Photo credit: Keith Allison via Flickr Creative Commons.
Introduction
When I originally launched Cleveland Sports Economics, one of my first posts was the development of a statistical methodology that produced a list of historical comparables for NFL quarterbacks based on the player’s age and previous season’s statistics. With the 2018 NFL season about to kick off, I revisit this work and apply the methodology to last season’s statistics to offer each NFL QB’s closest historical comparisons.
To be clear, these comparisons do not factor in the quality of a quarterback’s supporting cast or offensive coordinator. As such, the results of this approach are hardly foolproof and typically feature a wide range of outcomes. But it is a tool that can be used to provide some insight through a historical lens and offer a glimpse of what we can expect from NFL quarterbacks—both in 2018 and beyond.
Methodology
While a full description of the statistical approach is presented here, this methodology measures each quarterback’s rate statistics in five passing categories as they compare to league average (i.e., a Z-score) in each respective season for those with at least 150 pass attempts:
- Completion percentage (comp / att)
- Yards per completion (yards / comp)
- Touchdown percentage (td / att)
- Interception percentage (int / att)
- Sack percentage [sack / (sack+att)]
Developing a list of historical comparisons between two different quarterbacks based solely on passing statistics can be problematic given that passers of two vastly different styles can produce similar numbers (think Cody Kessler vs. Cam Newton). As a result, to factor in differences in body build and athleticism, this analysis factors in two other variables as they compare to league average (again, a Z-score):
- Rush yards per game (yards / game)
- Height (inches)
After calculating the Z-score for each of the seven categories, this approach calculates the Euclidean distance between the values for every player (i.e., the square root of the sum of squared differences in Z-scores). A smaller distance score represents greater similarity between two players’ seven-category profiles; this means they are a better historical match. This is roughly equivalent to how Baseball Reference’s “Similarity Score” has been developed to find historical comparables for players in that sport; to my knowledge no such statistical approach has been created for public consumption for football.
Restricting results to players by age (e.g., only comparing Tyrod Taylor’s 2017 season to other 28-year old QBs in the NFL since 1970), the distance scores that follow can be evaluated as follows:
- 0.00-1.00 – Exceptional comparison
- 1.00-1.50 – Excellent comparison
- 1.50-2.00 – Good comparison
- 2.00-2.50 – Fair comparison
- 2.50-3.00 – Weak comparison
- 3.00 and up – Poor comparison
Results
While reviewing the results of this approach, remember two critical issues:
- These comparisons are based solely on each player’s 2017 NFL season for those with at least 150 pass attempts.
- These comparisons do not factor in a quarterback’s supporting cast, coaching staff, or strength of schedule.
Also, for comparison’s sake, the results of this approach based on the 2016 NFL season can be found at the following links: AFC and NFC.
Finally, the tables below provide the number of games started by each comparison quarterback after the season in question.
Now onto the results…
