A written version appears below. To listen, click on the right pointing triangle or go to Apple Podcasts.
You signed up for my free email newsletter, and you’ve got the bowl cheat sheet that lets you fill out your bowl pool entry with confidence points
At the top, you see Appalachian State over UAB as the most confident game. However, App State just lost head coach Eliah Drinkwitz to Missouri.
Should you assign less confidence to this game? In this article, we’ll dig into the data on interim coaches in bowl games.
A recent study from Bill Connelly
This week, Bill Connolly of ESPN published a study on interim coaches. Based on data from 2010 to 2018, he asked whether there was betting value on fading interim coaches.
Connelly took the market values for point spreads in bowl games and asked how well it predicted the actual result of the game. He asked how well the market predicts games with the interim coach versus all the bowl games.
To measure the predictive power of the markets, he looked at the average absolute error. As an example of this metric, consider the Army vs Navy game.
The markets favored Navy by 11.5 points. Navy won the game by 24 points, so the absolute error is 24 minus 11.5. For this game, the absolute error is 12.5.
Connelly compared the average absolute error in all bowl games versus those with an interim coach. After eliminating two games with two interim coaches, he had 49 games with a single interim coach.
The average absolute error was 13.2 for all games and 13.0 for games an interim coach. This suggests that it is not more difficult for the market to predict games with an interim coach.
This study has two issues:
- a small sample size of 49 games with an interim coach
- the market could make an adjustment for the interim coach
If the market makes an adjustment, this could be why it predicts games with an interim coach with roughly the same error as in all bowl games.
Another study using analytics
In 2015, Ross Benes at Sports on Earth did a study of interim coaches. Instead of using market data, he used my team rankings at The Power Rank that take margin of victory and adjust for strength of schedule.
Over a decade ago, these team rankings allowed me to start this site and make football predictions. With my numbers, there’s no adjustment for an interim coach, as the algorithm only considers margin of victory in games before bowl season.
Benes compared 42 games with an interim to coach to 295 games without. Like Connelly, he looked at the average absolute error between The Power Rank prediction and the actual game result.
The average absolute error was 13.0 in games without an interim coach but 11.4 for games with an interim coach. Again, this study has a small sample size issue because of only 42 games with an interim coach.
However, the results of both of these studies suggest that an interim coach does not make it more difficult to predict a bowl game. They also suggests not making an adjustment to a prediction because of an interim coach.
Leave a Reply