Writers of Pro Football Prospectus 2008

Most Recent FO Features

WentzCar16-1.jpg

» Week 11 DVOA Ratings

DVOA has finally climbed on board the Wentz Wagon! The Eagles move into the No. 1 spot, but they aren't the only strong, well-balanced team in the NFL this year. New Orleans, Pittsburgh, and the Los Angeles Rams make this one of the best seasons ever for multiple teams over 30% in DVOA, and Minnesota isn't far behind.

19 May 2016

Snap-Weighted Age: 2015 NFL Rosters

by Scott Kacsmar

"Age is just a number."

We deal with many numbers at Football Outsiders, but in this case age really may be just another number, one that does not measure up to other, more telling metrics. Whether you are looking at a 21-year-old cornerback, a 30-year-old running back, or a 43-year-old kicker, a team will take all three if those players are good. That fictional team's average age would be 31.3 years, but we know kickers play fewer snaps than most position players. By accounting for how many snaps those players played in the regular season, we are able to give each team a Snap-Weighted Age (SWA).

What is the "sweet spot" for your team to rank in SWA? "Beats me" would be the type of brutally honest answer I would have written years ago, but we at least know a few things about age in the NFL. A very high SWA may signal that a team's window is nearly shut, so you'd better get in your last hurrah in while you can. (Just ask the 2002 Raiders.) It could also mean a past-their-prime team about to dump a lot of bad contracts with a rebuild around the corner. Meanwhile, a very young team may have just gone through such a rebuild, loading up on rookies and cheap free agents. That also sounds bad, but wouldn't it be a great thing if you dominated a draft or two and had so much young talent to develop?

In the NFL, youth is not inherently a positive feature, while experience is not necessarily negative (and vice versa). This is why we struggle so much each year to answer where a team should rank in SWA. After all, Carolina (third) and Denver (seventh) ranked among the oldest teams in the league on their way to Super Bowl 50, while the seven youngest teams in 2015 all failed to break .500. However, things rarely turn out that neatly. Consider this comparison of playoff teams to non-playoff teams since 2011:

  • Playoff teams' average SWA was 26.8, with an average rank of 16.0.
  • Non-playoff teams' average SWA was 26.7, with an average rank of 16.8.

So you may not want to rank first or 32nd in SWA, but anywhere in between might be fine. Any other time when we feature a table on here, it is rather clear where a team would prefer to rank. With age, you really cannot praise or criticize a team's ranking without understanding the makeup of their roster or their vision. We will rank the units from oldest to youngest below, but it is really done for reader convenience rather than making any definitive statement of quality.

We listened to your comments from last year's SWA article and will try to address a few of those thoughts here. We know there have been a recent rash of early retirements, and that SWA has been trending younger over time. In 2006, the league average SWA was 27.2; that has fallen to 26.6 in 2015, the lowest season yet. Offense has seen a steeper drop than defense, going from 27.6 in 2006 to 26.8 this past year as most teams have no problem with starting young quarterbacks now. Of course, this golden age has seen the quarterbacks in their thirties outplay their younger counterparts, but we know players generally decline with age. We also know that experience is valuable and that players rarely peak in their early 20s. As intriguing as it sounds, we probably have not yet seen the best of Odell Beckham Jr. or Marcus Peters.

There is consistently non-zero correlation between SWA and total DVOA. In 2015, that correlation coefficient was 0.27, which is not very strong, but a familiar result. Typically, offensive SWA and offensive DVOA had correlation around 0.30, but the last two seasons have not even been half of that. This past year was 0.14, while the defense was again stronger at minus-0.26. In 2013, the strongest correlation between SWA and DVOA was on special teams (minus-0.348). In 2014 that correlation was close to zero (minus-0.07), and last year was similar (minus-0.07).

2015 Snap-Weighted Age: By Unit

The following table shows SWA for the overall team (TOT) along with the unit breakdown for offense, defense, and special teams. Teams are ranked from oldest to youngest.

Team TOT SWA Rk OFF SWA Rk DEF SWA Rk ST SWA Rk
ARI 26.9 11 28.2 3 26.0 26 25.8 21
ATL 27.1 6 27.5 8 26.9 14 26.7 5
BAL 26.6 16 26.5 18 27.1 10 25.6 27
BUF 26.3 23 26.2 21 26.4 21 26.5 9
CAR 27.4 3 27.0 14 28.1 3 26.8 3
CHI 26.6 17 27.4 10 25.8 31 26.2 13
CIN 27.0 8 26.2 23 28.1 2 26.5 10
CLE 27.0 9 27.4 9 27.1 8 25.6 25
DAL 26.3 26 26.9 15 25.9 29 25.7 23
DEN 27.1 7 28.3 2 26.5 19 25.6 26
DET 26.6 14 26.2 24 27.3 6 26.2 15
GB 26.3 24 26.7 16 26.3 23 25.5 28
HOU 26.3 25 26.5 17 26.2 24 25.8 20
IND 27.6 2 27.1 13 28.6 1 26.3 11
JAC 25.9 31 25.6 30 26.4 22 25.4 29
KC 26.5 19 25.8 28 27.4 5 25.8 19
Team TOT SWA Rk OFF SWA Rk DEF SWA Rk ST SWA Rk
MIA 26.0 29 25.5 31 26.6 16 25.3 30
MIN 26.7 13 26.4 19 27.5 4 25.7 22
NE 26.5 20 27.2 12 25.9 27 25.9 18
NO 27.3 5 28.2 4 26.5 20 26.7 4
NYG 26.6 15 26.2 22 27.0 13 26.5 8
NYJ 27.6 1 28.5 1 27.1 9 26.6 6
OAK 26.5 18 26.2 20 26.6 17 27.2 1
PHI 26.9 10 27.2 11 26.7 15 26.9 2
PIT 27.4 4 28.2 5 27.0 11 26.2 14
SD 26.8 12 27.6 7 25.9 30 26.5 7
SEA 26.4 22 25.9 25 27.0 12 26.3 12
SF 26.2 27 27.7 6 25.4 32 25.1 31
STL 25.5 32 25.2 32 26.1 25 24.9 32
TB 25.9 30 25.9 27 25.9 28 26.2 16
TEN 26.0 28 25.6 29 26.5 18 25.7 24
WAS 26.4 21 25.9 26 27.1 7 26.0 17
NFL 26.6 26.8 26.7 26.0

The Saints were 2014's oldest team, but fell to fifth last year after more veterans departed from Sean Payton's golden era. The Jets took over as the oldest team thanks to the league's oldest offense, comprised largely of transplanted veterans, including Ryan Fitzpatrick (33), Brandon Marshall (31), Eric Decker (28), and Chris Ivory (27). While a 31-year-old Matt Forte will age this group in 2016, they stand to get younger quickly with D'Brickashaw Ferguson's retirement and at quarterback, as Fitzpatrick remains a free agent. Second-round pick Christian Hackenberg would be a 21-year-old rookie starter should he ultimately win that job. Similarly, the Broncos had the second-oldest offense with a 39-year-old Peyton Manning starting half the year, but have changed their offensive line again and could end up starting 22-year-old rookie quarterback Paxton Lynch.

Arizona is a smart bet for 2016's oldest offense, assuming Carson Palmer (37) and Larry Fitzgerald (33) stay healthy. D.J. Humphries was 2015's only healthy first-round pick to never see the field. While he may win the right tackle job this year (in his age-23 season), the Cardinals also have two veteran guards expected to start in Mike Iupati (29) and Evan Mathis (35 in November).

We may never see another run like that of the recent Steelers, who had the oldest defense for seven seasons in a row (2007-2013), but the Colts have done so in back-to-back years now. This has been the result of general manager Ryan Grigson's inability to draft defensive starters and his heavy reliance on acquiring older free agents to fill out Chuck Pagano's middling defense. While Henry Anderson was a fine pickup last year, the Colts have a great shot at the three-peat as they still plan to use Robert Mathis, Trent Cole, Mike Adams, D'Qwell Jackson, and Erik Walden this year. Each player will be 31 or older.

Indianapolis did not repeat as the most defensively-skewed team in the league when it comes to age; the Colts finished second to Cincinnati in that differential. The Bengals paired the second-oldest defense with the 23rd-ranked offense. The offense is still going to be young, but there are some pieces in place to replenish the defense. Leon Hall is a free agent, but the Bengals drafted cornerback William Jackson III in the first round. Andrew Billings could be a fourth-round steal as an eventual replacement for Domata Peko, who turns 32 in November. San Francisco paired the league's youngest defense with the sixth-oldest offense, but look for that to even out more under Chip Kelly. Anquan Boldin remains a free agent, while the 49ers had some defensive injuries to veterans such as Glenn Dorsey and Antoine Bethea last year.

A year ago, we observed the Rams and Jaguars as the two youngest teams in the league. The only thing that has changed is the amount of hype each brings to the 2016 season. We know about the talent the Rams have acquired, mostly in the front seven, but is No. 1 overall pick Jared Goff finally the answer at quarterback as the team returns to Los Angeles? Todd Gurley sure looks poised to become the league's next great running back, so things could be getting interesting again with the Rams. Meanwhile, has any team added more premium, young defensive talent than Jacksonville? Due to Dante Fowler's minicamp injury last year, the Jaguars can basically boast three stud rookies on defense with his return and the 2016 draft additions of Jalen Ramsey and Myles Jack. On paper, Jack may only be the 36th pick due to injury concerns, but let the record show that just about every major draft analyst had him as one of the very best players available this year. In a winnable AFC South, if Gus Bradley cannot turn this defense (which also picked up Malik Jackson in free agency) around, then it is time to move on.

Of course, it would be wise to temper expectations for an inexperienced pass-rusher coming off a major knee injury like Fowler. We also know rookie cornerbacks like Ramsey rarely impress, as there is a steep learning curve at that position. Maybe Jack's medical history will prove to be his downfall. Finally, we know high-priced free agents like Jackson, who has little track record of success and was part of an incredible Denver defense, are far from sure things when they jump ship. Not to rain on Jacksonville's parade here, but 2016 seems unlikely to be a huge year for this team. The future is looking bright, though, with the defense's potential and that the offense's young core of Blake Bortles, T.J. Yeldon, Allen Robinson, and Allen Hurns.

(For the record, last year we calculated Jacksonville's SWA with an incorrect birth date for Bortles, but even the Jaguars aged him differently in year two.)

Age is just a number that sometimes gets misreported.

2015 Snap-Weighted Age: By Range

You probably noticed that a lot of the ages are tightly bunched together around the league. In other words, it would be very hard to build a team with an average age of 24 or 30, especially given the salary cap. In the pre-cap era, maybe there was a larger split in age between the teams. Hall of Fame coach George Allen was notorious for trading away draft picks to acquire older veterans. But we are rather limited in the years for which we have snap data.

Generally, a team's SWA will not change much over the course of the season unless a rash of injuries occurs. For 2015, every team had a standard deviation under 1.0 for SWA on each of the three units. Pittsburgh had the highest standard deviation on offense at 0.97, which can largely be explained by its heavy use of one running back. DeAngelo Williams is nine years older than Le'veon Bell, who started the season suspended and ended it on injured reserve. The Vikings (0.91) had the highest deviation on defense. You may recall a late stretch of the season where Harrison Smith, Anthony Barr and Linval Joseph were injured.

Injuries can drastically change a team's SWA. Recall the 2015 Ravens, who lost multiple skill players at each position. You probably are not surprised to learn that the Ravens had the largest range of SWA on offense at 2.5, which was calculated by taking the team's "oldest" game and subtracting its "youngest" game for the season. Atlanta's offense had the smallest range (0.45). Below is a graph that charts the SWA for Baltimore's offense (blue) and defense (red) for each week.

The defense was at its oldest in Week 1, which was when Terrell Suggs tore his Achilles tendon in Denver. The offense eventually ended up losing Steve Smith, Joe Flacco, Eugene Monroe, Jeremy Zuttah, and Justin Forsett for the season, finishing with a much younger cast.

St. Louis had the most consistent SWA from week to week. The offense had the second-smallest range (0.46), and the defense had the smallest range (0.54). Carolina's defense actually had the largest range (2.25), which makes sense when you think of the injuries to older veterans like Charles Tillman and in-season acquisition of Jared Allen.

2015 Snap-Weighted Age: By Position

Finally, we adjusted SWA by position to get a better picture of where teams are younger or older compared to the rest of the league. We excluded fullbacks since some teams never use them. "ST" includes the specialists on special teams.

The method used was to create a z-score for each team's SWA by position. For example, Denver had three quarterbacks combine to take 1,109 snaps last year. A 39-year-old Peyton Manning took 593 of those snaps, or 53.5 percent. We multiply each player's snap percentage by his age and sum them together for each team and position. Denver's quarterback SWA was 32.5, compared to the NFL's average of 29.3 in 2015. We take that difference and divide it by the standard deviation in quarterback SWA (4.16) to get a z-score of 0.78 for Denver. The same was done for each team for each position to get a ranking in z-scores.

Instead of going nuts with z-score tables, here is a summary chart to show where each team's z-score ranks by position. The 12 playoff teams are highlighted, and each team's rank is broken down into quarters where dark green are the eight youngest and dark red are the eight oldest. In the final two columns, the average rank of the positions is shown. The Rams are still the league's youngest team and the Jets are the oldest, but hopefully this does a better job of highlighting why that is.

Posted by: Scott Kacsmar on 19 May 2016

19 comments, Last at 23 May 2016, 8:15am by Ferguson1015

Comments

1
by tuluse :: Thu, 05/19/2016 - 10:57pm

Chicago's offense and defense DVOA and SWA ordinal rankings match exactly.

Also, the position chart is cool info, but hard as hell to read.

3
by Tomlin_Is_Infallible :: Fri, 05/20/2016 - 11:43am

"Also, the position chart is cool info, but hard as hell to read."

an easy fix, if the site hires someone who actually knows how to use anything other than excel.

/agreed
// double plus good.

--------------------------------------
The standard is the standard!

5
by Scott Kacsmar :: Fri, 05/20/2016 - 1:07pm

Most of this was done in Access, but what specifically is hard to read? Do we need wider columns? Looks pretty clear on my PC, but maybe it's not so good on a phone or tablet.

7
by Tomlin_Is_Infallible :: Fri, 05/20/2016 - 1:11pm

My bad if true. That said, Access=Excel= crap.

Viewing on a 17.3" wqxga laptop.
It isn't the size or resolution.

Make no mistake- it's the fact it's a table (not a graph). It's a table with limited color palette. It's great info, but a picture tells 1000 words.

--------------------------------------
The standard is the standard!

10
by Scott Kacsmar :: Fri, 05/20/2016 - 1:18pm

Not sure how you'd show 32 teams and 9 positions without it being a total mess. That's like the section with the Baltimore graph. I couldn't do all 32 teams on one graph, it'd look like shite, even if you ran one for offense and one for defense. At best you'd have to make eight of them for one division each. So I just took the most interesting example and called it a day. If we were running a high-tech site with plenty of staff, we'd probably have a drop-down menu where you can select a team and see their graph, but you know that's not FO.

12
by Tomlin_Is_Infallible :: Fri, 05/20/2016 - 1:23pm

At the risk of getting banned (again) for disagreeing with a site author, sure you could post all 32 teams. etc.

You just need to stop using scatter/bar charts, 8 color palette options, (microsoft) etc.

I know you data-dumped the 1995-era draft value stuff a few days ago. I'll post my version of a plot of that data later tonight to illustrate.

--------------------------------------
The standard is the standard!

16
by justanothersteve :: Fri, 05/20/2016 - 3:19pm

I'm not complaining. Tomlin_Is_Infallible has a point. The table could be more clear without much effort. Here's a bit of what I believe would help.

1) Use a better color scheme in the tables, especially the second table. The two shades of red are too similar. A red-orange-yellow-green (and a more soothing shade of green than the puke green used) would be a better option. Use pastels instead of the default colors. Also, never use black as a background in any cell unless it's to call attention to something. A shade of gray would work better. Possibly alternate with a light blue on your table to highlight playoff vs non-playoff teams. I'd also have the emphasis color as blue with the non-playoff teams gray. The long red bar is very distracting. A light shade of gray would break the table without drawing attention to the break. The overall layout isn't bad. But the colors are poorly chosen. (Note: To improve the experience for colorblind readers, a red-orange-yellow-blue scheme could be used.)

2) Group the positions better. ST is in the middle of the offensive positions. Position groups jump around. Nobody thinks of defense as DB, DL, LB since defensive groups are usually described as DL-LB-DB; think of a 3-3-5 or 3-4-4. A column order of (for example) DL, LB, CB, QB, RB, TE, WR, OL, ST would make more sense. RB-TE is again a quick method of describing offensive sets (e.g., 1-2 or 2-2 sets) so it might be better to keep them together. A small gap or wider line between defense, offense, and special teams would also help.

You may prefer to place the QB and ST columns next to each other on the right, since they are the positions that seem least influenced by age. It doesn't matter if your QB is as old as Brady or as young as Luck, as long as he's good. Same with ST. Having a kicker as old as Morten Andersen or George Blanda at the end of a career could greatly skew the reality. I'm not even sure ST age should matter here as it may be more telling that a team has screwed up the position if they are reliant on younger players. If you do put the QB next to ST, try ordering the offense columns something like OL, RB, TE, WR, QB. You still want the skill positions together.

You may have a different way of how you want to list the columns. DB, LB, DL, OL, TE, WR, RB, QB, ST is essentially how you see teams if you are scanning across the field starting with the back of the defense. I kept QB last based on my previous paragraph. But the important thing is to keep a logical order.

I don't think this changes anything about the column. I'm fine with just one example of how a team's age can vary over the course of a season, and that graph is just fine (assuming you chose the purple box to highlight it's the Ravens and would use other colors for other teams). I'm very appreciative of the info regardless. Just pointing out how to make it look better and be more understandable.

9
by tuluse :: Fri, 05/20/2016 - 1:17pm

It's not hard to read as in each cell, but it's just information overload or something. I don't know, I have no suggestions on how to improve it and I'm very grateful to have the info at all.

11
by Tomlin_Is_Infallible :: Fri, 05/20/2016 - 1:20pm

Part of what I do for science research is catalysis reaction mechanisms across suites of materials. We can post graphs (or other, ugh) with reaction pathway data, activation barriers, etc. Yet- what people get funding for and what gets a paper into a 10+ impact factor journal is condensing that data down to a descriptor-based Volcano plot. (100 variables, correlate into 2 or 3, etc).

I see this table in a smaller-yet-similar basis.
--------------------------------------
The standard is the standard!

14
by SandyRiver :: Fri, 05/20/2016 - 3:00pm

Info overload is what I thought as well. It's all there but these old eyes struggle a bit to take it all in. However, given the modest (at best) correlation between SWA and team performance, I don't think any deeper/more detailed analysis would tell us much.

2
by Ferguson1015 :: Fri, 05/20/2016 - 11:21am

I'm a little surprised that SD's STs is considered the 9th oldest with a 2nd year kicker and a rookie punter. Is it projecting the Special Teams as it was last year without Donald Brown and Eric Weddle?

8
by Scott Kacsmar :: Fri, 05/20/2016 - 1:11pm

These are not projections, they are the 2015 results. So Scifres was a 35-year-old punter.

19
by Ferguson1015 :: Mon, 05/23/2016 - 8:15am

Of course they are. You only make that abundantly clear in every part of the article. I don't know what I was thinking. I blame lack of sleep due to my newborn. Yeah, clearly it is that and not me being an idiot.

4
by Hoberk Unce :: Fri, 05/20/2016 - 12:34pm

"The following table shows SWA for the overall team (TOT) along with the unit breakdown for offense, defense, and special teams. Teams are ranked from oldest to youngest."

Looks like the table is ranked alphabetically.

6
by Scott Kacsmar :: Fri, 05/20/2016 - 1:08pm

That's just semantics. It's sorted alphabetically, ranked oldest to youngest.

13
by nat :: Fri, 05/20/2016 - 2:16pm

Is there a point to computing the z-scores for a team's units, and then averaging their rankings rather than the z-scores themselves?

Doesn't that lose all the benefits of using a z-score?

15
by Scott Kacsmar :: Fri, 05/20/2016 - 3:07pm

Being quicker to do was the benefit to me. You get a similar order from that.

Rk Team Avg_Z-Score
1 IND 0.70
2 NYJ 0.66
3 PIT 0.55
4 PHI 0.40
5 CAR 0.40
6 NO 0.31
7 ARI 0.30
8 ATL 0.29
9 CIN 0.27
10 SD 0.23
11 DEN 0.17
12 NYG 0.07
13 CLE 0.07
14 SEA 0.06
15 HOU 0.04
16 CHI 0.04
17 DET 0.03
18 OAK -0.02
19 NE -0.07
20 BAL -0.09
21 MIN -0.12
22 DAL -0.12
23 WAS -0.15
24 GB -0.16
25 KC -0.18
26 SF -0.26
27 BUF -0.27
28 TEN -0.47
29 TB -0.59
30 MIA -0.63
31 JAC -0.70
32 STL -0.75

17
by nat :: Fri, 05/20/2016 - 4:25pm

Thanks.

As you say, it doesn't change the rankings too much. The differences come from weighting different players' snaps differently. Smaller units (e.g. QBs) naturally have higher standard deviations to drive their z-scores lower, but get weighted more per snap. The overall effect is.... I have no idea, other than noting that the two rankings are somewhat different.

18
by justanothersteve :: Sun, 05/22/2016 - 11:29am

Thanks for the changes on the second table. It's both easier on the eyes and you can quickly see where teams are young or old. Much better.