lines, and slightly better than the records predicted by the computers. Michigan State (over Pittsburgh) and Maryland (over Virginia Tech) were both upset winners. Penn State and Iowa evened the scales by picking up upset losses to Arkansas and Kentucky, respectively.
As for the rest of the Big Ten, Minnesota, Ohio State, Purdue and Wisconsin all celebrated bowl victories. Meanwhile, Rutgers and Michigan both were blown out, which was not a surprise on either front.
While the SEC did claim yet another national title and both finalists, the rest of the conference did not fare well at all. In fact, eight of the remaining 10 bowl games including SEC teams ended in an upset. In five of those eight cases, it was the SEC team that lost (Auburn, Florida, LSU, Ole Miss and Mississippi State, as mentioned above)
Teams from the ACC had an even tougher bowl season. My algorithm was the most pessimistic about the conference’s prospects, but the ACC finished with two fewer wins than even I projected. The conference failed to pick up any upset wins while North Carolina (to South Carolina), Pittsburgh and Virginia Tech were all upset.
But, the basement when it comes to bowl performance belongs to the “conference of champions,” the Pac-12. The conference was not exactly expected to do well, but upset losses by Oregon State and Washington State doomed the conference to a winless holiday season.
Against all odds, we have reached the end of the college football season. But, fear not, I will be back next fall with a freshly oiled spreadsheet, ready to dispense yet another season of bad betting advice. Until next time, enjoy, and Go Green.