Russell Wilson wins all over the field

We heard it a thousand times. “Russell Wilson is too short to play QB in the NFL.”

Nevermind the fact that the average QB is 2-3 inches shorter than the average offensive lineman. Nevermind that Wilson played behind some big dudes at Wisconsin. Nevermind that Wilson has been one of the most productive QBs in the league over the past five years. Yet, some evaluators continue to claim that Wilson is physically incapable of making certain throws.

The most common is that, due to his height, he does not pass to the short middle of the field.

Thus we have a chicken-and-egg type problem. On one hand, Pete Carroll’s defensive philosophy is all about forcing teams to pass to the short middle of the field, so it stands to reason that his offensive philosophy would be the opposite. Fortunately, we can examine these questions due to the work of the #nflscrapR project, the result of Carnegie Mellon University’s sports analytics club, which provided seven seasons of NFL data for free.

To do this, I initialized a bunch of vectors and a couple arrays, then built a for loop in R and… Wait you don’t want to hear about the coding stuff? Anyways, the dataset came with a prebuilt ‘deep’ and ‘short’ designations. A ‘short’ pass is anything 15 air yards and under, anything longer than 15 air yards is a ‘deep’ pass. This table has the raw totals for Wilson over the years. If you like charts, here it is in chart form.

However, Wilson’s raw totals are not very meaningful without comparison to league average. Therefore, I converted Wilson’s numbers into something I’m calling ‘offensive %’. Offensive% simply measures what fraction of the passing offense was in a particular location. For example, we can see that in 2016, passes to the deep left were 9.8% of the Seahawks’ passing offense. I then converted the league-wide numbers into offensive% so that we can compare Wilson’s numbers to the rest of the league. From this ratio, we can determine how much more (or less) frequently Wilson passes to areas of the field. Sticking with our earlier example, Wilson passed to the deep left on 9.8% of his attempts. The league average is just under 7%. Therefore, Wilson passed to the deep left 41.3% more often than league average in 2016.

This chart shows that Wilson absolutely avoided the short middle of the field from 2012-2014. He threw to the short middle about 40% less than league average. However in 2015-2016, he has been within 10% of league average. I feel comfortable, if not confident, in saying that a 10% margin is explainable by offensive preference rather than offensive necessity. In 2015, we had two big changes to the offense. A) the OL sucked and B) Seattle added Jimmy Graham. I don’t know if 11 games of Graham is enough to offset the hypothetical effect of a terrible OL on passing to the middle of the field or if Wilson got better at attacking the short middle of the field or if Bevell/Carroll learned to trust Wilson to do so more often.

Here is the above table in a chart, because pictures are worth 1000 words.

While pass attempts can tell us about the intentions of an offense, completions offer a bit of a window in how successful that offense was at completing (pun intended) its objectives. After some more coding mumbo jumbo, I figured our Wilson’s (and the league’s) completion percentage to various parts of the field. The following table shows Wilson’s numbers for the past five years. The cells with negative values are where Wilson was less accurate than the league. I have also included a graph of Wilson’s completion percentage relative to the league figure. (Remember, this is not a difference).

As you can see, Wilson has been adept at completing passes all over the field. Only six out of a possible thirty cells are highlighted, and none in the past 2 seasons. If Wilson came into the league as a flawed passer, he has corrected a lot of those flaws by now.

Of course, these numbers are very rudimentary. Knowing where he passes and where he completes passes does not tell us if these completions are particularly useful. Fortunately, we have a stat to track that.

Expected Points Added (EPA) quantifies how many points a team is expected to earn based on field position and down and distance. Getting zero points on a drive that starts at your team’s 5-yard line is fairly expected, while getting zero points on a drive that starts at your opponent’s 5-yard line is unexpected. These situations have different Expected Points. EPA is simply the difference in expected points from before and after a play. A 90-yard run from one 5-yard line to the other has a massive EPA because the Expected Points are so different between those states. Every football fan has an intuitive understanding of this concept. EPA merely quantifies it.

Now, I don’t know the specifics of the math that the #nflscrapR team used to get the EPA numbers. That warrants a grain of salt or two. However, I will be using EPA from the same dataset, so these numbers should be comparable to other EPA findings from this project. With this caveat, here is the table of Wilson’s average EPA’s for his career, along with his weighted career averages.

The more meaningful table, is one that compares Wilson to the league. This table is the difference between Wilson’s mean EPA to the mean EPA of the league (for each depth/direction bucket). Negative numbers are where Wilson performed worse than the league.

You can see Wilson’s evolution as a passer over time with the above table in chart form. In 2014 and 2015, Wilson’s passes generated above average EPA in every single depth/direction bucket. You can see the effect of his injury when he declined relative to the league in 5/6 categories and was below average in 3 depth/direction bucks. However, he was below average by some pretty small margins. In the Short Right and Deep Right bucket, his throws were less than 0.05 EPA worse than league average.

I love charts, so here is the last one. Here is the weighted average of each of Wilson’s years. This means that his high EPA/low frequency deep passes don’t unduly influence the sample. This is compared the league average (note that passing EPA in the league is creeping up) and Russell Wilson has always been above average at generating expected points. Even with his injury last year, he had his second best season by passing EPA, and 3^rd best (and 3^rd worst) by average passing EPA relative to the league. Compared to another QB recently touted as an MVP candidate illustrates that, defying all reason, Russell Wilson is still underrated.