Usually, quality computer predictions tend to match up. For example, my member predictions are close to the SP+ predictions of Bill Connelly.
Not for this semi-final match up:
- SP+ has Michigan by 8.4 points over Alabama.
- The Power Rank has Alabama by 0.3 points over Michigan.
Bill’s numbers tend to look at season averages in making a prediction. It considers Alabama vs South Florida along with Alabama vs Georgia. You make some schedule adjustments and out pops a prediction.
This is how I used to run my college football predictions at The Power Rank. In fact, my entire company started with an algorithm that takes statistics like success rate and adjusts for opponents.
Based on data from the current season, this model at The Power Rank likes Michigan by 6.1 points, in the same neighborhood as SP+.
However, I couldn’t run this model in 2020. The pandemic didn’t allow for non-conference games that my algorithm relies on to compare teams from different conferences. I had to develop a new model.
In 2020, I started running a college football model that changed the rating of teams based on game results. It used the same data like success rate and yards per play as the old model, but now the model tended to react more to recent performance.
To my surprise, the new model had a slightly better error metric than the model that used season long averages. It also seems to track how the markets react to games.
This new model implies that Alabama, not Michigan, should be favored when the two teams meet in the Rose Bowl. This is closer to the markets that favor Michigan by 1.5 points.
Alabama isn’t the same program as before. After an amazing run NFL caliber quarterbacks, they came into the season with questions at football’s most important position.
Jalen Milroe has been good. However, even looking at the Alabama’s data without the first three games, which includes the South Florida game in which Milroe didn’t play, the offense ranks 21st in my adjusted success rate. Good, but not elite.
Also, in 2020 to 2022 NFL drafts, Alabama had five WRs drafted in the 1st round. This talent at WR has fallen off the past two seasons. Alabama didn’t have any WR taken in 2023 and only Jermaine Burton looks like a prospect in 2024.
However, Nick Saban always has a great defense, and this unit ranks 6th in my adjusted success rate. Alabama is not the same team as earlier this season, and a win over Georgia helps their rating heading into the playoff.
In contrast to the QB issues at Alabama, JJ McCarthy has had a great season for Michigan, as the offense ranks 7th in my adjusted success rate. While the run game led the way last season, the passing game has been better this season (3rd in adjusted passing success rate compared to 23rd for rushing).
Michigan peaked during the middle of the season as they hammered teams like Minnesota on the road. Despite wins over Penn State and Ohio State, they haven’t played as well in the past four games. They should have pulled away from Maryland but instead had a 50-50 type game.
Since my model has Alabama by 0.3 points, I would lean towards Alabama +1.5. However, I might be hedging my bias towards Michigan (I live in Ann Arbor).
Smarter sports betting in less than 5 minutes
This article was sent to The Power Rank’s email newsletter around noon Eastern on Friday, December 29, 2023. This newsletter is a free service that strives to be:
Every week, I send out 5-Nugget Saturday, my curated list of sports betting tips, analytics and humor. If you’re looking for action on a weekend, this is the service for you.
To get this free service delivered straight into your inbox, enter your best email address and click on “Sign up now!”