Writers of Pro Football Prospectus 2008

Most Recent FO Features

JohnsonCal07.jpg

» Catch Radius: Best of the NFC

Part I of our catch radius season finale spotlights the NFC kings of double coverage (Calvin Johnson), the sideline (Jordy Nelson), the drag route (DeSean Jackson) and the red zone (Dez Bryant).

28 Jul 2010

Adjusting Plus-Minus For Team, Part II

by Bill Barnwell

If you haven't already, please read the previous entry in this series on adjusting our plus-minus statistic for a player's team before going through this piece.

What makes this concept of team-adjusted plus-minus so interesting is the promise of producing a measure of hands that is universal across schemes, one that accounts for the benefits of playing with Peyton Manning and the hindrances of catching passes from Derek Anderson.

Before getting into the second part of the research, I wanted to congratulate reader "Peregrine" on successfully guessing that the player since 2006 with the biggest gap between his per-play plus-minus and his team's per-play plus-minus was Packers halfback Noah Herron. In 2006, Herron caught 28 of the 30 qualifying targets thrown to him, producing a 93 percent catch rate and a plus-minus of 4.6 catches above expectation. He was the only Packer that year to be thrown a single pass and not end up with a negative expectation, thanks to Brett Favre's 56 percent completion percentage. The rest of the team had a -47.9 plus-minus on 529 targets, yielding a per-play plus-minus of -.09. Since Herron was at .15, the resulting difference was .24, the highest-such figure over the past four seasons with a minimum of 30 targets.

People asked about the identity of the Darkside Herron in the comment thread. The worst running back with team adjustments built in would be from that same year: Larry Johnson of the Chiefs, who had a -8.7 plus-minus on 61 targets. He had a per-play plus-minus of -0.14, while the rest of the Chiefs were at 0.02. The difference rounds to -0.17, which is the biggest difference of any back.

If we're expanding the scope to players of any stripe, some familiar favorites come into play. It's no surprise that Darrius Heyward-Bey's 2009 season and Ashley Lelie's 2008 campaign are at the top of the charts; by this metric, DHB's season actually figures as the fractionally worse endeavor on a per-play basis. Although Lelie had a worse per-play plus-minus (-.29 to Heyward-Bey's -.28), the absence of JaMarcus Russell for a fair chunk of 2009 means that the baseline for DHB's season (-.04) is slightly better than Lelie's (-.05). The difference is three one-hundreths of a catch per play. Just to mix it up, slightly behind them is the 2008 season by Marty Booker of the Bears, who had a whopping -9.6 plus-minus on 41 targets, which eventually yielded a team-adjusted per-play plus-minus (TAPPPM?) of -.23 catches.

I'd be happy to field questions about the best and worst players over the past four years in the comment thread, but I think it's more relevant here to talk about some particular teams and situations from 2009, and what plus-minus suggests about the performance of both quarterbacks and receivers going forward.

Atlanta

I didn't produce figures that split out performance by a given quarterback, which may color the numbers of the receivers here, since Chris Redman threw 21 percent of the passes by the Falcons a year ago. On the other hand, Redman completed 56 percent of his passes and Matt Ryan was only at 56.3 percent, so it wasn't as if there was a significant difference in accuracy between the two.

By almost any measure, Roddy White didn't do his best work a year ago. This metric goes a step further and says that he was the worst in the league among guys with 100 targets or more. He had a -9.1 plus-minus on 150 targets, and after adjusting that figure for the context of the team (a -.06 per-play figure versus .02 for the rest of the Falcons), his cumulative team-adjusted total of -12.7 catches below expectation was the worst in football.

Now, here's where we need to clarify using this statistic in context. Does that figure mean White was an awful receiver a year ago? Absolutely not. We know from watching games that the team, especially with Redman at quarterback, was desperate to force throws to White. The 2009 figure is out of character with his career team-adjusted plus-minus (-.02 in 2006, 7.0 in 2007, and -2.5 in 2008). It was a down year for White, no question, but it has to be considered in context. His 52 percent catch rate is all but sure to rise in 2010, and it should drive a healthy improvement in White's statistics.

Baltimore

Here's a quick case where a player's skills might be overrated. Ray Rice's primary role in the Ravens' passing offense is to serve as the safety valve/dumpoff option/quick read, and he catches most of his passes at or near the line of scrimmage. As a result, although he caught 79 of his 97 qualifying targets, he only had a plus-minus of 2.7, and once you adjust for the team context, that figure falls to 0.14. In other words, he caught passes at exactly the rate an average player might in this offense.

Cincinnati

Fluke alert: Brian Leonard caught 30 of the 34 targets thrown in his direction, putting up a per-play plus-minus of .11 when the rest of the team was at -.03. That's not going to happen again, so expect his catch rate to decline this year.

Dallas

Just behind White in the annals of poor catch rates is Roy Williams, who went 38-for-80 (-10.3 plus-minus) on a team that was above-average on passes not thrown to him. He finished with a team-adjusted plus-minus of -12.4 catches below expectation, the second year in a row he's been well below expectations.

Detroit

Here's one where I think the team adjustments make sense. Calvin Johnson has a raw plus-minus of -5.9 on 128 targets; we can't adjust that figure for Matthew Stafford lobbing the ball at him in triple coverage, but we can adjust it for the misthrows Stafford makes elsewhere. Johnson's per-play plus-minus is -.05, but the rest of the team is at -.07; adjusting the figure for team has Johnson at 3.3 catches above expectation.

Green Bay

Another reason to favor Jordy Nelson over James Jones in the battle for the Packers' third receiver job: Nelson had a team-adjusted plus-minus of 4.0 catches above expectation on 28 targets, while Jones was at -6.6 on 56 targets. This camp battle was something we debated a lot in the process of writing the book and putting together the projected role variables in KUBIAK. Nelson's almost certainly the better receiver already, and I firmly believe that Nelson will be playing the larger role this season.

Houston

And here's one where we're not modeling the relationship between receivers properly. Andre Johnson has a plus-minus of 4.2 catches above expectation on 156 targets. That's fine. However, Johnson plays on a team where everyone else catches a high percentage of the passes thrown to them, and once we adjust the figure for the team context, Johnson is actually 8.1 catches below expectation.

Of course, that's not realistic. One of the reasons why those guys are so open to make those catches is that Johnson draws attention away from them. On the other hand, his numbers haven't consistently been that bad -- his team-adjusted plus-minus figures go from -13.0 in 2006 to 7.2 in 2007 and 5.0 in 2008. Last year's drop mirrors a decline in his catch rate, which fell to 59 percent after being at 68 percent in 2008 and 70 percent in 2007. This is one where Texans fans would be able to help: Did Johnson struggle to catch the ball in 2009? His catch rate didn't drop significantly in games without Owen Daniels.

Jacksonville

Another reason to like Mike Thomas: He had a team-adjusted plus-minus of 6.6 on 59 targets as a rookie last year. Second wideout in the Jaguars' offense isn't a particularly glamorous role, but there's a lot of potential there as a late-round guy in fantasy drafts.

Kansas City

Dwayne Bowe raw plus-minus figures: -0.9, -9.4, -2.4. Dwayne Bowe team-adjusted plus-minus figures: 4.1, -4.6, 2.5. I think the latter is a better representation of Bowe's abilities, no?

Miami

Greg Camarillo had a great fluky season, with a plus-minus of 8.9 catches above expectation on 68 targets. His figure of 19.6 catches above expectation per 150 targets was fifth-best in the league. I don't think he's going to catch 69 percent of the passes thrown to him again this year.

Miami/New England

Since one of the reasons this statistic was developed was to try and track players like Wes Welker, I'll mention him here. In 2006, Welker had a plus-minus of 3.6 on 93 targets; after adjusting that figure for the miserable Dolphins offense of that year, Welker's plus-minus more than doubled, to 8.8 catches above expectation. He had a difference between his per-play plus-minus and the rest of the team's per-play plus-minus of .09; good, but not extraordinary. Had we done this statistic in 2006, I don't think people would have recognized Welker as a star waiting to break out, just a good player in a bad offense.

On the other hand, that was the year of Chris Chambers's remarkably terrible season; he had 59 catches on 134 qualifying targets, which yielded a staggering plus-minus of -21.4. The rest of the team was just about average, so his team-adjusted plus-minus was also -21.4. That is the worst single-season team-adjusted plus-minus of any receiver over the four-year stretch by a staggering margin; the next-closest receiver is Torry Holt, who put up a plus-minus of -14.2 for the Rams that year.

New York Giants

This is one that really interested me. The player who has the largest team-adjusted plus-minus of any player over the past four years? It's the Giants' Steve Smith. Last year, Smith had a raw plus-minus of 13.8 on 149 targets, which is a very impressive figure. He had a .09 per-play plus-minus, but the rest of the Giants were at -.02! None of the other players with big target totals were very good at catching the ball -- Mario Manningham was almost exactly average (-.1 -- total -- over 96 targets), Hakeem Nicks was only slightly above-average (2.9 on 69 targets), and Kevin Boss was slightly below-average (-0.3 total on 64 targets). Of course, Ahmad Bradshaw and Brandon Jacobs were both terrible.

As a result, Smith had a team-adjusted plus-minus of 16.7, while everyone else with more than 10 targets besides Nicks ended up with a figure below zero. That raises a question to me: Is the "improvement" in Eli Manning's statistical record really just the effect of replacing Plaxico Burress as his top receiver with Smith? I don't know if the statistical record supports it -- Burress had a -0.3 plus-minus on 117 targets in 2006 and was at -3.4 on 138 targets in 2007 -- but I think there's probably at least some cause there.

Oakland

Because the wideouts were so bad, Zach Miller ends up looking like one of the great tight ends in league history. He had a 5.8 plus-minus on 90 targets when the rest of the team combined for a -32.7 plus-minus on 348 targets. The resulting difference pushes Miller up to 14.2 catches above expectation, a figure topped by only Jason Witten and one other tight end, both in 2009. (I'll get to that tight end in a minute.)

San Diego

In 2008, Malcom Floyd had one of the great outlier seasons, with 27 catches on 35 targets and a staggering 8.2 catches above expectation; scaled to 150 targets, his figure of 35.1 catches above expectation was the best of any receiver with 30 targets or more over the four-year stretch.

In 2009? He dropped back to earth. He put up a reasonable plus-minus of 5.2 on 73 targets, but once you strip out the team context, his plus-minus drops to -0.6. Not everyone will bounce back towards the mean the way Floyd did, but staggering outliers in his vision just have to. The aforementioned Camarillo and Miller are two candidates that stand out as almost sure to regress, and the third is the tight end that we promised we'd get to in a minute.

St. Louis

On the Welker hunt? Danny Amendola had a raw plus-minus of 0.1 catches above expectation on 61 targets, but adjust that for the suck of the St. Louis offense, and he was actually at 4.0 catches above expectation.

Tampa Bay

Minute's up. Kellen Winslow has lived a rough pro existence; in addition to the injuries, he's yet to have a season where his teammates have put up an per-play plus-minus greater than zero, while he's had a positive figure in every season but one (2007, when he was at -.02). Last year, as an example, he put up a 3.5 plus-minus on 117 targets, while the rest of his teammates stumbled through their 367 targets with a combined plus-minus of -38.1. The resulting difference between his per-play plus-minus and the rest of his team's was .13, and his team-adjusted plus-minus of 15.7 catches above expectation was the best of any tight end over the past four years.

It's not going to happen again. Even if Winslow stays healthy and the team goes with two rookie receivers, not even Winslow is likely to sustain a plus-minus rate that represents 20 percent of his actual catches (15.7 versus his 78 receptions, which is 20.1 percent). He's going to drop off, and he's probably overvalued in fantasy drafts at this time.

We aren't able to update plus-minus during the season because it involves charting data, but we'll keep working on improving the methodology for next season, and report back on what we saw in the 2010 season in Football Outsiders Almanac 2011.

But now that we've focused on the catch itself, what's next should be obvious: Next week, we'll be taking a look at how effective receivers are once they've got the ball in their hands.

Posted by: Bill Barnwell on 28 Jul 2010

31 comments, Last at 25 Jul 2011, 8:04pm by nath

Comments

1
by Jerry :: Wed, 07/28/2010 - 5:32pm

On the other hand, Redman completed 56 percent of his passes and Matt Ryan was only at 56.3 percent, so it wasn't as if there was a significant difference in accuracy between the two.

Bill,

I realize that you may not have the data for quarterbacks, but isn't the whole point of this exercise that 56% completion rates can be very different depending on what passes are thrown?

14
by Bill Barnwell :: Thu, 07/29/2010 - 6:08am

Their completion percentage and YPA were almost identical. They weren't really playing very differently, I promise!

2
by Jon :: Wed, 07/28/2010 - 5:57pm

Leonard isn't a fluke. Pretty much all he does is catch passes out of the backfield, he's always been really good at that.

5
by Nathan :: Wed, 07/28/2010 - 10:51pm

I'm all about the pass catching fullback.

10
by Hari-Kiri Bengals Fan (not verified) :: Wed, 07/28/2010 - 11:52pm

Leonard isn't actually a fullback. He was our #2 running back, with Bernard Scott as #2.

Bratkowski's offense uses a weird third-down back, who mainly blocks and goes out for passes--it used to be Kenny Watson; now, it's Leonard.

11
by Nathan :: Thu, 07/29/2010 - 12:16am

Yeah, I actually googled it cause I wasn't sure and saw lots of articles saying "Cinci trades for fullback Leonard". So I went with it. You know what I mean though, those hybrid fb/rbs like Leonard, Peyton Hillis, Le'Ron McLain, etc.

12
by tuluse :: Thu, 07/29/2010 - 1:45am

I think that's actually a pretty normal 3rd down back.

Is that what Faulk has been doing for years? Well plus a draw every now and then.

15
by Bill Barnwell :: Thu, 07/29/2010 - 6:10am

I promise you, having a plus-minus that high is a fluke, even if he was the best receiving back in football. And he's not.

3
by kpickles :: Wed, 07/28/2010 - 6:45pm

good work guys

4
by Fan in Exile :: Wed, 07/28/2010 - 7:08pm

I just wanted to chime in as well, and say this is awesome work. Any chance for a breakdown on the Bronco's recievers? I'd love to know what we really lost with Brandon Marshall. I mean besides the stuff I could see of course.

7
by Nathan :: Wed, 07/28/2010 - 10:53pm

I'd also be curious about that, since FO's stats seem to treat Marshall pretty harsly. I'd also like to see the change from Cutler to Orton. Cutler chucked it up to Marshall A LOT while Orton seemed a little more judicious.

17
by Bill Barnwell :: Thu, 07/29/2010 - 6:22am

OK, here are the Broncos. Player Name / targets / raw plus-minus / team-adjusted plus minus.

MARSHALL 147 / 5.4 / 9.3
GAFFNEY 84 / 0.1 / 0.9
ROYAL 68 / -6.8 / -7.1
SCHEFFLER 45 / 1.2 / 1.7
GRAHAM 37 / 0.1 / 0.4
MORENO 37 / -1.5 / -1.3
BUCKHALTER 36 / 2.3 / 2.8
STOKLEY 28 / 0.5 / 0.8
LLOYD 18 / -2.1 / -2.1

20
by Noah of Arkadia :: Thu, 07/29/2010 - 10:45am

In your face, DVOA!

25
by Dan :: Thu, 07/29/2010 - 3:14pm

deleted

22
by Scott C :: Thu, 07/29/2010 - 10:59am

Why would Royal's team-adjusted number go down compared to raw, while others went up?

24
by BigCheese :: Thu, 07/29/2010 - 2:38pm

Looking at the numbers, I would assume it's because he was the Broncos' worst reciever BY FAR, and thus the only one below the team average (Lloyd seems to be right AT the average), and by a significant margin as well, so when compared to the rest of hte team (even Lloyd) he looks that much worse, while everyone else looks that much better.

- Alvaro

26
by Dan :: Thu, 07/29/2010 - 3:14pm

Because the team adjustment is based on the rest of the team. The rest of the team (excluding Royal) was above average, so Royal gets adjusted down. Marhsall gets adjusted up because the rest of the team (excluding Marshall) was well below average. Anyone else gets adjusted slightly up because the rest of the team (excluding him) was slightly below average.

31
by nath :: Mon, 07/25/2011 - 8:04pm

Just read this after you linked it in your Grantland article, Bill. Just wanted to pile on after the fact that it's not too surprising that McDaniels dumped the only two receivers with significant positive +/-.

6
by Nathan :: Wed, 07/28/2010 - 10:51pm

Can we see the full numbers or was this done for ESPN? The link to the original is borked so I can't check. I'd be curious to see how Jerome Harrison fared last year.

16
by Bill Barnwell :: Thu, 07/29/2010 - 6:11am

Harrison's raw plus-minus was -1.83 on 45 targets. That's -.04 per play, but the rest of the Browns were -.09, so Harrison rates out as 2.4 catches above team-adjusted expectation.

8
by Chase (not verified) :: Wed, 07/28/2010 - 11:01pm

The link for Part I is incorrect: should be this.

Great stuff as always.

9
by Dales :: Wed, 07/28/2010 - 11:01pm

"Is the "improvement" in Eli Manning's statistical record really just the effect of replacing Plaxico Burress as his top receiver with Smith?"

If his improvement was all of 13 catches, then perhaps. If it was more than that, then no, clearly.

30
by BigDerf :: Wed, 08/04/2010 - 9:08pm

It could really work either way. I thought Manning improved most as a quarterback during the superbowl year... which was also Smith's rookie year. 2008 he was fine until Plaxgate and then he "broke out" last year... but that really did all start in 07 when Smith arrived. But at the same time Smith is one of the most targeted players in the league which helps his stats and is because he is Manning's favorite receiver. Almost every quarterback has one. Eli has the football intelligence and arm strength to see the field and make all the throws but not the accuracy to hit all the tighter windows. When Smith arrived he gave Manning a serious check-down option who reliably gets open in space on third downs and that made him Manning's "favorite" receiver because when the first option isn't there he goes right to Steve Smith. Both guy's skill sets treated each other's well and thats why they have both been so successful.

13
by Mr Shush :: Thu, 07/29/2010 - 5:49am

As a Texans fan who watched almost every game, I don't remember noticing any particular fall-off in Johnson's catching. I do remember thinking that he was running more deep routes, and the jump in his yards per catch (from 13.7 to a career high of 15.5) would seem to bear that out. Note too that concurrently, Kevin Walter's YPC fell from 15.0 to 11.5, while slot possession man David Anderson's fell from 12.7 to 9.7. Essentially, I think that in 2009, Walter and Andre Davis "lost" deep targets to Johnson and Jacoby Jones, while Johnson and the tight end group (after Daniels went down) "lost" possession type targets to Anderson. I know Plus-Minus seeks to adjust for length of target, but no adjustment of that sort is ever perfect, and I think what we're seeing here is a combination of change of usage that the system for whatever reason fails to fully account for and plain old random variation.

18
by ChaosOnion (not verified) :: Thu, 07/29/2010 - 8:44am

The "previous entry" link in the first sentence above is broken. Well, it is broken when I try to click on it.

http://www.footballoutsiders.com/fo-espn-feature-columns/2010/adjusting-...

19
by Brandon (not verified) :: Thu, 07/29/2010 - 9:42am
23
by ChaosOnion (not verified) :: Thu, 07/29/2010 - 11:38am

Hah! I just had a typo like that.
> i vs I
> minSize vs. minSide

21
by George (not verified) :: Thu, 07/29/2010 - 10:55am

How does plus-minus take into account a receiver facing a #1 CB vs. a nickel back? Thanks, great work.

27
by aaa (not verified) :: Thu, 07/29/2010 - 5:12pm

Isn't this like the issue with pitching and defense? What if the other receivers suck? Does this tell us anything? Especially in the case of Winslow. If all the receivers suck, does this tell us anything useful?

28
by tuluse :: Thu, 07/29/2010 - 5:51pm

It tells you Winslow is better at catching than anyone else on the team.

Look at the Bronco's numbers Bill posted above. Marshall has pretty average DVOA yet a high +/-. So maybe this tells you that Marshall was used in a way where he didn't really have a chance to make a valuable play in a lot of situations, but he was still really good at catching the ball.

29
by TheHerminator :: Fri, 07/30/2010 - 12:18am

Could you post the Chiefs' +/- numbers from last year? I have to see how bad everyone was if Bowe jumped up that high.