≡ Menu

Johnson's target ratio is no joking matter

Johnson's target ratio is no joking matter.

Yards per Route Run, a metric tabulated by Pro Football Focus, is one of my favorite statistics to use to examine wide receiver performance.  To me, it’s the wide receiver version of yards per pass, as it takes production and divides that by opportunity.  However, there are some folks who prefer Yards per Target to YPRR, under the idea that a target is a better way to define an opportunity than a route.

Which view is correct?  Fortunately for our analysis, Yards per Route Run can be broken down into two metrics: Yards per Target and Targets per Route Run.  In other words, YPRR already incorporates Yards per Target, but it adjusts that statistic for Targets Per Route Run.  This makes it very easy for us to compare the two statistics: essentially, the question boils down to how valuable it is to know a receiver’s number of Targets per Route Run.

For example, Kenny Stills had the most extreme breakdown of any player in the NFL in 2013. He was off-the-charts good in yards per target (13.9), but saw targets on just 9% of his routes run last year. As a result, Stills averaged just 1.29 yards per route run, a pretty unimpressive figure.

Steve Johnson was the anti-Stills. While Johnson had the worst year of his career since becoming a Bills starter, he still managed to pull down targets on 25% of his snaps. However, he averaged only 6.3 yards target, leaving Johnson with a poor 1.56 yards per route run average. Of course, when comparing Stills’ numbers to Johnson’s, one might note that Johnson was playing with EJ Manuel and Thaddeus Lewis while Stills was playing with Drew Brees, which provides some explanation for the drastic differences between the two receivers in yards per target.1 But putting the quarterbacks issue aside, the question today is a more global one.

Since the only difference between YPRR and Y/T is the metric “targets per route run,” it’s worth asking: is Targets Per Route Run a metric worth looking at? Is it more useful than Yards per Target? Well, the word “useful” will mean different things to different people. What I’m curious about is the stickiness of each metric. And there is a pretty clear answer to that question.

Among the three metrics — YPRR, Y/T, and TPRR — it’s Targets Per Route Run that’s the most consistent from year to year. From 2007 to 2012, there were 344 wide receivers who saw at least 40 targets in Year N, and then played for the same team and saw at least 40 targets in Year N+1.2

To predict Year N+1 Yards per Route Run using Year N Yards per Route Run, the best fit formula is

N+1 YPRR = 0.843 + 0.474 * Yr N YPRR (R^2 = 0.21)

In other words, only 47.4% of a receiver’s Yards per Route Run is predictive of his YPRR in Year N+1.

What about Yards per Target? The best fit formula is:

N+1 Yd/Tar = 5.84 + 0.28 * Yr N Yd/Tar (R^2 = 0.08)

Here, we see that Y/T is not very sticky. To predict a receiver’s future yards per target, we use only 28% of his prior yards per target average.

Now, by definition, if yards per target is less sticky than yards per route run, than targets per route run has to be the stickiest. Here’s the best-fit formula:

N+1 TPRR = 0.062 + 0.671 * TPRR (R^2 = 0.41)

The number of targets a player sees per route happens to be a very sticky metric. Now, by itself, that doesn’t make Targets per Route Run a good metric.3 But it is interesting to know, and it is useful in making predictions.

A few moments ago, we used Year N YPRR to predict Year N+1 YPRR. But we can gain some precision by instead using Year N Yards per Target and Year N Targets per Route Run to predict Year N+1 Yards per Route Run:

Yr N+1 YPRR = 0.062 + 5.09 * Yr_N_TPRR + 0.0656 * Yr_N_Y/T (R^2 = 0.23)

How would this formula work for Stills and Johnson? Last year, Still averaged 1.29 YPRR, Johnson 1.56. If we wanted to predict each player’s Yards per Route Run in 20144 using just their YPRR from 2013, we would project Stills at 1.455 and Johnson at 1.58 YPRR. But if we use each player’s TPRR and Y/T from last year, Stills’ projection stays at 1.45, while Johnson’s rises to 1.74. In other words, the regression thinks Johnson’s much more likely to maintain his elite TPRR than Stills is to maintain his elite Y/T. And that makes sense, at least to me.

Finally, I thought it would be fun to use the regression formula above to predict the wide receivers with the top YPRR averages in 2014. In the table below, I’ve listed the 20 wide receivers with the highest YPRR projection based on their 2013 Yards per Target and Targets per Route Run averages (minimum 40 targets, with all data coming from Pro Football Focus). For reasons that will become evident in a moment, the far right column lists each player’s routes per team pass attempts in 2014.

RkNameTeam2013 Y/T2013 TPRR2014 Proj YPRR2013 R/TPA
1Justin BlackmonJAX90.292.110.25
2Calvin JohnsonDET10.
3Julio JonesATL10.
4Andre JohnsonHOU80.292.050.91
5Anquan BoldinSFO9.
6Josh GordonCLE110.242.020.84
7Vincent JacksonTAM7.
8Pierre GarconWAS7.
9Jerrel JerniganNYG80.282.010.24
10A.J. GreenCIN8.30.2721.01
11Antonio BrownPIT9.10.261.990.97
12Brandon MarshallCHI8.20.261.921
13Alshon JefferyCHI10.20.231.910.99
14DeSean JacksonPHI11.20.221.910.98
15Dez BryantDAL80.261.890.98
16T.Y. HiltonIND8.10.251.870.87
17Cole BeasleyDAL7.10.261.860.32
18Kendall WrightTEN8.10.251.860.95
19Victor CruzNYG8.50.241.820.83
20Demaryius ThomasDEN10.40.211.810.95

While sixteen of the wide receivers saw at least 80% of their teams passing snaps6, four receivers saw fewer than 35% of their team’s snaps. Of those, one was Atlanta’s Julio Jones, who ran routes on 93% of Atlanta’s passing plays through the first five games of the year, but missed the final eleven with a fractured foot.

The other three are kind of interesting. Justin Blackmon was in the same boat as Jones in that he only played in four games. But in those games he gained 415 yards, and a 103.8 yard per game average while playing with Chad Henne is pretty incredible. Of course, Blackmon missed the first four games of the season for violating the NFL’s substance abuse policy, and then received an indefinite suspension for again being in violation of that policy after the Jaguars’ eighth game of the year. He is unlikely to see the field in 2014, and will almost certainly go down as a colossal bust. But if he never plays again, he have caught 83 passes for 1,201 yards in his final 16 NFL games, split between his rookie and sophomore seasons. He may be a bust, but it wasn’t because of what he did on the field.

The other two “weird” names on the list are cut from different clothes than Jones or Blackmon. The Giants’ Jerrel Jernigan not only had a 66% catch rate — that’s excellent for someone playing with the 2013 version of Eli Manning — but he was targeted on a remarkable 28% of his routes in 2013! So, uh, what’s up with that? Through thirteen games, Jernigan had just 10 catches for 92 yards, and was a forgotten man behind Victor Cruz, Hakeem Nicks, and Rueben Randle. At the time, Jernigan had barely seen the field, so he hadn’t run many routes, either. Then, in week 15, Cruz was injured in the third quarter against Seattle, and did not play again in 2014. Over the final 2.5 games of the year, Jernigan caught 19 passes for 237 yards and two touchdowns. Those numbers came in garbage time against Seattle and against Detroit and Washington, so they should be taken with a grain of salt. But what we have here is a backup thrust into a starting role due to injury, and a player who then produced like a star receiver over the final 10 quarters of the season. What does that mean? Who knows, but that at least explains why Jernigan was on the list.

Cole Beasley may be the weirdest case in the group. He saw significant time in games sporadically throughout the year. Beasley spent nearly all of his time in the slot, and as a result, his playing time was tied to the health of Miles Austin. Austin missed five games: weeks 4 and 5 against Denver and Washington, and then weeks 8 through ten against the Lions, Vikings, and Saints. Beasley also saw significant action against Philadelphia in week seven, when the Cowboys were without their top two running backs. In those six games, Beasley caught 21 passes for 210 yards. While that’s not a great stat line, it is a pretty good stat line for a player who was still only getting about 20 snaps per game.

Beasley caught 72% of his targets in 2013; that sounds good, but it’s not out of line with the catch rate of a lot of other slot receivers. What’s more impressive to me is that Tony Romo threw to Beasley on 26% of his routes, which is an extremely high figure. While it’s unlikely we’ll see Beasley see significant playing time in 2014, I’m a little more interested in watching him now than I was before this post.

As for the other 16 receivers on the list? There aren’t too many surprises there, but it might be interesting to compare that list to the True Receiving Yards leader board.

  1. I suppose one counter to that would be that Stills was competing with Jimmy Graham, Marques Colston, and the Saints obsession with throwing passes to running backs, while Johnson was competing with Scott Chandler, Robert Woods, and Fred Jackson for targets. []
  2. While there are some issues with survivorship bias here, I’m not sure (1) how to get around them, and (2) that those concerns bias the results in a way that’s more biased towards one of the variables we’re examining than the others. []
  3. In some ways, TPRR is like completion percentage. It’s not very sensitive to outliers, which makes it sticky. Yards per Target, of course, is very sensitive to outliers. []
  4. Ignoring the fact that Johnson is now a 49er, of course. []
  5. Because he was below average for this data set, and everyone in the set gets regressed to the mean of the group, he benefits. []
  6. Note: A couple of wide receivers have more pass routes than their teams had pass plays. I believe this is because PFF includes passing plays called back due to offensive penalties in the number of routes run by each wide receiver. I did include sacks when calculating routes per team pass attempt, so that seems to be the only explanation. []
  • Dan

    > Yr N+1 YPRR = 0.062 + 5.09 * Yr_N_TPRR + 0.0656 * Yr_N_Y/T (R^2 = 0.23)

    It seems a little strange to be adding TPRR and Y/T to predict YPRR, when YPRR is equal to their product rather than their sum. Could you run a regression that is designed to look at their product (e.g., by taking logs)?

    • Chase Stuart

      That’s a very good comment, Dan. Unfortunately, we’re getting to the border of my ability to handle these sorts of questions in Excel. Can you walk me through what you’re suggesting?

      • Dan

        If your data set has four columns:
        Yr N Y/T
        Yr N TPRR
        Yr N YPRR
        Yr N+1 YPRR

        Then create 4 new columns by taking the natural log (ln(x)) of each of those columns:
        =ln(Yr N Y/T)
        =ln(Yr N TPRR)
        =ln(Yr N YPRR)
        =ln(Yr N+1 YPRR)

        Then run a standard linear regression (like you did with the original variables) predicting ln(Yr N+1 YPRR) based on ln(Yr N Y/T) & ln(Yr N TPRR). You can compare the R^2 to what you get when running a regression with ln(Yr N YPRR) as the only independent variable (and to the R^2 that you got in this post).

        In order to get each player’s projected YPRR for 2014, you create one column that has each player’s projected ln(Yr N+1 YPRR) (from the formula). Then you create another column with the formula =exp(x), (where x is the projected ln(Yr N+1 YPRR)). That is the projection for 2014 YPRR.

        The math here is that:
        if a = (b)(c), then ln(a) = ln(b) + ln(c)
        and e^(ln(x)) = x

        (If any of the original variables are equal to zero for any player, then this won’t work. One alternative if that is the case is to take ln(x+1) instead of ln(x).)

        • Chase Stuart

          That’s really cool. Thanks — I am using it in today’s post!

  • The PFF numbers seem pretty iffy. Kenny Stills: 641 yards, 50 targets. That’s a 12.8 average, not 13.9. Steve Johnson: 597 yards, 101 targets. That’s a 5.9 average, not 6.3. Justin Blackmon: 415 yards, 48 targets. That’s a 8.6 average, not 9.

    These are pretty sizable differences.

    And those are just the first 3 guys I checked. While I don’t know what definition PFF uses for a route, if we go with a simple “player was on the field for a passing attempt or sack, doesn’t matter if there was a penalty” I get: 1.25 yards/ route for Stills, not 1.29, 1.57 for Johnson, not 1.56, and a TPRR for Blackmon of .30, not .29.

    While the above numbers are difficult for most to calculate, the yards per target numbers aren’t. Since those seem suspect, I don’t know how much faith I have in the more difficult numbers that require nonpublic data to calculate correctly.

    • Chase Stuart

      Fair question, Chris. I know target numbers differ from site to site. PFF has Larry Fitzgerald with 129 targets. NFLGSIS has him with 136. Footballguys has him with 135. Football Outsiders has him with 134. Frustrating, but that’s life. I agree that while I respect the work of many of the PFF guys, it’s legitimate to question the faith in tabulating the nonpublic data correctly.

      • Targets are listed in the game book, there’s no reason why sites should differ from what’s on nflgsis.com.

        I have Fitzgerald with targets last year, not 136, same as FO.

        • Chase Stuart

          Sorry, I had you two reversed: FO has him with 136 http://www.footballoutsiders.com/stats/wr

        • James

          Because targets aren’t all created equal. If I’m a scrambling QB and want to throw the ball near a receiver to avoid a sack, is that considered a target? If I get the ball within an arm’s reach, sure, but what if it’s 3 yards away, or 5 yards? Or 5 yards out of bounds?

          I can easily see some sites counting those as targets, and others ignoring them because they aren’t “representative” of the player’s abilities.

        • Arthuro

          NFL game book counts throw away as target.
          They even sometimes count a target on intentional grounding penalties …. not sure what the logic of that is : intentional grounding means there was no receiver right ?

  • One other thing: it seems unusual to count plays nullified by penalty in the denominator, but not in the numerator (since a player can’t get a target or earn yards if the play didn’t count). If that’s the explanation for why there are more passing plays for some players than team passing plays I’m not sure I agree that’s valid, since it will affect players who play for different teams unequally.

    If PFF doesn’t count plays that are nullified by penalty then I would’ve stop writing as soon as footnote 6 was written.

    • David L

      Are scrambles counted as team passing plays? What about packaged plays where the QB has a choice to handoff or to make a quick throw? The WR still runs a route even if the QB hands it off since he can’t necessarily predict what the QB will choose, but will that be picked up in the count? If PFF is trying to count such plays, it could make the numbers difficult to interpret

      • Chase Stuart

        That’s a good point: PFF might be counting scrambles as pass plays. In fact, I think they probably do. My guess is on handoffs they are counting those as running plays, but that’s only a guess.

      • When checking the numbers I got closer values to what PFF listed by including plays nullified by penalty and excluding scrambles.

    • Chase Stuart

      I would agree that it’s a bit odd to include plays called back by penalty in the denominator. I have sent an email seeking clarification.

      • Andy

        (just thinking aloud, since this article is older)

        I’d be interested to hear what they say. (IMO, PFF has gotten very bad at “public relations”. They promote but don’t interact with “the rabble” often (though you’re not likely in that, as a fellow writer)).

        It IS weird to include penalty plays. On their QB accuracy pages they list types of passes they don’t intend to include (throw-aways, spikes, etc)…..seems logical you’d do the same with routes (scrambles, penalties, sacks). ALTHOUGH, I would also wonder if removing them would create a statistically significant change – if the totals are so low compared to total-routes that it wouldn’t change someone’s YPRR past the 100th’s place (and even then….between 2 players would a difference of less than 3.6″ (0.1 yards) even be meaningful?).

  • Chase Stuart

    Not to detract from what is an interesting side discussion in the comments…. are you convinced that YPRR (to the extent the underlying information is accurate) is better than Y/T?

  • GMC

    I’m interested in Beasley, because these stats confirm what I saw in the few Dallas games I watched last year:

    Cole Beasley is pretty good at getting open, but isn’t take seriously by the coaching staff because he’s half the size of a normal football player. Seriously, he looks like a child out there. Cornerbacks tower over him. Linebackers obscure him from view entirely. He looks like an oompa-loompa in a helmet.

    You can see Jason Garrett turning to the O.C. and saying “What the hell is the ball boy doing out there in the slot?”

    But he’s Really Rather Good. it will be interesting to see if he can hang on, Julian Edelman style, and make an impact.

    • Chase Stuart

      I agree with you. He both passes and fails the Eye Test.

  • Richie

    and [Blackmon] will almost certainly go down as a colossal bust.

    If true, that would be quite sad, because it appears that he has the talent to be exactly the opposite of a bust.

  • Nick Bradley

    I still prefer yards per target because of decoy routes.

    • James

      But if you are a receiver running a decoy route, doesn’t that say something about your abilities? If I’m the Colts, I’m using DHB on a lot more decoy routes than I am Hilton.

      • Nick Bradley

        I don’t think so — if DHB could catch, they’d actually throw to him.

        • Andy

          (a little late to the party, but I liked this article so I want to comment :))

          I think that’s what James was saying, that DHB CAN’T catch (ie: “he’s not good”) so they’re sending him on decoy routes & therefore deserved to be “punished” by having a route-total inflated by routes where he was “never” eligible to get yards.

  • Nick Bradley

    Breaking out Y/T or YPRR by route type is pretty ideal.

    There are a limited number of route types and depths, and you can just create those states.

    You then look at Y/T above route average, or YTARA.

  • Make sure you know how long it requires to recover from a plastic surgery process ahead
    of agreeing to it. And a proficient plastic surgeon with a lot of experience, can take that even further.
    With an type of plastic surgery, there will be a recovery process.

    Here is my page; plastic surgery breast