2016 Midseason Report: QB Productivity and Analysis
The 2016 NFL Regular Season is now halfway over, and every team in the league has played at least eight games. While the NFL has seen its ratings drop and the amount of penalization by the officials rise, the quality of football being played has not necessarily dropped with the ratings. We have seen plenty of good football, and that’s evidenced by the fact that 24 of the 32 teams have a 0.444 record or better, meaning that most of the teams are floating around that 0.500 record rather than several teams way above 0.500 and several way below (there are only three teams with two wins or less).
With that said, we have also a seen a lot of good quarterback play. In March, I presented my new metric, the NFL Quarterback Productivity Rating (QPR), taking the best aspects of the NFL Passer Rating and ESPN’s QB Rating (you can read all about the QPR here). I’ve worked on a few things since March with the QPR, including creating a model that projects win percentage for a quarterback given their average QPR, using the qualifying starting quarterbacks from 2013 through 2015, but that is something we will talk about later. With the season halfway in the bag, it is time for a Midseason Report.
First, I would like to show the QB Productivity Ratings for the 2016 Season thus far. The quarterbacks included are only the quarterbacks that have started at least four games in a row or five games total. They are sorted in descending order by their QPRs (Note: for the quarterbacks shown, the “2016 Productivity” is the same is the “Qualifying QPR”).
First and foremost, we can see that the average QPR of the qualifying starters is a 33.91, which is slightly higher than the 31.29 average of 2015. The standard deviation of these QPRs is 12.97, which is also slightly higher than 2015’s 11.32, meaning that the quarterbacks’ averages are a little more spread out from that average of 31.29, which is evident by Tom Brady and Drew Brees, who are two standard deviations above that average.
Now, looking at the average standard deviation for each quarterback, meaning how consistent each quarterback is performing with respect to their own individual averages, the value is a 21.07, which is slightly lower than the 22.97 from 2015, meaning that the quarterbacks have actually been more consistent week-to-week this season than last year.
We will save any actual statistical tests of the data for the end of the season, but at first glance, it appears that quarterback play is improved, as the average starting QPR is higher, and the play is also more consistent.
Plotting It Out
We can plot where all the quarterbacks stand with respect to each other, based on their standard deviations (x-axis) and their average productivity ratings (y-axis). Those above the red x-axis are above average in productivity and those below are below average. Those to the left of the green y-axis are more consistent than the average QB this season, and those to the right are less consistent than the average.
Those in Quadrant I are those performing above average but not as consistently around their average. Those in Quadrant II are performing above average but are more consistently around their average performance. Quadrant III shows quarterbacks that are consistently performing around their average, which is below the league's average, and Quadrant IV shows quarterbacks inconsistently performing around their mean that is below the league average.
The most obvious and insane thing about this all is that Tom Brady is out on his own in the Nowhere Land. He has clearly been the most productive quarterback, with his average QPR of 70.52 through four games, but he is also the most consistent quarterback. Typically, we see the top quarterbacks ending up more to the right because they have some really productive games and some less-productive games which will average out to somewhere near the 50s (like Aaron Rodgers and Ben Roethlisberger this year), but not Brady, who is on pace to have a more productive season than Peyton Manning had in 2013 (69.44 average) while being two times more consistent.
Sophomore Slumps? Not really.
On the other hand, and unsurprisingly, second-year passers Marcus Mariota, Jameis Winston and Trevor Siemian are all on the more inconsistent side, which makes sense for young quarterbacks. Mariota was the most inconsistent quarterback last year, but we’ve seen him approach his average more consistently this year. Winston has been the opposite, becoming much more inconsistent in his second year, where he was very consistent his rookie year. He has had a top five performance in three weeks this season, while having a bottom five performance twice, which makes sense why he has averaged out to just above the league mean. Overall, the first and second overall picks of 2015 are playing well, and Siemian is not playing too poorly for a first-time starter.
Photo Credit: Donn Jones, AP
Speaking of first-time starters, let’s talk about the rookies, and wow, have they been relatively good! Cody Kessler has been the worst of the trio with a 20.15 average, which is still higher than the average productivity rating of qualifying rookies from 2012 through 2015. If the season ended today, Kessler would be the 11th ranked rookie over that span, just below Mike Glennon and above Matt McGloin; Carson Wentz would be ranked 7th, just below Mariota and above Derek Carr; and Dak Prescott would be the highest ranked, a staggering 11 points above Russell Wilson. If these three continue playing around their averages, and if another rookie like Jared Goff does not come in, this rookie class will go down as the most productive rookie class of the past several years.
We can also look at the top five improvements (and regressions) from 2015 to 2016. Looking at the improvements first, there are some surprises.
Matt Ryan has been playing lights-out football for the Atlanta Falcons, and there is a huge difference between his 2015 numbers and 2016 numbers, actually doubling his 2015 average productivity rating through nine games this season for the biggest improvement of the season. What is even crazier, however, is that Brady and Brees, who were the most and third-most productive quarterbacks in 2015, respectively, have been able to improve so much in 2016, despite being 39 and 37 years old, respectively.
As far as Rodgers is concerned, his numbers have improved drastically form last season to this season, but the Packers’ win total has not. We’ve also already discussed how much Mariota has improved, but it is interesting to see that he has numerically been one of the most improved quarterbacks.
The most regressed players should not be too surprising, namely Ryan Fitzpatrick, whose 2016 average productivity rating is a third of what it was in 2015. With the exception of Fitzpatrick, the other four quarterbacks were half of the top eight most productive passers in 2015 and have all fallen below average in 2016. Wilson is dealing with an injury, but the others are just not experiencing the productivity they have in the past. I would not be surprised to see all four quarterbacks rebound in the second half of the season to go above average, but as of now, they are far worse than they were in 2015.
Our win percentage model uses the data from the 2012 through 2015 seasons, examining the correlation between the productivity ratings and win percentages. The model has an R-squared value of 0.328, meaning that 32.8% of the variance of win percentages can be explained by our model, which is just about 1.3% less than the model correlating win percentage with the NFL passer rating; therefore, this model projects win percentage about as well as a passer rating model would.
The biggest difference between the actual win percentage and the projected win percentage belongs to Kessler and the Cleveland Browns. Based on our model, the Browns should have won two of the six games that Kessler has started, but that’s not the case, given that the team has not won a game yet. The Denver Broncos, on the other hand, have outperformed their projection, winning six of the eight games that Siemian has started, where they should have only won four.
Football is a game that is played by 11 players at a time, and these differences are a nice reminder of that. While the model does give us a nice idea for the typical team, such as the Los Angeles Rams and Washington Redskins, who have won pretty much as many games as they should have, given their quarterback play, team wins are dependent on more than just the guy taking snaps.
Top and Bottom Fives
Throughout the season, I have been tweeting out the top and bottom five performances each week (you can follow me on Twitter at @KennethGoit to keep up with these every Tuesday afternoon). Now, let’s look at the top and bottom five performances of the entire season, starting with the top.
There has been a lot of good quarterback play this season (as stated earlier), and it should not be surprising to see Brees in the top five twice this season. He has played like textbook Brees, but the New Orleans Saints defense has not helped him enough.
The bottom five is headlined by Fitzpatrick, who had such a terrible in game in Week 3 against the Kansas City Chiefs, throwing six interceptions. He’s been the worst quarterback this season, and it is fitting that he has had the worst performance of the year as well.
It will be interesting to see how the rest of the season pans out, but I have found this study to be interesting and fun to watch as the season has progressed. Some quarterbacks’ productivity ratings do not give the full picture of how well or poorly they have played, namely Blake Bortles of the Jacksonville Jaguars, who has been quite bad despite not having the worst statistics, but others, such as Brady, have been summed up pretty perfectly by their Quarterback Productivity Ratings. Always remember, however, that there is more than one way to skin a cat, and that definitely applies to evaluating quarterback play in the NFL.
Kenneth Goit is an analyst for Pro Football Spot. You can send all questions and comments to him on Twitter at @KennethGoit or via email at firstname.lastname@example.org.