Sports betting with a PhD edge

Hey, my name is Ed Feng. I’m a huge fan of the work of Jordan Sperber at Hoop Vision.

At The Power Rank, I took my PhD from Stanford in applied math and turned it into a predictive algorithm for sports. I specialize in football (both college football and the NFL) predictions and March Madness bracket advice.

In my free email newsletter, I distill the essence of this analysis into sports betting information that strives to be:

Valuable
Concise
Entertaining

As a bonus for signing up, you’ll get my NFL quarterbacks to fade in 2021. Check out the analytics behind my contrarian take on Tennessee QB Ryan Tannehill.

To sign up for this free service, enter your best email and click on “Sign up now!”

Not sure yet? Check out this recent newsletter.

The hidden importance of NFL quarterbacks

In the NFL, passing is king. The proof of this statement comes not only from analytics but in the dramatic rise in NFL QB salaries.

The quarterback’s impact on the passing game includes not only moving the ball down field with his arm but also minimizing interceptions. However, the QB also has a critical role in another department: fumbles.

This visuals shows the distribution of fumbles lost based on type of play.

A significant 31.1% of fumbles lost happened on sacks even though only 3.6% of offensive plays ended in sacks. A higher fraction of 32.2% of fumbles lost happened on runs, but runs constituted 41.3% of offensive plays.

From the 2015 through 2020 seasons, NFL quarterbacks fumbled on 13.4% of sacks, and the offense lost 54.0% of these fumbles to the defense. In contrast, fumbles happened on 1.6% of run plays, and the defense recovered 42.2% of these fumbles.

With the abundance of sack fumbles, you might think to look at quarterback fumble rates on sacks to get that betting edge. However, the research shows that fumble rate is not predictive from year to year.

Instead, consider predicting sack rate. The more often a QB gets sacked, the more often he might fumble and potentially turn the ball over.

To predict sack rate, you might think first about analyzing the offensive line. However, analytics suggest that quarterbacks have a bigger role in sack rates.

For example, consider sack rate, or sacks divided by the sum of pass attempts and sacks, for an individual quarterback. For NFL QBs, sack rate is predictive from year to year. Sack rate has an R-squared of 24.2%, only slightly lower than the R-squared of 25.4% for completion percentage.

Dr. Eric Eager of Pro Football Focus did the same analysis with pressure rates. He found a year to year R-squared value of 32.5%.

As an example of the QB impact on sacks, consider Tom Brady. In New England from 2015 through 2019, he got sacked on 4.4% of dropbacks, an excellent rate compared to the 6.2% NFL average.

In 2020, Brady signed with Tampa Bay, and questions arose as to whether the Bucs had an offensive line that could protect him. Brady got sacked on 3.5% of dropbacks in 2020, while his offensive line had only the 14th best pass block grade out of 32 teams according to PFF.

Tom Brady is an all time great at throwing the football. To find out that he is also great at avoiding sacks is like learning that Michael Phelps is also a world class synchronized swimmer.

My analysis suggests a five year sample of sack rate is best at predicting the sack rate for next season. Let’s first look at QBs with the highest sack rates from 2016-20 compared to an NFL average of 6.2%.

Russell Wilson, 8.2%.
Tyrod Taylor, 9.4%.
Daniel Jones, 8.3%.
Ryan Tannehill, 7.5% (although he had a 4.7% rate in 2020)

These quarterbacks have some of the lowest sack rates compared to the NFL average of 6.2%.

Ben Roethlisberger, 3.0%.
Tom Brady, 3.9%.
Patrick Mahomes, 4.0%.
Derek Carr, 5.0%.
Drew Lock, 3.9%.
Jared Goff, 5.1%.