Playmaker Score 2013

by Vince Verhei
Sometimes, there simply aren't many good wide receivers in a given draft. Take the 2002 class, for example. Eleven wideouts went in the first two rounds of the draft that year, but a decade later we find that only three of those men ever gained more than 1,000 yards receiving in an NFL season. Deion Branch leads that class with a paltry sum of 6,644 yards, an average of less than 50 yards per game over his career.
I bring this up to point out that sometimes even the best wide receivers available don't warrant early draft picks -– and the numbers say that 2013 could be one of those years.
At Football Outsiders, we've been working on our Playmaker system for projecting wide receiver success for several years now, refining it to try to more accurately forecast which receivers will succeed in the NFL, and which are likely to fade away. Playmaker is based purely on empirical data, not scouting info or subjective opinions. The goal is simple: to identify those wide receivers with a track record of success and the athletic gifts necessary for success in the NFL.
Playmaker looks at each player's receiving yards and touchdowns in college, adjusted for the number of passes that player's team threw. There is also a slight bonus for yards per reception. Only a player's best season is considered, rather than career totals, to avoid penalizing players who transferred from junior college or declared early for the draft.
Playmaker also uses some data from the NFL scouting combine and various pro days, checking each player's 40-yard dash time (or, when available, the 10-yard split) and vertical leap.
None of these numbers were chosen arbitrarily. All collegiate receiving numbers and combine data were checked for their ability to project NFL success, and those with the most predictive ability are used in the formula.
Is it accurate? The following table sorts all wide receivers drafted from 2005-09 by their Playmaker Score coming out of college, along with their average receiving yards per season in the NFL:
|
|
It's important to remember that most drafted players don't accomplish much in the NFL, so the standards for success here are awfully low. The key number for Playmaker appears to be 400. Those with Playmaker Scores of 400 or higher have about a 50-50 chance of becoming a legitimate No. 1 NFL wide receiver, while the odds are much shorter below that level, and quickly plummet when scores drop below 200 or so.
Which brings us to this year's class of receivers. The following table takes the 13 receivers in the draft with a grade of 70 or higher from ESPN's scouts, and sorts them by Playmaker. None of them cross the critical 400 threshold (though one comes very close):
Name | School | Grade | Playmaker |
Stedman Bailey | West Virginia | 75 | 399 |
DeAndre Hopkins | Clemson | 85 | 368 |
Terrance Williams | Baylor | 83 | 317 |
Justin Hunter | Tennessee | 89 | 296 |
Robert Woods | USC | 82 | 285 |
Markus Wheaton | Oregon State | 77 | 274 |
Tavarres King | Georgia | 71 | 268 |
Aaron Dobson* | Marshall | 84 | 213 |
Quinton Patton | Louisiana Tech | 86 | 197 |
Tavon Austin | West Virginia | 93 | 190 |
Keenan Allen* | California | 91 | 168 |
Cordarrelle Patterson | Tennessee | 90 | 146 |
Kenny Stills | Oklahoma | 70 | 133 |
*-Vertical leap not available; projected using average vertical leap. |
Why the dearth of quality wideouts? For starters, there's not a physical freak in the class like Julio Jones or Calvin Johnson, a guy whose raw athletic gifts jump off the page. The fastest players in this group in the 40-yard dash (Tavon Austin and Kenny Stills) were both below average in vertical leap, while the best leaper (Justin Hunter) was just average in the 40.
Meanwhile, the best statistical guys in this class seem to come up short athletically. The average 40 time among these receivers was 4.46 seconds, but the guys with the best NCAA numbers (Stedman Bailey, Terrance Williams, DeAndre Hopkins, Robert Woods) were all slower than that.
There's another problem here: we're looking only at receiving data, not rushing or special teams numbers. Cordarrelle Patterson averaged 12.3 yards on 25 rushes for Tennessee last season, and he also scored three touchdowns on the ground and two more on special teams. Tavon Austin had 643 yards and three touchdowns as a rusher for West Virginia, and he too scored twice on returns. None of those numbers mean anything in Playmaker.
Should they? A long touchdown on a kick return or an end-around demonstrates "football speed" and open-field running talent, but it may not tell us much about beating professional cornerbacks. Success in the NFL is primarily a matter of bursting off the line (as measured in our combine numbers), followed by running precise routes and actually catching the ball (which we can measure with NCAA stats). Return touchdowns show lightning-in-a-bottle home-run ability, but they may not show the kind of every-down talent around which you can build an offense.
Patterson, in particular, is a polarizing case. His big-play ability is tantalizing, but he often disappeared from games. He didn't crack the SEC's top 10 in receiving yards, despite playing for a team that had the second-most pass attempts in the conference. He had just one 100-yard game with the Volunteers, and that came against Troy. Against SEC competition, he never gained more than 88 receiving yards, and averaged less than 50 yards per game on a pass-heavy team. For all his gifts, Patterson has done very little as a receiver to show he can play in the NFL.
Austin and Patterson both finish below the 200 level in Playmaker Score. What have other players with similar scores done in the NFL? From 2005-09, there were 62 receivers drafted with Playmaker Scores below 200. The best of those in the NFL has been Steve Johnson, who has gone over 1,000 yards in three straight seasons now. Some other big names in the group include Steve Breaston and Eddie Royal, both of whom were also threats to score on special teams in their collegiate careers. Those players, though, are the exceptions. As a group, these 62 receivers have averaged less than 100 yards per season in the NFL. Patterson and Austin could buck those odds, but those are not the kind of numbers around which NFL general managers should want to base their team's fortunes.
(Ed. Note: Could Playmaker Score be right about Patterson, a consensus first-round pick? Matt Waldman will have a rebuttal on Wednesday. Also, check out Waldman's Futures columns scouting out Tavon Austin, DeAndre Hopkins, and Terrance Williams.
Comments
36 comments, Last at 27 Apr 2013, 12:21pm
#1 by JonFrum // Apr 22, 2013 - 12:08pm
I'm not big on Austin as a first round pick as a receiver, but I suspect he'll have a good NFL career - at least until he gets killed - as long as he's used right. If his coaches get him the ball, like Belichick has done with Aaron Hernandez, he'll be a valuable playmaker. Put him in the slot, in the backfield, throw him screens and get him the ball in space. Add returns, and you may get a lot more out of him than you'd expect based on statistical analysis.
#2 by Mr. X (not verified) // Apr 22, 2013 - 12:29pm
If that chart is sorted by Playmaker Score, should Keenan Allen be ranked behind Tavon Austin?
#4 by Aaron Schatz // Apr 22, 2013 - 1:32pm
Fixed.
#3 by K (not verified) // Apr 22, 2013 - 12:45pm
Table 1 is screaming for a scatter plot instead of coarse binning at arbitrary cut-offs.
#5 by Aaron Schatz // Apr 22, 2013 - 2:00pm
Added above.
#7 by K (not verified) // Apr 22, 2013 - 2:37pm
Thanks!
#6 by johnlimberakis // Apr 22, 2013 - 2:12pm
I wonder what:
1) Justin Swope's score is.
2) What Randall Cobb's score was.
John
#8 by Aaron Brooks G… // Apr 22, 2013 - 3:09pm
1. You mean Ryan Swope?
#9 by Jgr514 // Apr 22, 2013 - 4:31pm
Any chance we can get a sneak peek at the new formula?
#10 by johnlimberakis // Apr 22, 2013 - 4:36pm
Yes,
Ryan Swope.
Sorry... not sure how I got Justin.
#11 by fb29 // Apr 22, 2013 - 4:53pm
Vince,
I was disappointed last year when you didn't specifically address the huge miss regarding predicting the success of Julio Jones when you switched to Playmaker 3.0. This year, you even cite Julio as a physical freak while failing to mention that your prediction for him in the NFL based on Playmaker 2.0 was dead wrong. I assume Julio had to be a big reason for the switch to 3.0.
Could you share Julio's playmaker 3.0 score? Also, have you considered adding a "1st round" adjustor similar to LCF? Maybe a top 10 adjustor?
Thanks
#14 by Aaron Schatz // Apr 22, 2013 - 5:59pm
Basically, Julio Jones' Playmaker Score climbed substantially when Playmaker v3.0 added the use of combine data to go along with college production. His score in the current system is 381, which still does not properly indicate how good he would become but is still better than 85 percent of drafted wide receivers over the past eight years.
#15 by Vincent Verhei // Apr 22, 2013 - 6:18pm
Here are the revised PM scores for all WRs drafted in the first three rounds in 2011 (the Julio Jones class). It's better on the whole, although Jonathan Baldwin comes out way too high and AJ Green embarrassingly low, but you can't win 'em all.
Name PLAYMAKER
Jonathan Baldwin 464
Torrey Smith 448
Julio Jones 381
Leonard Hankerson 328
Titus Young 321
Greg Little 236
A.J. Green 227
Vincent Brown 224
Austin Pettis 171
Jerrel Jernigan 168
Randall Cobb 136
Jones' score is so low because his touchdown total was miniscule -- never more than 7 in a season. That's partly because those Alabama teams were running for about 30 touchdowns each season, but his percentage of team touchdowns, compared to other top prospects, was still very low.
Green, I can't really explain. His combine scores and NCAA numbers were all mediocre.
#18 by justanothersteve // Apr 22, 2013 - 8:13pm
I think Cobb's score may be another anomalous outlier. With Jennings and Driver gone, there's a very good chance he'll be averaging close to 1000 yards/year for his career in just a couple years.
#21 by DisplacedPackerFan // Apr 23, 2013 - 12:05am
I'm not sure the lack of Driver or Jennings matters. He hit 954 yards in just 15 games played last year, so his 63.6 yards per game is already ahead of the 62.5 that you need to be a 1000 yard receiver. Of course the Nelson/Jones/Cobb receiver set that we expect to see a lot next year, was the most common last year too due to injuries.
#22 by justanothersteve // Apr 23, 2013 - 7:51am
The lack of Driver or Jennings means there won't be anyone coming back who might reduce his yardage. I think even with Jones, Nelson, and Finley in the lineup, Cobb will be averaging close to 1200 yds/season for the next few seasons. Finley is probably gone after this year, but I expect someone else to step up (and TT will probably draft a TE on day 1 or 2 of the draft).
#20 by Insancipitory // Apr 22, 2013 - 11:06pm
I just remember watching Jon Baldwin a few times last year and thinking I really liked the way he'd would play occasionally very good corners. At somepoint I just wonder "Maybe he's not the problem."
#25 by fb29 // Apr 23, 2013 - 4:23pm
This is great, thanks.
When you put this into the appendix of FOA 2013, would you mind throwing the rest of the columns in the table that are used for playmaker score?
#12 by peterplaysbass // Apr 22, 2013 - 4:59pm
This backs up all of the mocks I've seen with Minnesota taking Hopkins in the first round. I've been hoping for LB & DT in the first round, but I've been hearing scenarios where Te'o and/or Ogletree will be available in the 2nd, freeing the Vikings up to look at DT / WR (Hopkins!) in round 1.
#13 by Sifter // Apr 22, 2013 - 5:48pm
Seems like Tavarres King is slipping down the Big Boards eg. 149 at CBS/NFL draft scout. One guy I'd like to see scored is Da'Rick Rogers, I think he'll be picked in rounds 2-4. Otherwise, good stuff! Always interesting to get some extra statistical info instead of the usual airy-fairy draft talk about upside/explosiveness/intangibles blah blah blah
#16 by KJG520 (not verified) // Apr 22, 2013 - 6:37pm
I have a concern about the graph used in this post. The standard typically is that the independent variable is displayed on the x (horizontal) axis and the dependent variable on the y (vertical) axis. Thus, the playmaker score should be put on the bottom and receiving yards on the side in order for the causation relationship to make sense.
#19 by Noso Opforu (not verified) // Apr 22, 2013 - 9:21pm
Rotate your head 90% to the side and look at your computer screen in a mirror. Then the causation will make sense to you.
#23 by Dean // Apr 23, 2013 - 7:57am
Did you mean 90°?
#24 by Aaron Schatz // Apr 23, 2013 - 10:30am
We fixed this.
#17 by johnlimberakis // Apr 22, 2013 - 7:40pm
Yes... I'd love to see Da'Rick Rogers and Ryan Swope scored here.
#26 by Vincent Verhei // Apr 23, 2013 - 5:27pm
Because I have gotten many, many requests for it, Ryan Swope's Playmaker is 270. He has the second-best Combine numbers in the class (behind Justin Hunter), but his NCAA numbers, especially TDs, were nothing special given how often Texas A&M threw the ball.
#27 by Karl Cuba // Apr 23, 2013 - 6:20pm
Is there any control over how many receivers are on the field when they throw? A receiver for a team that uses 21 personnel versus a receiver on a team that uses 10 personnel could expect very different results even if they threw the ball a similar amount.
#28 by Vincent Verhei // Apr 23, 2013 - 6:29pm
No there is not. I'm intrigued to hear a practical idea for how we could do that.
#29 by Karl Cuba // Apr 23, 2013 - 6:43pm
No idea, I don't read your (or anyone else's) college stuff and so I have no clue whether or not there is anyone tracking that sort of thing in the manner FO does for NFL teams.
If you did have formation tendencies then a crude fix would be to multiply them by pass attempts, or you could try separating teams by spread option, pro-style and wishbone-types and see if that worked.
#30 by Vincent Verhei // Apr 23, 2013 - 7:33pm
That could work, but I don't think any of that information is available. As far as manually sorting teams by offense, there might be some value in it, but I am not the man to do it.
I apologize if my response came off as snippy, but this project, for years now, has been a never-ending pattern of "I spent all weekend working on this" followed by "here is what you should have done instead, do that next weekend," ad nauseam.
#31 by fb29 // Apr 24, 2013 - 2:40am
vince next weekend mow my lawn
#32 by Vincent Verhei // Apr 24, 2013 - 12:44pm
Depending on the weather and size of lawn, I just might do that.
I should also add that finding some of this college data can be difficult. It's gotten much better in the last year or two, but when I first started doing this I had to go through ESPN's gamelogs and count games manually for each guy just to get the number of games played. So if you think there's advanced data out there on formations and personnel groups, for every team, I'm afraid you'll be disappointed.
#33 by Calvin (not verified) // Apr 25, 2013 - 9:15pm
As someone who deals with statistics on a daily basis,that R^2 means that this is not the strongest model. Ranges between 0 and 1, with high being better.
#35 by otros // Apr 27, 2013 - 1:23am
As someone who deals with statistic on a daily basis, you should know that R¨2 is piss poor way to judge a model fit
#36 by Aaron Schatz // Apr 27, 2013 - 12:21pm
In NFL analysis, you have to accept lower correlations than you would accept with other statistical analysis.
#34 by Tanner (not verified) // Apr 26, 2013 - 10:22pm
Marquise Goodwin's score?