Teams spent a lot of high draft picks on players who can help their passing attack this year, both in the draft and in trades. Brandin Cooks netted the Saints a first-round pick, and Sammy Watkins gave the Bills a second-round pick with just one year left on his contract. During the draft, two quarterbacks, three wide receivers, and one running back (Christian McCaffrey) who figures to be heavily involved in the passing game were taken in the top 10 picks. As the reliance on passing grows in the NFL, the importance of quantifying the effect of each player involved grows with it.
Last year, I introduced a new metric called Wide Receiver Efficiency Rating (WRER). You can read the introduction on Pro Football Focus. The goal is to measure each wide receiver’s impact on a per route run basis. This is not a list of the top receivers, but rather a metric that gauges how well each receiver did given the playing time he received. Now that I have another year’s worth of data to look at, I did some more testing and drew conclusions for the upcoming NFL season.
I wanted to see if WRER fluctuates from year-to-year or if it is a relatively stable statistic. In other words, I wanted to know how much efficiency was influenced by skill versus luck (which I am loosely defining as factors that the receiver cannot control). The short answer is that it’s a lot of both, but knowing where those two components come into play is important.
What is luck and what is skill?
Before I get into the analysis of how efficiency is affected by luck and skill, remember that WRER has three components: a separation score, a hands score, and an open field score. Below are the formulas that I use to calculate each part:
** I should quickly note that Football Outsiders no longer tracks catchable balls in the way that they used to. I am currently using a regression to estimate catchable balls that involves the accuracy of the quarterback and the catch rate of the receiver. My regression isn’t perfect, but it does explain more than 60 percent of the variance in catchable balls, which is much more than drops plus catches.
In order to figure out how much efficiency is influenced by luck or skill, I regressed 2016-17 WRER onto 2015-16 WRER for receivers who played both seasons. The R-Squared values should give me a pretty good indication of how repeatable WRER scores are, and thus how much the value is related to skill.
The metric as a whole did not show great correlation from year to year. The R-Squared value in this case was just .35. That number isn’t the worst amongst sports metrics, but it isn’t good, either. It’s hard to argue that one’s efficiency is a good predictor of his efficiency the following season.
The separation score includes Clay’s aDOT in the calculation, so perhaps it is not surprising that the value is in the same area. Nevertheless, I am measuring something different with WRER, so my metric should be compared to other metrics that attempt to gauge how much receivers affect their teams when on the field. In the same article, Clay notes that the R-Squared value for Yards per Reception is .41, and the value for Yards per Target is .27. In other words, the separation component of WRER seems to be more predictable than Yards per Reception or Yards per Target. However, this is mostly because the separation component of WRER does not include yards after the catch.
According to Clay, Yards After Catch has an R-Squared value of just .15. That explains both why the separation component is so much more predictable than the total WRER and why the open field score is essentially unpredictable.
This analysis once again shows that the separation component of WRER is the most important. In that Pro Football Focus article, I explained that the separation score tends to have the greatest effect on the total WRER. If you believe that I created a sound formula, then you also believe that separation will improve efficiency more than having good hands or getting yards after the catch (except in extreme cases). This analysis shows that having a good separation score is much more repeatable from year to year than having a good hands score or open field score.
Perhaps that last point sounds counterintuitive, particularly with the hands part. Essentially, the influence of skill is has a greater impact on the separation score than it does on the hands and open field scores. On the other hand, the influence of luck (or at least factors that the receiver can’t control) has a greater impact on the hands and open field score than it does on the separation score. Luck and skill are not inversely proportional, as there is a good deal of both involved in all three scores, but it’s clear that luck dominates skill with the hands and open field scores.
This does not necessarily mean that catching balls is not a skill. A receiver’s ability to catch passes is absolutely a skill, and we know this because the same players tend to drop the most or the fewest easy passes. But look closely at the way I created the metric. WRER doesn’t care what passes the receiver did or didn’t catch as long as he could have caught them. This means that balls thrown at a receiver’s knees are counted the same as those thrown at his chest. If I were trying to rate a receiver’s ability to catch passes, I would change that. But I am measuring efficiency, and for that, all we want to know is how often he corralled balls when he had the chance to do so. Thus, the hands score is largely influenced by fluctuations with quarterback throws. The receiver may have great hands, but even Larry Fitzgerald and Odell Beckham Jr. will miss some passes thrown at or below their knees.
We see a similar story with the open field score. Certain players and receivers can make would-be tacklers miss more than others. However, WRER doesn’t care how many tackles a receiver breaks, it only cares how many yards after the catch you got. I suppose better separation can influence this at times, but it is likely more a product of play calling. Receivers who get many screen passes should be able to pick up more after the catch than those running go routes.
Upon completing this analysis, I came to a similar conclusion that I came to when I first introduced the model. There are many skills that can help a receiver, but his ability to get open downfield will help him and his team the most. Open field jukes look great on television, but a team can’t rely on them as an offense. Receivers who catch everything thrown their way are great, but you likely won’t get as much value from it as a guy who gets open often enough that the quarterback wants to throw everything his way.
I plan on doing a multi-year version of this analysis when I have enough data points for it. I only have three years of data, and it isn’t the same group of players, so I need more data points before I try to make more conclusions. I also want to use the metric to establish a trade-off between playing time and efficiency, but need more data for that, too.
For the rest of the article, I will highlight some guys who did very well or very poorly in WRER from the 2016-17 season.
Top WRER Performers
As usual, we have a mix of star talents and surprises at the top. Of the top 10, three (Sammie Coates, JJ Nelson, and Taylor Gabriel) did not even play 50 percent of their teams’ snaps. Gabriel came on strong at the end of the season, and this is actually the second time he has ended up in the top 10. Gabriel finished 7th in WRER in 2014 before ended up 110th in 2015. Because of that drop-off, I labeled Gabriel as one of the biggest disappointments the season before. Gabriel is likely somewhere in the middle, as his high open field score makes him an unlikely repeat for 2017. However, he was able to shake off that disappointment label two seasons ago and is likely in line for a larger role. If nothing else, Gabriel does appear to be one of the more efficient backups in the league.
Worst WRER Performers
I mentioned last year that players who show poor efficiency rarely have that “breakout” season. Agholor played a lot of snaps, and got nothing done for the Eagles. Agholor ranked 123rd out of 137 qualified receivers in separation score. Including all receivers, he had more targets per route run than just 10 other receivers (out of 204). If Agholor really is gearing up for a big season, then the entire Eagles coaching staff deserves a world of credit for their work with him.
Outside of Agholor, it’s notable that three receivers for Houston showed up in this space. The good news for Houston is that the two receivers who figure to be on the field all the time, Will Fuller (28th) and DeAndre Hopkins (48th) did just fine in WRER. They could certainly use an efficient backup, but they use their tight ends a lot in the passing game. Even though Houston shows up three times in the bottom ten receivers, this isn’t really a problem for the team right now.
Efficient While Playing Less Than 50 Percent of Snaps
Tyreek Hill is a name that excites many because of his big plays on both offense and special teams. Hill’s speed and ability to lose his coverage deep downfield makes him a good bet to be an efficient receiver. Alex Smith might not throw deep often enough to take full advantage of Hill’s skills, and my metric reflects that. Hill’s normalized WRER is 4th among qualified receivers, behind just Julio Jones, AJ Green, and Sammie Coates. If Hill keeps finding space downfield (like he did on Thursday night), then he will continue to post a high WRER.
Inefficient While Playing More Than 50 Percent of Snaps
A high WRER in a small sample size does not mean that a high WRER in a large sample size is forthcoming. However, a low WRER in a small sample size usually means that player will struggle in a large sample size.
Last year, I highlighted the players many would discuss as breakout picks. Along with their WRER ranking from 2015, they were: Stefon Diggs (30th), Tyler Lockett (41st), DeVante Parker (44th), Willie Snead (50th), Dontrelle Inman (71st), Jaelen Strong (81st), Chris Conley (97th), and Nelson Agholor (125th). In 2016 here is how those players performed by WRER: Parker (30th), Lockett (58th), Diggs (73rd), Snead (81st), Inman (89th), Conley (123rd), Agholor (130th), Strong (136th).
Once again, the bottom stayed the bottom and the top stayed the top, though you could argue that Diggs’ regression in WRER was more than expected (and perhaps injury-induced). This year, here are the names that analysts are talking about for breakout picks, along with their 2016 WRER ranking: Terrelle Pryor (11th), Tyreek Hill (18th), DeVante Parker (30th), Corey Coleman (59th), Stefon Diggs (73rd), Willie Snead (81st), Jamison Crowder (82nd), Nelson Agholor (130th).
First of all, the fact that we are saying many of the same names this year probably means that we missed last year. Secondly, this group of potential breakouts seems significantly older than in previous years. Pryor, in particular, is 28 years old. But only two players on that list (Hill and Coleman) were rookies last year. Many of these players probably won’t have expanded roles, so the assumption is that a new scheme (Pryor) or self-improvent (Parker, Crowder, Agholor) will drive the breakout. WRER says nothing about a player’s ability to improve, but suggests that an inefficient player on fewer snaps will not be efficient on more snaps. For this group of breakouts, WRER has less predictive power, but can still provide insights in a few areas.
As for the two rookies, Hill and Coleman, I think both have a decent chance of breaking out (or just getting even better in Hill’s case). Coleman was held back by a 0.759 hands score, but his separation ranked 36th among qualified receivers. Coleman’s average depth of target of 16 yards was in the top 15 among qualified receivers. Quarterback play undoubtedly contributed to some of his poor hands score, but Coleman did show an ability to get deep separation. Rookie quarterbacks sometimes are put on a leash, but DeShone Kizer was not a dink-and-dunk quarterback in college. Coleman has the skill set to be an efficient receiver, but we will see how much his quarterback holds him back.
As far as the others, I wouldn’t put much stock into Snead, Crowder, or Agholor. Snead played a larger workload last season, and it didn’t go so well (although his separation score was not far behind that of teammate Michael Thomas). New Orleans may need someone to replace Brandin Cooks, but Snead was already given that chance.
Jamison Crowder needs to start going deep more often if he wants to improve his efficiency. Slot receivers always look nice because they catch a high percentage of their targets and (sometimes) get yards after the catch. But it generally is not an efficient way to play football. As counterintuitive as it sounds, teams that don’t throw the ball deep are the least efficient offenses. Big plays and efficiency are not inversely related. Crowder ranked a lowly 114th in separation, but his WRER was held up by top 25 marks in both hands and open field. Based on my analysis at the beginning of the article, if one of those scores is to stay, it will be his poor separation.