Win. Your. League.

Receive 3 Free Downloads More Details

Dynasty, in Theory: Some Statistics are Better than Others

When evaluating player performance, it's only natural to emphasize some statistics over others. Let's make sure that we're emphasizing the right ones.

The Teaser

Hey guys and gals, I’ve been working on two new statistics for judging player quality, and I wanted to see what you thought. Before we get into the statistics, I wanted to show you a list of the top 20 running backs and wide receivers currently in the NFL, based on their average play over the last five seasons. (Note: in order to have a large enough sample to evaluate, a player must have received at least 100 rushes or 100 targets over that span).

Top 20 running backs by statistic #1:

  1. Todd Gurley
  2. Darren Sproles
  3. Jamaal Charles
  4. C.J. Spiller
  5. Adrian Peterson
  6. Justin Forsett
  7. Latavius Murray
  8. Marcel Reece
  9. Lamar Miller
  10. Demarco Murray
  11. Jerrick McKinnon
  12. LeSean McCoy
  13. DeAngelo Williams
  14. Ryan Mathews
  15. Reggie Bush
  16. Joseph Randle
  17. LeGarrette Blount
  18. Jeremy Hill
  19. Marshawn Lynch
  20. Jonas Gray

Top 20 running backs by statistic #2:

  1. Adrian Peterson
  2. Arian Foster
  3. Todd Gurley
  4. Marshawn Lynch
  5. Le’Veon Bell
  6. LeSean McCoy
  7. Doug Martin
  8. Alfred Morris
  9. DeMarco Murray
  10. Matt Forte
  11. Frank Gore
  12. T.J. Yeldon
  13. Eddie Lacy
  14. Jamaal Charles
  15. Chris Johnson
  16. Ryan Mathews
  17. Jeremy Hill
  18. Darren McFadden
  19. Chris Ivory
  20. Stevan Ridley

Top 20 wide receivers by statistic #1:

  1. Kenny Stills
  2. Jordy Nelson
  3. Malcom Floyd
  4. DeSean Jackson
  5. Julio Jones
  6. Calvin Johnson
  7. Odell Beckham Jr
  8. Randall Cobb
  9. Jarius Wright
  10. Marques Colston
  11. Demaryius Thomas
  12. Terrance Williams
  13. Josh Gordon
  14. Victor Cruz
  15. Dez Bryant
  16. Antonio Brown
  17. Doug Baldwin
  18. T.Y. Hilton
  19. Lance Moore
  20. DeAndre Hopkins

Top 20 wide receivers by statistic #2:

  1. Calvin Johnson
  2. Odell Beckham Jr
  3. Brandon Marshall
  4. Julio Jones
  5. Antonio Brown
  6. A.J. Green
  7. Demaryius Thomas
  8. Andre Johnson
  9. Kelvin Benjamin
  10. Allen Robinson
  11. Mike Evans
  12. Wes Welker
  13. Larry Fitzgerald
  14. Vincent Jackson
  15. Josh Gordon
  16. Alshon Jeffery
  17. Keenan Allen
  18. Dez Bryant
  19. Roddy White
  20. Victor Cruz

Go ahead and look at those lists for a little bit. Keep in mind that these lists represent average play over the last five years, so of course someone like Andre Johnson might feel a bit too high while someone like DeAndre Hopkins feels a bit low. Regardless, what do you think about my new advanced player-grading statistics?

And The Reveal

I’m willing to bet that you’re thinking that Statistic #2 looks like a decent-but-not-perfect list, while Statistic #1 looks like I was just pulling names randomly out of the hat, mixing all-time greats with end-of-roster scrubs. That’s certainly what I think when I look at the two sets of lists.

So what were my “advanced stats”? The first stat was merely yards per carry for an RB and yards per target for a wide receiver. The second was carries per game for an RB and targets per game for a wide receiver.

This, to me, is very interesting. We laud players all the time for having a high yards per carry or yards per target average. We say they’re being “efficient” with their opportunities. And when a player struggles in one of those metrics, (such as Le’Veon Bell his rookie year), we view this is a huge negative sign.

On the other hand, we’re almost disdainful of players who excel based on huge carry-per-game or target-per-game averages. We say they’re “volume-dependent”, and we speculate about what might happen when that volume dries up.

Obviously the players who have both volume and efficiency are the gold standard. But if a player isn’t going to have both, is it really better for him to be efficient than high-volume? I’d argue that the list of “volume-dependent” players is much higher-quality than the list of “efficient” players. In fact, let’s take a look at the players who appear exclusively on one list or the other, (but not both).

“Efficient” running backs:
Darren Sproles, C.J. Spiller, Justin Forsett, Latavius Murray, Marcel Reece, Lamar Miller, Jerrick McKinnon, DeAngelo Williams, Reggie Bush, Joseph Randle, LeGarrette Blount, Jonas Gray

“Volume-driven” running backs:
Arian Foster, Le’Veon Bell, Doug Martin, Alfred Morris, Matt Forte, Frank Gore, T.J. Yeldon, Eddie Lacy, Chris Johnson, Darren McFadden, Chris Ivory, Stevan Ridley

“Efficient” wide receivers:
Kenny Stills, Jordy Nelson, Malcom Floyd, DeSean Jackson, Randall Cobb, Jarius Wright, Marques Colston, Terrance Williams, Doug Baldwin, T.Y. Hilton, Lance Moore, DeAndre Hopkins

“Volume-driven” wide receivers:
Brandon Marshall, A.J. Green, Andre Johnson, Kelvin Benjamin, Allen Robinson, Mike Evans, Wes Welker, Larry Fitzgerald, Vincent Jackson, Alshon Jeffery, Keenan Allen, Roddy White

Perhaps it’s just me, but those “volume-dependent” players look markedly better than their low-volume, high-efficiency peers.

Turning Up The Volume

How does a receiver gain a target? To begin with, his team must decide to throw the ball. After that, one of three things must happen.
1. The play must be designed to go to him.
2. The wide receiver must be the most open player the quarterback sees.
3. The wide receiver must be the player most trusted to win his matchup despite being well covered.

All three of those things, you will notice, say something pretty good about the wide receiver who is getting the target. A player like Demaryius Thomas gets a lot of plays designed to go in his direction, (especially quick receiver screens). A player like Wes Welker is able to get open at-will in the middle of the field. A player like A.J. Green is dominant at making contested catches down the field. In all cases, the fact that these players get a ton of targets tells us a lot about how great they are.

Meanwhile, what does it mean if a player like Kenny Stills is absolutely lights-out on every target, but receives just a bare handful a game? Perhaps it means that the offense doesn’t see fit to feature him more. Perhaps it means that he’s only getting the ball when he’s wide open, which is not very common. Perhaps it means that other receivers on his team are more reliable in tough situations.

In the case of Kenny Stills, his lack of targets tells us a lot more about him as a player than his per-target efficiency.

This isn’t to say that volume is the perfect indicator of player skill. Teams that trail a lot tend to pass more, which means more targets to go around. (Teams that trail a lot often have below-average receivers, too, which contributes to them trailing so much.) Players compete against their own teammates for targets, so a player like DeAndre Hopkins who played with a target hog like Andre Johnson can sometimes struggle to get a lot of looks. (Obviously that hasn’t been a problem this year with Johnson gone.)

But this is to say that the fact that a player is getting targets tells us something positive about him as a player. The more targets he’s getting, the more positive information it’s giving us. Targets may not always be directly comparable across teams, but more targets is always more good.

(The process for getting carries doesn’t lend itself to easy illustration as much as the process for getting targets, but for running backs as well, more carries is always more good.)

A Big Problem with Efficiency Stats

When people talk about efficiency stats, they typically mean “per opportunity” efficiency stats, which for running backs means per carry and for wide receivers means per target. But the carry and the target are not the ideal measure of opportunity for these positions. Ryan Mathews has rushed for 409 yards on his 67 carries, giving him a 6.1 yard-per-carry average that ranks second in the NFL. But those 67 carries hardly represent the only opportunity he’s had to run the ball; indeed, Philadelphia’s 225 team rush attempts rank 10th in the NFL, and more than 2/3s of them have gone to someone other than Mathews.

Mathews has played fantastic in his limited role so far, and he’s making a very strong case to see that role expanded. But we have to ask why, if he’s so great, his role was limited in the first place. Is he really that much better than teammate DeMarco Murray, who averages 3.7 yards per carry this season?

Similarly, a wide receiver has an opportunity to gain positive yards every time his quarterback drops back to pass, not merely the times he winds up getting targeted. Consider, if you will, the 2013 Arizona Cardinals. That season was viewed by many as a passing of the baton, as Michael Floyd out-gained Larry Fitzgerald by nearly 100 yards, 1041 to 954. On a per-target basis, the gap was even more impressive; Floyd averaged a sterling 9.21 yards per target, while Fitzgerald averaged an entirely unimpressive 7.07.

The one problem with the passing-of-the-baton story, though? If Michael Floyd was so much better than Larry Fitzgerald in 2013, why did Fitzgerald get 135 targets to Floyd’s 113? Clearly the team felt that Fitzgerald was the better bet because they threw the ball his way 20% more often. And, with the full benefit of hindsight, it seems silly now to suggest that Michael Floyd is a better receiver than Larry Fitzgerald; their respective target total predicted their futures a lot better than their “efficiency”, or even their raw yardage totals.

(An interesting parallel this season is happening in Jacksonville; Allen Hurns' 10.76 yards per target figure is a big driving factor behind is surprise breakout, especially compared to Allen Robinson's good-but-not-amazing 8.84 mark. Robinson, however, is out-targeting Hurns by over a third, with 80 targets to 59 after eight games.)

Anecdote is interesting, but Fitzgerald vs. Floyd is not an isolated example. Study after study find that the “stickiest” metrics, (meaning the metrics that remains most stable from one sample to the next), are all volume metrics. A study by Chase Stuart found that targets per route run was by far the “stickiest” receiver stat over yards per route run and yards per target. Another study by Chase Stuart found that targets per team pass attempt was much stickier than yards per target. Similarly, for running backs, he has shown that rush attempts are several times stickier than yards per carry.

Danny Tuccitto has done a lot of work on his blog, Intentional Rounding, on how long it takes for various statistics to “stabilize” to the point where they represent a 50/50 mix between a player’s “true mean” and random chance. For wide receivers, Tuccitto found that yards per target took 39 games— two and a half years— to finally stabilize. Yards per route run, however, stabilized in just 14 games. And targets per route run? It only took, on average, 7 games before a player’s “targets per route run” reached a point where it was less than 50% random noise.

For fans of running back efficiency, things are especially grim. Tuccitto found that it took 177 games before a running back’s yards-per-carry average was more representative of his own underlying talent level than it was of random chance. That’s more than eleven years. Or, to put it more bluntly, that’s statistician speak for “never”.

I’ve even dabbled my own toes into these waters. After Le’Veon Bell’s rookie year, when many were worried that his low ypc meant he would be the next Trent Richardson, I wrote about the history of high-volume, low-efficiency rookie running backs.

The history, as it turns out, is amazing. Since volume is so stable and efficiency is so random, most of those high-volume rookie turned into high-volume sophomores, had a large increase in per-touch efficiency, and saw their production spike through the roof.

What Should We Take From This?

The question we should be asking is not “what would happen if this hyper-efficient player finally started getting more usage?”, it’s “why isn’t this hyper-efficient player getting more usage?” Sometimes the answer is obvious; Dez Bryant doesn't get more usage because his team passes so rarely. DeAndre Hopkins didn't get more usage because he was sharing the field with Andre Johnson, the greatest player in franchise history. Michael Turner got so few carries in San Diego because he was stuck behind a first-ballot Hall of Famer.

Other times, the answer is less obvious. Why didn't Kenny Stills get more targets in New Orleans? (If early returns in Miami are any indication, it's because he simply wasn't that good; Robert Meachem and Devery Henderson had both previously put up similarly stellar yards-per-target averages in the same role for the Saints.) Why isn't Lamar Miller getting more usage? Are his coaches merely incompetent, or do they know something we don't? It at least has to be a concern.

On the other side of the coin, the question we should be asking is not “What happens to this compiler when the volume dries up?” Instead, the proper question is “If efficiency is prone to such wild swings from game to game and from season to season, what would happen if this heavily-used player finally started getting better efficiency, too?”

The answer, more often than you might think, is "championships".