A mechanic does not start work on a car’s engine until he or she understands what is wrong with it. In the same way, I’m sure Sean Payton knows what is “wrong” with Russell Wilson, who set career lows in almost every category in 2022. One theory on what was “wrong” with Russ in 2022 was the he suddenly became less accurate with the ball.
In this study we are going to look at the four years of advanced passing accuracy data that is available for free at Pro-football-reference.com and see if it tells anything about a number of questions
- How accurate have Bronco QBs been over the last four seasons?
- How accurate has Russell Wilson been? Was 2022 just an aberration?
- How accurate have Saints QBs been? Is that a function of Sean Payton’s offense?
To address these questions I will use two data sets - on-target % and bad throw % for every QB over the past four season who had a minimum of 50 passing attempts. Both are fairly self-explanatory, but I’ll define them anyway. PFR defines a bad throw as a poor throw that was either “not catchable” by the receiver or could have been intercepted (or both). They define a throw as “on-target” if it could have been caught by the offensive player that was being targeted with the throw. Throws that are purposefully thrown out of bounds, throw-aways, are excluded.
Historically, Russell Wilson has always had a fairly high completion percentage, but completion percentage can be helped by receivers with large catch radii and good 50/50 ball skills. Russ generally had those in Seattle.
Wilson came into the season 10th all-time in career completion percentage and fell to 18th by the end of the 2022 season. In general when a QB makes an on-target pass, it is NOT a bad pass, so there is a strong negative correlation between on-target % and bad throw %. In other words, when one goes up the other almost always goes down. Few primary QBs are far from the trendline.
The NFL correlation between on-target % and bad throw % by season for QBs with min 50 attempts
- 2022: -0.786
- 2021: -0.732
- 2020: -0.872
- 2019: -0.871
If we plot all the data over the four seasons it looks like this below: (each circle is a QB in a given season)
You can see that most QBs fall somewhere close to the linear regression trendline. QBs want to be in the bottom right. The other takeaway from this plot is that almost all primary QBs starting QBs are fairly close to the trendline meaning accurate QBs, in general, don’t make many bad throws.
The R squared value for each year’s regression line is:
- 2022: 0.617
- 2021: 0.536
- 2020: 0.761
- 2019: 0.759
This was included because one commenter said no statistical article can be taken seriously without showing R squared and correlation values. So if these mean nothing to you, then just move along.
So let’s look at where various QBs of interest to Bronco fans fall on these plots for each season where we have data.
Joe Flacco and Drew Lock were the two primary QBs for the Denver Broncos in 2019. Both were almost on the regression line. Lock is hard to see because of the other QBs in that same region with him. Teddy Bridgewater, in New Orleans, was in the “elite” zone in 2019 along with Ryan Tannehill, Drew Brees, and Derek Carr. Nick Foles, Chase Daniel and Jimmy Garoppolo were just outside the elite zone. We should note that both Saint’s QBs were elite in terms of accuracy in 2019. More on this later.
In terms of accuracy Russell Wilson was basically average in 2019. His overall accuracy was similar to Drew Lock by this metric. You can see that his point falls fairly close to Lock’s in 2019.
Please note that the axis range for these plots is not static from year to year and has been adjusted to best frame the data. That is why I showed the plot of all four years on the same graph at the start.
2020 was a good year for accuracy for Russell Wilson. Teddy B had moved to Carolina, but was still very accurate with his throws as were both Saints QBs again this season - Taysom Hill and Brees in his final season. Other elite accuracy QBs in 2020 were Aaron Rodgers, Ryan Fitzpatrick, Kirk Cousins, and Kyle Allen (with an amazing 10.1% bad throw rate).
Now playing for the Jets instead of the Broncos, Flacco had a horrible year for accuracy, as did Drew Lock in his one and only season as a primary starting QB in the NFL.
This was a “normal” year for Russ in term’s of accuracy; he was back to where he was in 2019 - essentially.
Drew Lock was about where he had been in the past (maybe slightly better) and Teddy B had a terrible year by his standards.
I have purposefully not shown the aggressiveness data the NFL NextGen Stats has. Plotting Aggressiveness vs on-target% is useful, because it shows you who is accurate on both easy and difficult throws and who is ONLY accurate because they take the easy throws. I’ll dive more into this later.
The Saints passers in 2021, Taysom Hill, Trevor Siemian and Jameis Winston were all less accurate than they had been in previous seasons. Ian Book also started a game at QB for the Saints but he only threw 20 passes which means he didn’t qualify for this plot. In other words, in Sean Payton’s final season in New Orleans, which was his only one without Brees, his QBs were not very accurate with the ball. Taysom appeared to be fairly accurate in 2020, but that may have been a function of where he was throwing the ball, which was mostly redzone and short-yardage throws that were “easy”. In 2021 he had to make all the throws and his accuracy suffered.
The group of elite accuracy QBs in 2021 was really only three guys - Joe Burrow, Colt McCoy and Jimmy G. McCoy quietly played very well in relief of Kyler Murray, completing almost 75% of his passes in 2021 while leading the Cardinals to a 2-1 record in his three starts.
Russ in 2022 was less accurate than normal. His normal 76% would have put him right on the regression line for his bad throw % this season, but he was an outlier. In other words, his bad throw rate was good - comparable to Patrick Mahomes, Trevor Lawrence and Kirk Cousins, but his on-target rate was comparable to Mitch Trubisky, Josh Allen and Jameis Winston.
We should note that Daniel Jones was by himself this year in terms of elite accuracy. This was a big improvement for him as he was not very accurate in his previous seasons in the NFL. Jones made a huge jump in on-target % (76 to 81 %) and a huge jump in bad throw % (18 down to 12) year over year. So we have precedent that big changes can happen in one off-season, but Daniel Jones is a relatively young QB. Can an “old” QB be fixed once he starts to lose his skills?
Back to Russ, has he started to lose his skills? Here’s a look at his year-by-year accuracy data:
- Season On Target % Bad Throw %
- 2019 76.3% 18.5%
- 2020 79.5% 15.8%
- 2021 76.5% 18.6%
- 2022 73.3% 15.6%
I would expect him to bounce back to his normal 76% next season, possibly reaching 80% with Sean Payton’s guidance. While that would be a big year over year jump in on-target%, it would not be the biggest among QBs with 50 or more attempts over the last four seasons. The biggest year-over-year improvement in on-target % belongs to Super Bowl starting QB, Jalen Hurts. He improved his on-target % by 17.5% (absolute) in 2021 relative to 2020. His on-target % 60.7 in 2020 on 148 attempts. That jumped to 78.2 in 2021 (and it was 77.8 in 2022). At the same time his bad throw rate dropped from 26.7 in 2020 down to 14.0 in 2021. So big changes can happen with “young” QBs, what about with “old” potential Hall of Fame QBs like Russ?
For the entire league over the last four seasons the average year over year change in on-target % is positive 0.4. Meaning that most QBs get a little bit more accurate every year, but that’s right on the edge of being statistically significant.
If we look at bad throw rate (for which we have five seasons of data), most QBs stay pretty much the same year over year (decrease on average by 0.15%). That average change is not statistically significant.
Aggressiveness AND Accuracy
Russ had a down year for accuracy, but was that a function on him being more aggressive than normal? Let’s look at the plot of aggressive throw % (throws into tight windows, double coverage or deep throws) vs on-target % for 2022 (below). You want to be in the top right on this one.
Roughly 15% of Russ’ throws in 2022 were deemed aggressive throws by NFL NGS. That was about average for primary starters in the league. You can see that Dak Prescott, Derek Carr and Ryan Tannehill were the most aggressive among guys with 8 or more starts, while Justin Fields and Patrick Mahomes (surprisingly) were both not very aggressive in 2022.
If we look at Russ’ aggressive throw % by season we find this (data only goes back to 2016).
- 2022 - 14.9%
- 2021 - 15.5%
- 2020 - 11.1%
- 2019 - 18.6%
- 2018 - 14.3%
- 2017 - 16.1%
- 2016 - 17.9%
So in terms of aggressive throw %, 2022 was an average year for Russ. NGS has their own accuracy measurement that they call expected completion % which factors in catchable passes that were not caught by the QB’s receivers. So how did Russ compare to other starters in terms of xComp% vs Agg%? See below. There are 80 unique QBs for which NGS has data from the past seven regular seasons. Using a color coding table allows is to see how Russ ranks in terms of accuracy and aggressiveness by season going back to 2016.
I have highlighted Russ in orange. You can see that he is generally average (orange or yellow) in aggressiveness and in the low end for accuracy (by xComp%), but that had a definite inflexion point. In 2020 he was on the low end for aggressiveness at 11.1%.
For the first three years of data he is on the low end for expected completion percentage, meaning that he was aided by his receivers to get his high completion percentage numbers. He then jump from roughly 60% to roughly 64% in expected completion % in his last four seasons. So he had actually gotten more accurate while staying at basically the same level of aggressiveness. The plot of xComp vs Aggressiveness for 2022 is below with QB callouts on the data points.
You can see that Russ was right in the middle of the pack in this analysis which tracks pretty well with the on-target % data for 2022. In fact if we look at the entire data set plot, you find that Russ 2022 is about right in the middle of the pack for both xComp and Agg.
We also find that the fit for this data is much worse than the fit for on-target vs bad throw %. Admittedly we have more data with seven seasons, but the R squared value for the regression line above is 0.27, which is a fairly poor fit. There are 272 data points in the plot above.
It should also be mentioned that Russ spent three seasons in the low end of the xComp spectrum, but he has improved significantly over the past four seasons. With Sean Payton calling the shots now, it is informative to see where Drew Brees plotted above. Drew Brees data by season (xComp, Agg)
- 2020 - 69.3, 12.3
- 2019 - 68.0, 13.8
- 2018 - 67.6, 15.5
- 2017 - 68.7, 16.6
- 2016 - 65.4, 17.5
Drew Brees was in the high end (green) for xComp% for every season while being average to below average in terms of his aggressiveness. So anyone expecting Russ to turn in Drew Brees, will most likely be disappointed. Drew’s worst xComp% season is better than Russ’ best for the seven seasons from which we have data.
The second part will be more history than data. It will focus on QBs who are “similar” to Russ in that they used their legs and the threat of the run as a big part of their games.
So let’s look at comparable QBs for whom we have data to see if others had down years in their early 30s and then recovered.
- Doug Flutie
- Mark Brunell
- Steve McNair
- Rich Gannon
- Warren Moon
- Steve Young
- Tony Banks
- Kordell Stewart
- Randall Cunningham
- Michael Vick
- John Elway
- Fran Tarkenton
- Cam Newton
- others who you mention in the comments - I’m sure I’ve missed some