FBS Against All Odds: Full Season Recap of College Football

It’s hard to be a college football fan in the summer. After the spring game ends, there are four long months and a few more before the next season kicks off. In that time, fans and pundits alike have more than four months to simply think, talk and dream about what the future holds for their team of choice.

In the summer of 2021, I made a series of predictions about the upcoming season based on a mathematical analysis of the preseason rankings and some historical trends. As we close the book on the 2021 season, it’s time to revisit those predictions to see how accurate they were and what that might teach us about the predictions we’ll soon hear about the 2022 season.

Last summer, the experts had very low expectations of the Michigan State Spartans. The preseason magazines had ranked MSU as high as #56 (by ESPN’s FPI) and as low as #87 (by Phil Steele). Based on this ranking, my calculations and most experts predicted that the Spartans would only win about four regular season games. Had this ranking been correct, Michigan State would have had less than a one percent chance of winning just eight regular-season games, let alone the right number of regular-season wins: 10. Of course, Michigan State would have finished 11-2 in the season after the Peach Bowl win over Pittsburgh.

How did Michigan State manage to win six games more than expected in the regular season? When a team performs above (or below) it is essentially due to a combination of three factors: the team is either better (or worse) than expected, the team’s schedule is easier (or more difficult) than expected, and the teams had good (or bad luck.”

In this case, “luck” is essentially a team’s ability to beat the odds and win more games than expected. For example, if a team is favored with exactly one touchdown in 10 consecutive games, the math says that team should win seven of those games on average and get upset in the other three games. If the team won eight or nine games, they beat the odds. Some people call this “luck,” while others call it “grit” or “execution.”

However, it is possible to independently calculate the impact of skill changes, schedule changes, and luck/grit changes on the number of expected wins relative to the pre-season values. Figure 1 below shows the results of this calculation for both the Michigan State Spartans and the Michigan Wolverines, for reference.

Figure 1: Impact of the change in skill, schedule and luck on the final win tally for both MSU and Michigan relative to the preseason forecasts.

As the figure shows, the story is similar for both the Spartans and the Wolverines in 2021. Both teams were a lot better than expected. They both won about 3.5 games more than the preseason polls suggested based purely on skill. But that would only bring Michigan State to about eight wins and Michigan to just 10 wins.

Figure 1 shows that changes in schedule strength for each team were essentially negligible. Instead, both teams were able to increase their win tally based on that mysterious factor of luck, guts, and/or execution.

For either team, this unexpected pre-season success was unlikely to happen. My full-season simulation in July, taking into account the historical uncertainty of the rankings for the season, gave Michigan State only a one percent chance of winning 10 or more games and Michigan’s chance of winning at least 11 games. was calculated as only five. per cent.

For context, Figure 2 below compares the expected post-season profit total with the expected pre-season profit total for all 130 teams at the FBS level. The deviation from the middle diagonal line represents the change in skill from the preseason rankings.

Figure 2: Comparison of projected post-season and pre-season earnings for all 130 FBS teams.

As Figure 2 shows, both the Spartans and Wolverines were among the top five or six teams for skill-based overachievement. Only Utah State, Baylor, Western Kentucky and UTEP had comparable or better numbers. Indiana, on the other hand, was the biggest underperformer of the year, with San Jose State a notable but distant second.

Figure 3 visualizes the impact of the power of schedule changes and luck on each team’s final win tally.

Figure 3: Impact of changes in schedule strength and luck/grit/execution on total wins for all 130 FBS teams.

In this case, the Spartans were the sixth happiest (gritest?) team in the country, behind only Oregon, San Diego State, Iowa, BYU, and Ole Miss. The Wolverine’s luck was good enough for the top-25.

Figure 3 also tells us that Nebraska and Colorado State were the two least fortunate teams in the country in 2021. UCLA and California both benefited the most from an easier-than-expected schedule, while Clemson and Indiana both had more difficult schedules than they first appeared. Then there’s the poor state of Boise, which had both bad luck and a tougher program than expected.

Full Season Predictions Summary

Part of any preseason analysis is a projection of which team will win each conference and division and how the New Year’s Six and College Football Playoff will shake out. In late July I presented a similar analysis where I provided two different sets of predictions for each division race based on the overall probabilities and a single “disruptive” scenario where I made some predictions regarding major disruptions that could affect each race. Tables 1 and 2 below summarize my predictions for the summer compared to the actual result.

Table 1: Summary of the expected division champions of the summer of 2021 compared to the actual winners.

Table 2: Summary of the expected Summer 2021 Congressional Champions versus the actual winners.

In all cases, the yellow shaded teams represent a correct choice. For the division races shown in Table 1, the odds-based picks only predicted the winner of five out of a total of 18 races. Based on the sum of the probabilities for the season (9.4), this value is only about half of what was expected. But the odds-based choices were more accurate than the disruptive ones this year. That method yielded only three correct choices, only one of which (Texas San Antonio) was unique.

Predictions for the conference weren’t much better, as only the Cincinnati and Louisiana-Lafayette picks were correct. Again, this was only half as much as the preseason odds suggested based on the expected value of 3.58.

Both tables also list the pre-season odds each of the eventual champions had at the start of the season. As we can see, several division and conference champions had very long chances in August. Six of the eventual champions in the division had a less than 10 percent chance in the preseason, and five of the eventual conference champions started with a less than a five percent chance.

This data suggests that 2021 was a particularly unpredictable college football season, especially at the top. The unusual nature of the 2020 season (shortened by the COVID-19 pandemic) is the most likely explanation.

As for the College Football Playoff and New Year’s Six predictions, I predicted the playoffs would feature Clemson, Ohio State, Oklahoma, and Texas A&M. In reality, none of those teams would make it to the CFP, and only the state of Ohio would bid for a New Year’s Six Bowl. My only other correct New Year’s projections were Alabama, Notre Dame, Cincinnati, and Utah.

However, the odds-based projection correctly suggested Alabama would claim a playoff spot over Texas A&M. In addition, my analysis of the odds suggested that the most likely outcome was that only one team from Clemson, Ohio State, Oklahoma, and Alabama would make the playoffs. This prediction turned out to be correct.

As for the teams that made it to the playoffs in 2021, Alabama (24 percent) and Georgia (23 percent) both had chances to be in the top five. Cincinnati (2.1 percent) and Michigan (2.4 percent) both had significantly more chances at the start of the season.

Rating of Bad Gambling Advice

The last math analysis I did before the start of the season was an analysis of some Las Vegas season bets, including the division, conference, college football playoff, and national title odds, as well as bets on the total number of regular season wins for each team. (over/under).

For each type of bet, I compared the Vegas money lines to my generated odds and calculated a return on investment (ROI) if my chances for the season were correct. Based on this analysis, I have made a series of betting recommendations. How does this advice come about? Table 3 lists the results for the highlighted Big Ten bets.

Table 3: Revision of the recommended Big Ten bets in the preseason.

The results of the Big Ten were not great. Of the 20 bets in the Big Ten with a positive ROI, only two were correct: Michigan won the Big Ten East and Illinois won more than 3.5 games. Several of these bets were in direct conflict with each other, so it would have been difficult to get more than a few correct, but the math suggested that about five bets should have been correct. Using this data to place bets would have been a losing proposition.

Table 4 below lists the results of the other high ROI bets for the rest of the country.

Table 4: Revision of National Pre-Season Recommended Bets.

At first glance, these results seem worse than the Big Ten results. Of the 24 bets in this table, only one turned out to be correct: Northern Illinois won the MAC. However, this was also one of the highest value bets on the board. A $100 bet on the Huskies would have paid $25,000, which would have more than compensated for any incorrect bets in both tables 3 and 4.

Finally, Table 5 lists the results of the recommended total win bets for the regular season.

Table 5: Revision of recommended total win bets in the regular season.

In general, the bets in tables 3 and 4 were often long shots. In contrast, the seasonal over/under bets in Table 5 were much lower risk. However, of the 30 possible bets in Table 5, only 14 were correct, which was less than the expected value of 19.2. Again, the preseason rankings (which are the main input for my calculations) this year seemed less accurate than usual. Placing all 30 of these bets would also have been a losing proposition.

That said, the choices with the higher calculated ROI were more likely to be correct. In fact, 9 of the top 14 picks came true. If someone had placed just these top-14 bets (where the ROI was over $20.00), that person would have made $421 profit on a run of $100 bets.

Well, that’s all I have to say about the 2021-2022 college football season. It was fun. I plan to be back in the summer of 2022 to review the numbers for the upcoming college football season. As always I will share with you all the advice I have to give. Until then, enjoy and go green.

Leave a Comment

Your email address will not be published.