I have posted multiple snippets on this subject, but I wanted to dive a bit deeper into the subject with recent posts on the matter. Quarterback is one of the positions where some believe data has no place in evaluation. Or at the very least data should be constrained to charting and game planning only. However, there is actually a lot to statistical data at quarterback. So before I go into my findings, I want to make it absolutely clear how I gathered the data, how far back I went and what I did to the data.
First off, I gathered the data through college football reference and started off by gathering everything. Touchdowns, interceptions, yards per attempt and every other stat for quarterback. I went all the way back to 1958. In order to test the variables in different eras. And the way I normalized the eras was by taking a 10 year college football sample for each quarterback.
So a quarterback in 2016 was compared to every quarterback since 2007. While a quarterback in 1983 was compared to every quarterback since 1974. So on and so forth. This was the best way to compare eras, because what we are trying to figure out is how each quarterback performed relative to their era. Warren Moon’s stats today would look worse than Christian Hackenberg in 2015, but when compared to his era, he was a Peyton Manningesque performer.
And the last point I added was strength of schedule data using college football reference. I added this layer in order to see if how well a quarterback performed schedule wise had impact to their stats. And it would normalize quarterbacks who played weak schedules versus quarterbacks who faced tough schedules in each era. The entire sample included 4,528 quarterback performances.
After testing each variable, two variables stuck out the most correlation wise in TD/INT ratio and completion percentage. TD/INT ratio is taking into account how often a quarterback is scoring versus turning over the ball. While completion percentage speaks to a quarterback’s ability to minimize dead plays. After normalizing each quarterback performance with TD/INT ratio+completion percentage+SOS scores on a ten year sample for each performance. The results speak for themselves.
These were the quarterbacks who scored at least one 90 percentile or higher score in Complete QB Stat.
The vast majority of multiple All-Pro quarterbacks since the 1958 NFL draft class scored a 91.40 or higher score in Total QB Stat.
The vast majority of multiple Pro Bowl quarterbacks scored in the 80 percentile or higher of Total QB Stat.
And surprisingly, or not so surprisingly Brett Favre was the only All-Pro quarterback since the 1958 NFL Draft Score to score lower than a 90 or 80 percentile score in Total QB Stat. Showing that the system isn’t perfect. However, it should be clear that quality quarterback score high end scores. While average starters typically score average or below average college scores.
Matt Hasselbeck was the only multiple Pro Bowl quarterback to score lower than 50. But the question now is what to do with this data? Well it’s pretty simply like this. When projecting quarterbacks film, character profiles and age need to be examined in projection. But so does the Total QB stat. Not every quarterback who scores high in this metric will become highly impactful.
However, the majority of All-Pro and Pro Bowl quarterbacks scored in the high end. Not every QB who scores high becomes a great quarterback. Very few quarterbacks who score low become good quarterbacks. But every great quarterback scored high. Not the mention the high school data that further helps to filter out the bad from the good.
But that’s for another post. I hope explaining this data further can help you to understand it better. This is a very useful piece of data that is more correlative than any quarterback stat data out there. I don’t have a fancy acronym for it, because I don’t need one.
The results speak for themselves.