RotoViz vs. The Machine: Pitting Staff vs. App Wide Receiver Projections

RotoViz has a great new Screener App, that helps you do some sophisticated analysis and regression work in just a few clicks. Here’s an exercise I did, using the Screener to create a wide receiver projection model, which I then compared to our composite staff projections. There are some interesting takeaways.

First, here’s how I set up the model.


Just as I did with running backs, I’m using per-game numbers to give better treatment to players who were good when healthy, but may have missed time. I’m also using just the 2010 – 2015 seasons. Here are the results.


Again, just like with RBs, I tried various combinations of explanatory variables before settling on these three as having the best mix of predictiveness without overfitting or multicollinearity. Age had a small but significant negative value, indicating that younger is slightly better than older.

You can peruse the table below to see which players our staff composite projections prefer more or less than this naive statistical model; I’ve added some quick thoughts below.

121Antonio Brown23.918.8-5.1
231Odell Beckham Jr.21.318.36-2.94
31-2Julio Jones23.218.83-4.37
4106A.J. Green17.214.52-2.68
54-1DeAndre Hopkins20.617.38-3.22
65751Dez Bryant9.98.8-1.1
76-1Brandon Marshall21.416.54-4.86
87-1Keenan Allen20.416.06-4.34
95-4Allen Robinson1917.36-1.64
10177Demaryius Thomas1713.64-3.36
11121Brandin Cooks15.914.17-1.73
13218Amari Cooper13.513.15-0.35
14162Mike Evans14.213.86-0.34
158-7Sammy Watkins16.815.73-1.07
164428Golden Tate13.29.96-3.24
172912Ty Hilton13.212.2-1
18224Jeremy Maclin16.313.06-3.24
199-10Alshon Jeffery17.615.21-2.39
204020Randall Cobb12.710.44-2.26
21265Jarvis Landry16.812.64-4.16
226442Torrey Smith7.78.20.5
2311-12Julian Edelman19.414.39-5.01
24273Jordan Matthews14.512.5-2
2523-2Larry Fitzgerald17.812.97-4.83
2615-11Doug Baldwin16.813.86-2.94
2718-9Eric Decker1713.51-3.49
284214Donte Moncrief10.810.23-0.57
293910DeSean Jackson11.910.59-1.31
30355Michael Floyd11.510.87-0.63
3128-3DeVante Parker11.712.240.54
3224-8Emmanuel Sanders15.212.91-2.29
3325-8John Brown1412.72-1.28
3430-4Willie Snead12.511.84-0.66
354914Marvin Jones10.99.59-1.31
3631-5Michael Crabtree14.411.39-3.01
3814-24Allen Hurns15.113.88-1.22
3933-6Stefon Diggs11.511.06-0.44
415312Dorial Green-Beckham7.99.021.12
4313-30Steve Smith18.713.99-4.71
44506Tyler Lockett9.79.51-0.19
4737-10Kamar Aiken12.510.79-1.71
488234Tavon Austin12.37.46-4.84
50522Vincent Jackson10.59.12-1.38
5234-18Travis Benjamin12.210.97-1.23
53596Kendall Wright9.78.5-1.2
5747-10Markus Wheaton9.39.620.32
5851-7Pierre Garcon11.69.32-2.28
5948-11Terrance Williams9.69.60
6032-28Rishard Matthews12.111.32-0.78
6010848Mike Wallace6.66.04-0.56
618928Chris Hogan7.27.17-0.03
63663Kenny Britt7.68.170.57
64706Devin Funchess6.881.2
6660-6Robert Woods8.68.49-0.11
6845-23Steve Johnson12.59.77-2.73
6954-15Jermaine Kearse9.28.94-0.26
7068-2Jamison Crowder8.28.09-0.11
7110938Nelson Agholor4.46.041.64
7316693Sammie Coates2.14.342.24
7515681Jeff Janis24.592.59
76793Brandon LaFell8.17.59-0.51
7736-41Ted Ginn13.110.8-2.3
7871-7Davante Adams8.17.95-0.15
8238-44Rueben Randle11.610.79-0.81
8367-16Danny Amendola10.88.1-2.7
8577-8J.J. Nelson5.97.711.81
9019-71Calvin Johnson16.513.33-3.17
9020-70Martavis Bryant15.713.23-2.47
9041-49Tyrell Williams8.510.331.83
9043-47James Jones11.610.09-1.51
9055-35Brice Butler7.68.911.31
9056-34Anquan Boldin12.38.85-3.45
9058-32Nate Washington10.58.66-1.84
9061-29Marquess Wilson7.38.471.17
9062-28Jeremy Butler8.48.31-0.09
9063-27Percy Harvin108.3-1.7
9065-25Cecil Shorts10.28.17-2.03
9069-21Dontrelle Inman8.58.09-0.41
9072-18Brian Hartline9.27.94-1.26
9073-17Corey Brown7.47.920.52
9074-16Ty Montgomery8.47.89-0.51
9075-15Paul Richardson57.812.81
9076-14Rashad Ross6.57.81.3
9078-12Darius Jennings8.67.69-0.91
9080-10Cole Beasley8.57.59-0.91
9081-9Seth Roberts6.97.510.61
9083-7Jeremy Kerley8.67.39-1.21
9084-6Deonte Thompson57.332.33
9085-5Marques Colston9.37.32-1.98
9086-4Greg Salas7.17.220.12
9087-3Kenny Stills5.97.221.32
9088-2Albert Wilson6.87.210.41
9946-53Leonard Hankerson119.62-1.38

These Screener projections, like those for RBs, exhibit some reversion to the mean, in that the highest projected players are expected to score less, while the lowest projected players are expected to score more. It’s also the case that the model doesn’t know about team or coaching changes, so many of the differences in rank can be discounted for that reason. One interesting thing is that the r^2 between staff and screener WR projections is higher than the r^ between staff and screener RB projections. Perhaps, and this makes sense, we’re more confident projecting WRs. On to the commentary.

  • Torrey Smith has been a favorite buy low candidate of ours for quite some time. The model doesn’t know he has a new coach, but the model does know that his per-game production was not good last year. We’ve got him projected as WR22, vs. WR64 for the model. I definitely think he’ll end up much closer to the top end of that range, but it’s a reminder that he does have a very low floor. I’ll still be taking shots at acquiring him, but that downside should get factored into your considerations.
  • Neither the staff rankings nor the Screener model like Nelson Agholor. But the model has him 38 spots lower than we do. Like Smith, Agholor has a new coaching staff. But unlike Smith, Agholor’s Eagles are expected to trend towards a slower-paced, run-heavy offense. I’d have no hesitation unloading Agholor, but Brian Malone still likes him.
  • More pertinent is Golden Tate, whose Screener rank (44) comes in well below the staff projection (16). Our staff projection is basically 2014 Tate: 145 targets, 97 receptions, 1140 yards, 7 TDs.1 With Calvin Johnson out of the picture, maybe that makes sense. On the other hand, most of his 2014 was accumulated with Johnson playing. So maybe the loss of Johnson hurts the offense so much that there won’t be enough drives for Tate to return value. Also, Tate’s yards per target was near a career low last year; positive regression there would boost his yardage quite a bit. I guess my point here is the same as with Torrey Smith. Tate’s ceiling (he was WR12 in 2014 on the same number of targets we expect him to get this year) is tantalizing, but it’s also true we don’t have a really great idea of how this offense will look, so beware of his floor.
  • Basically the same thing can be said for Randall Cobb (20th in staff projections, 40th in Screener). Maybe a return to health, the return of Jordy Nelson, and a better schedule makes all the difference. But 2015 did happen, and it wasn’t pretty. I don’t think it’ll happen again, but I don’t know that it won’t. Smith, Tate, and Cobb are three WRs I’ll be happy to have on my teams, but only at the lower ends of their valuations.
  • Dorial Green-Beckham is close to WR3 value in our staff projections, but much lower overall. I’ll side with the Screener here. He wasn’t that good last year, and Rishard Matthews offers significant competition.
  • Amari Cooper also comes in lower in the Screener model. Rich Hribar made some great points about Cooper’s outlook on a recent RotoViz Radio episode.
  • In Denver, the model likes Emmanuel Sanders more than we do, and likes Demaryius Thomas less.
  • Our staff projections have Doug Baldwin fairly valued at WR26. The Screener has him all the way up at WR15. I know, I know, “last year was a fluke!” But the Screener is still projecting him to finish lower than his WR10 finish last year. Unlike some of the players above where the Screener is highlighting potential downside, here it seems to be highlighting Baldwin’s upside. Even if he just splits the difference between the Staff and Screener projections, he’s a great value.
  • Julian Edelman is another intriguing player. How he fares without Tom Brady, and whether or not he’s fully healthy, are legitimate questions. But the Screener pegs him as WR11, well above our staff projection. I’m happy to take a shot on him at his current ADP given that upside.
  • Travis Benjamin is another player the Screener loves (34), relative to our staff projections (52). He put up similar numbers to Steve Johnson last year, except he did it in Cleveland, with a lesser cast of quarterbacks. If you buy the argument that Benjamin can be better with a better QB in Philip Rivers, then he should be better than Johnson this year. Certainly San Diego is paying him to be better than Johnson, and if he gets the WR2 role to himself, he could have a lot of upside.
  • I mentioned Rishard Matthews above, but the Screener also really likes Allen Hurns, as do I.
  • The Screener spots Steve Smith a rank of WR13. We know he was on a fantastic pace last year before his injury, so from that point of view that makes sense. There are legitimate reasons to be concerned about Smith in 2016, but at a current ADP of WR51, I’ll take a shot at that upside all day long.
  • Ted Ginn is another slam dunk pick at his current ADP. The Screener remembers that he was a WR3 last year, and even if we all want Devin Funchess to emerge, and expect Kelvin Benjamin to have a big role, those things aren’t guaranteed to happen. For the price of WR67, the chance of Ginn reprising his 2015 is a worthwhile gamble. Is it likely to happen? Maybe not. But I think it’s an asymmetric bet worth making.

Subscribe to the best.

  1. In 2014, Tate posted a 143/99/1331/4 line.  (back)
By Charles Kleinheksel | @Spidr2ybanana | Archive

No Comment

Leave a reply