This article, written in August of 2021, was sent to those who joined The Power Rank’s email newsletter during the 2021 NFL season. The predictions about interception rate are evaluated.
Turnovers play a massive role in football.
Suppose a defender gets his hand on a pass. Sometimes, the ball flutters harmlessly to the ground. Other times, the ball lands in the hands of a defender who runs the other direction for a touchdown.
Despite the impact of these plays, interceptions are notoriously difficult to predict. Consider a quarterback’s interception rate, or interceptions divided by pass attempts.
An NFL quarterback’s interception rate one season has very little ability to predict interception rate the following season. For those obsessed with the beauty of least squares, the R-squared value between consecutive seasons is 7%.
However, you can make better predictions by looking at more than interceptions. As suggested earlier, the key idea is looking at plays in which defenders get a hand on a pass.
The NFL play by play data tracks passes defended, or incomplete passes in which a defender gets a hand on the ball or jars it loose with a hit. The quarterback put the ball in a dangerous situation but got lucky that the play didn’t end in a turnover.
To predict interceptions, consider bad balls, or the sum of interceptions and passes defended. For a QB, bad balls count up the times a QB puts the ball in a dangerous situation whether it ends up a turnover or not.
Now let’s consider bad ball rate, or bad balls divided pass attempts. While interception rate was not sticky between consecutive seasons, bad ball rate is predictive for NFL QBs. From one season to the next, bad ball rate has an R-squared of 27%, almost as high as completion percentage (28%).
In addition, the fraction of bad balls that turn into interceptions is random from one season to the next.
If you are going to remember one thing, it should be this: quarterbacks have control over the rate at which they put the football into a dangerous situation. Once they put the ball in danger, they have no control whether the play ends as an interception.
My research shows that 3 years of bad ball rate is the best predictor of next year’s interception rate.
To see how this works, consider the predictions heading into the 2020 season. Philadelphia QB Carson Wentz had a stellar pick rate of 1.5% the previous 3 seasons compared to the 2.4% NFL average. However, he had a bad ball rate closer to the NFL average, which suggests the un-sustainability of his low interception rate.
Wentz had an awful 2020 campaign. His interception rate skyrocketed to above 5% before settling to 3.4% for the season. He lost his starting job to rookie Jalen Hurts. In 2020, Wentz’s status fell like Julius Caesar on the floor of the Roman senate.
Let’s take a look at data on NFL QBs from 2018 through 2020 to predict the upcoming 2021 season. Over these 3 seasons, NFL QBs have averaged a pick rate of 2.3% and a bad ball rate of 10.7%.
The analytics suggests these quarterbacks have interception rates that might get even better. While we expect greatness from the top two on this list, the others might surprise you.
- Tom Brady: 1.8% pick, 8.4% bad ball. (1.7% in 2021)
- Russell Wilson: 1.6% pick, 8.2% bad ball. (1.5% in 2021)
- Marcus Mariota: 2.1% pick, 8.3% bad ball. (only 2 attempts as Carr’s back up)
- Kyler Murray: 2.2% pick, 9.3% bad ball. (2.1% in 2021)
- Cam Newton: 2.3% pick, 10.0% bad ball. (4.0% in 2021)
These quarterbacks might struggle with interceptions because of past bad ball rate.
- Ryan Tannehill: 2.1% pick, 13.6% bad ball. (2.6% in 2021)
- Daniel Jones: 2.4% pick, 14.2% bad ball. (1.9% in 2021)
- Blake Bortles: 2.7% pick, 15.6% bad ball. (out of NFL)
- Carson Wentz: 2.0% pick, 12.2% bad ball. (1.4% in 2021)
- Joe Burrow: 1.2% pick, 11.6% bad ball in rookie season in 2020. (2.7% in 2021)
- Matt Ryan: 1.7% pick, 11.3% bad ball. (2.1% in 2021)
- Gardner Minshew: 1.4% pick, 11.2% bad ball. (1.7% on 60 attempts in 2021)
Note: I’ve added the interception rates for 2021 regular season to the predictions above.
Football betting with a PhD edge
This article was sent to people that subscribed to The Power Rank’s email newsletter, a free service that strives to be:
The predictions are based on my PhD research from Stanford, and the content covers the NFL and March Madness bracket advice.
To get this free service delivered straight into your inbox, enter your best email address and click on “Sign up now!”