We all know Yards-Per-Reception. When comparing NFL pass catchers, it’s one of the first stats we look at. Player A has a 20.0 YPR? Guy is a down field threat. Player B is at 9.0? He must work from the slot. Recently, the wave of advanced statistics has brought us even newer ways to analyze receiving data. The addition of Targets to most boxscores has allowed for Yards-Per-Target. Additionally, we’ve seen Air Yards pop up for quarterbacks, which essentially removes the always unpredictable Yards-After-Catch (YAC) from the equation.
Today, I’m going to introduce a statistic that will make all of that obsolete. The stat is Average Depth of Target, or, as it shall be referred to going forward “aDOT”.
There are several issues with a stat like YPR that aDOT circumvents. For one, sample size. Considering that wide receivers generally catch only 60% of their targets, we now have, for example, 100 targets to study, rather than just the 60 receptions. That’s big.
Secondly, we can better overcome the impact of outliers. For example, if our Player A from the example earlier was targeted on a bubble screen, slipped past a poor tackle attempt, and went 80 yards for a touchdown, his YPR is going to be sky high. If that same player was targeted on three more bubble screens in that game and was tackled after the catch on each one, his YPR would still be 20.0. However, his aDOT would be right around zero. Considering the player was stopped three out of four times and how unlikely long bubble screen touchdowns are, the aDOT gives us a much better picture of this player’s role. Additionally, that same player’s YPR is going to be inflated for weeks during what is already a short season, which will distort your image of what kind of player you’re dealing with.
On the other hand, let’s again assume that Player A caught all four passes and has a 20.0 YPR, but this time he has an aDOT of 14.0 (on the high side for a pass-catcher). This player likely ran a handful of intermediate routes, and maybe even had a quick slant mixed in with a deep route. Regardless, we know that – overall – he is a player who runs intermediate-to-deep routes.
The next – and most important – reason to make the switch to aDOT is that it’s actually very predictable, which is not something that can be said for any of the aforementioned stats. In an off-season study, I ran regression and correlation analysis on all of the pass-catching stats we’re used to: catch rate, YPR, YPT, drop rate, TD rate, etc. At the end of the day, none really stood out as predictable. A normalized version of aDOT, however, had stunning results.
0.95! If you’re not familiar with a regression analysis, what this is telling us is that we can comfortably predict normalized aDOT (n-aDOT) with 95% confidence. That is almost unheard of in statistics, especially football stats.
So where did I come up with this “normalized” version of aDOT and what are my X and Y for the regression test? Without going into a ton of detail, the normalization process simply involves removing the impact of the quarterback from the player’s stat line. For example, there’s quite a bit of a difference between a 15.0 aDOT with Tim Tebow (13.2 career aDOT) and the same mark with Colt McCoy (8.2 career aDOT). For the record, regular aDOT r-squared works out to .55, which is obviously a lot less reliable, but still better than the .41 for YPR and .27 for YPT.
As for my sample for this test, I looked at wide receiver and tight end data from 2008-10 and compared it to 2011. I included only players at those positions who had a significant number of targets during three timespans: 2008-09, 2010, and 2011. 67 samples were formed, which is solid enough to run a fair analysis. We’ve only been tracking aDOT since 2008, which is why I can’t go back any further.
Obviously n-aDOT isn’t a perfect stat, so let’s examine two key issues. The first is a less severe version of what we discussed with YPR. It can be inflated (or deflated) by a few outliers. If a player sees five bubble screens and one 60 yard target, his aDOT is right around 10.0, which doesn’t give us the full picture. However, as mentioned earlier, we’re almost doubling our sample size when we move from YPR to aDOT, so it’s a major improvement.
The second issue isn’t necessarily a problem when using aDOT as a reference tool, but it is when running projections. Unfortunately, because of the unpredictability of YAC, aDOT and YPR don’t correlate as well as one would hope. If we compare our historical aDOT for each of our 67 players to their actual 2011 YPR, we get an r-squared of .36. It’s so low because YAC’s r-squared is a pitiful .15.
So, again, aDOT, especially the normalized kind, is great as a reference tool and, although it’s not a perfect solution, is a better way to run YPR projections.
Now let’s take a look at some data from 2008-2011. For some perspective, the NFL average aDOT is 0.4 for a running back, 12.0 for a wide receiver, and 8.0 for a tight end.
First, we have our wide receivers and tight ends. Shown are the Top 10 and Bottom 10 2008-11 aDOTs at these positions among players who saw, at least, 50 targets during the 2011 season.
Rk |
Player |
Targ |
n-aDOT |
Rk |
Player |
Targ |
n-aDOT |
|
1 |
Denarius Moore |
73 |
18.7 |
108 |
Antonio Gates |
87 |
7.8 |
|
2 |
Torrey Smith |
99 |
17.4 |
109 |
Aaron Hernandez |
134 |
7.8 |
|
3 |
Brandon Lloyd |
144 |
16.6 |
110 |
Owen Daniels |
86 |
7.7 |
|
4 |
Robert Meachem |
72 |
15.9 |
111 |
Preston Parker |
60 |
7.6 |
|
5 |
Darrius Heyward-Bey |
106 |
15.5 |
112 |
Jermaine Gresham |
91 |
7.4 |
|
6 |
Jonathan Baldwin |
51 |
15.5 |
113 |
Wes Welker |
193 |
7.4 |
|
7 |
Mohamed Massaquoi |
69 |
15.3 |
114 |
Heath Miller |
81 |
7.2 |
|
8 |
Mike Wallace |
122 |
15.2 |
115 |
Brandon Pettigrew |
122 |
7.1 |
|
9 |
Devery Henderson |
58 |
14.9 |
116 |
Brent Celek |
94 |
7.0 |
|
10 |
Vincent Jackson |
110 |
14.8 |
117 |
Dennis Pitta |
64 |
5.9 |
And aDOT is not just for pass catchers. Although four years doesn’t give us enough of a sample size to run a strong regression analysis, we can still learn more about quarterbacks from this new stat. We'll wrap up with the the chart below, which shows the Top 10 and Bottom 10 2008-11 aDOT figures among quarterbacks with at least 100 Aimed Throws during the 2011 NFL season.
Rk |
Quarterback |
Aimed |
aDOT |
Rk |
Quarterback |
Aimed |
aDOT |
|
1 |
Tim Tebow |
286 |
13.3 |
33 |
Matt Hasselbeck |
490 |
8.3 |
|
2 |
Vince Young |
111 |
11.6 |
34 |
Drew Brees |
730 |
8.2 |
|
3 |
Jason Campbell |
151 |
10.5 |
35 |
Blaine Gabbert |
381 |
8.1 |
|
4 |
Matt Moore |
328 |
10.4 |
36 |
Tony Romo |
497 |
8.1 |
|
5 |
Carson Palmer |
312 |
10.3 |
37 |
Alex D. Smith |
463 |
8.1 |
|
6 |
Eli Manning |
698 |
10.1 |
38 |
Ryan Fitzpatrick |
544 |
8.1 |
|
7 |
Cam Newton |
494 |
10.0 |
39 |
Donovan McNabb |
145 |
7.9 |
|
8 |
Joe Flacco |
568 |
9.8 |
40 |
Colt McCoy |
434 |
7.8 |
|
9 |
Ben Roethlisberger |
529 |
9.8 |
41 |
Josh Freeman |
519 |
7.4 |
|
10 |
Chad Henne |
102 |
9.7 |
42 |
Tyler Palko |
127 |
7.4 |
Check out Mary Pa0letti's response to this piece, which examines the Patriots 2011 aDOT numbers.
Follow Mike Clay on Twitter at @PFF_MikeClay