:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/61034493/usa_today_11102581.0.jpg)
This came from the discussion section of my last article. Do preseason quarterback stats have any predictive value for the regular season? To answer this question I looked at 49 “starting QBs” - defined as players who started five or more games for two of the five seasons evaluated - over the past five seasons.
I did a correlation measurement on preseason QB rating to the regular season QB rating. I wanted to look at other stats, but there was no way to pull those stats into Excel for me to do a direct comparison so I didn’t go beyond QB rating. I had to pull every data point “by hand”.
So for the all 49 quarterbacks for the 2013 through the 2017 seasons the correlation value of their preseason QB rating to their regular QB rating is 0.07: meaning that there is essentially no correlation. So a starting quarterback may look great in the preseason and then suck in the regular season, or he may look poor in the preseason and the play great in the regular season.
For those who are interested here is a table of the correlation value for all of the 49 quarterbacks. It’s interesting that there are some quarterbacks that have very strong negative and some with very strong positive correlations during this time frame.
| QB | correlation |
| Aaron Rodgers | 0.80 |
| Alex Smith | -0.08 |
| Andrew Luck | -0.16 |
| Andy Dalton | -0.60 |
| Ben Roethlisberger | 0.19 |
| Blaine Gabbert | 0.69 |
| Blake Bortles | -0.73 |
| Brandon Weeden | -1.00 |
| Brian Hoyer | -0.11 |
| Brock Osweiler | -1.00 |
| Cam Newton | -0.20 |
| Carson Palmer | -0.35 |
| Carson Wentz | 1.00 |
| Case Keenum | -0.52 |
| Colin Kaepernick | 0.77 |
| Dak Prescott | -1.00 |
| Derek Anderson | 1.00 |
| Derek Carr | -0.49 |
| Drew Brees | -0.30 |
| Drew Stanton | 1.00 |
| EJ Manuel | 0.18 |
| Eli Manning | -0.57 |
| Geno Smith | 1.00 |
| Jake Locker | -1.00 |
| Jameis Winston | 0.27 |
| Jared Goff | 1.00 |
| Jay Cutler | 0.11 |
| Joe Flacco | 0.64 |
| Josh McCown | 0.00 |
| Kirk Cousins | -0.65 |
| Marcus Mariota | -0.92 |
| Matt Hasselbeck | 1.00 |
| Matt Ryan | -0.63 |
| Matthew Stafford | 0.59 |
| Mike Vick | -0.69 |
| Nick Foles | -0.19 |
| Paxton Lynch | 1.00 |
| Peyton Manning | 0.85 |
| Philip Rivers | -0.39 |
| Robert Griffin III | -1.00 |
| Russell Wilson | -0.55 |
| Ryan Fitzpatrick | -0.28 |
| Ryan Tannehill | -0.40 |
| Sam Bradford | -0.85 |
| Teddy Bridgewater | 1.00 |
| Tom Brady | -0.29 |
| Tony Romo | -0.98 |
| Trevor Siemian | -1.00 |
| Tyrod Taylor | 0.46 |
Note the Trevor Siemian had a -1.00 correlation (admittedly from only two seasons). How did this occur? Siemian’s ratings in 2016 were (preseason/regular season) 70.4/84.6. In 2017 these were 84.5/73.3. With only two data points this is an extremely strong inverse correlation.
Peyton Manning also had a pretty strong correlation in the three seasons that he played that were included in the study - +0.85. He had a good preseason and regular in 2013 and 2014 then he played poorly in both the preseason and the regular season in 2015.
As is always the case (say it with me): Correlation is not causation. That being said, feel free to use this data the next time someone wants to rant and rave about a starting quarterback’s preseason stats. They have about as much predictive power as the magic 8-ball in your closet.