Snap-Weighted Age: 2014 NFL Rosters

by Scott Kacsmar
With the free-agency ink dry and the draft quickly approaching, we are deep into the team-building part of the football year. Teams are planning around the salary cap, getting younger and making moves for the future.
What is the right mixture of youth and experience for an NFL team to be successful?
This is a question we are not able to definitively answer, but age is just a number, so we can quantify it. Football Outsiders uses a metric called Snap-Weighted Age (SWA) that's designed to depict a team's average age based on snap counts. Some of the oldest players in the league are usually on special teams, but since they play a limited number of snaps, that should be factored into the calculation. Every player's age is weighted by how many regular-season snaps they played.
Going back to that question of a right mixture of age, 2014 did not provide much help. The correlation between a team's SWA and DVOA was -0.03. In recent seasons we observed a correlation slightly above 0.30 between offensive SWA and offensive DVOA. Last season was just 0.12, lower than defense (0.19). The league-wide SWA for defense was 26.7, same as 2013. Offense dropped a tick from 27.0 to 26.9 thanks in part to an incredible class of rookie wide receivers.
The answer may simply be that there is no superior way of building a team. While 2014's three oldest teams (Saints, Bears and Chargers) all missed the playoffs, the league's third-youngest team was one yard away from back-to-back championships. Old or young, drafted or a free agent, good players with good coaching is what drives success in the NFL. Teams have to explore every avenue to maximize the number of good players on their roster. That's always the goal.
In presenting SWA, we try to withhold any comments on whether a team's ranking is good or bad. Their methods of arriving at that ranking matter much more. If a team has a great defense with a lot of veterans, who are we to criticize them? That just may be a signal that this team needs to start investing into the future on that side of the ball.
Nothing lasts forever.
Snap-Weighted Ages for 2014
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 Age | Rank | OFF Age | Rank | DEF Age | Rank | ST Age | Rank |
ARI | 27.1 | 4 | 27.3 | 10 | 27.1 | 9 | 26.4 | 5 |
ATL | 26.6 | 20 | 26.8 | 16 | 26.6 | 21 | 26.4 | 7 |
BAL | 26.8 | 17 | 27.4 | 8 | 26.8 | 15 | 25.4 | 31 |
BUF | 26.3 | 26 | 26.6 | 21 | 26.1 | 26 | 26.1 | 15 |
CAR | 26.7 | 19 | 26.4 | 26 | 27.2 | 8 | 26.4 | 6 |
CHI | 27.2 | 2 | 27.9 | 3 | 27.0 | 12 | 26.3 | 9 |
CIN | 26.9 | 16 | 25.9 | 29 | 28.1 | 2 | 26.1 | 14 |
CLE | 26.4 | 25 | 26.6 | 18 | 26.4 | 22 | 25.8 | 23 |
DAL | 26.1 | 29 | 26.4 | 25 | 26.1 | 25 | 25.6 | 29 |
DEN | 26.9 | 15 | 28.6 | 2 | 25.7 | 31 | 25.6 | 27 |
DET | 27.0 | 7 | 27.0 | 15 | 27.5 | 6 | 25.9 | 20 |
GB | 26.2 | 28 | 25.7 | 30 | 26.7 | 18 | 25.9 | 19 |
HOU | 26.5 | 22 | 27.2 | 12 | 26.0 | 28 | 26.1 | 17 |
IND | 27.0 | 8 | 26.2 | 28 | 28.3 | 1 | 26.1 | 12 |
JAC | 25.4 | 32 | 24.7 | 32 | 26.1 | 27 | 25.5 | 30 |
KC | 26.4 | 24 | 26.6 | 19 | 26.6 | 20 | 25.7 | 26 |
Team | TOT Age | Rank | OFF Age | Rank | DEF Age | Rank | ST Age | Rank |
MIA | 26.6 | 21 | 26.2 | 27 | 27.3 | 7 | 25.7 | 25 |
MIN | 26.2 | 27 | 26.7 | 17 | 25.9 | 29 | 25.6 | 28 |
NE | 27.0 | 10 | 27.7 | 5 | 26.6 | 19 | 26.1 | 16 |
NO | 27.3 | 1 | 29.0 | 1 | 26.2 | 24 | 25.9 | 22 |
NYG | 27.0 | 11 | 26.6 | 20 | 27.6 | 5 | 26.7 | 3 |
NYJ | 27.0 | 9 | 27.3 | 9 | 27.0 | 11 | 26.1 | 13 |
OAK | 26.9 | 14 | 26.5 | 22 | 27.7 | 4 | 26.2 | 11 |
PHI | 27.0 | 6 | 27.2 | 11 | 26.9 | 13 | 26.9 | 1 |
PIT | 26.9 | 13 | 26.5 | 24 | 27.8 | 3 | 26.2 | 10 |
SD | 27.2 | 3 | 27.9 | 4 | 26.7 | 17 | 26.6 | 4 |
SEA | 25.8 | 30 | 25.3 | 31 | 26.3 | 23 | 25.8 | 24 |
SF | 27.1 | 5 | 27.6 | 6 | 26.8 | 16 | 26.4 | 8 |
STL | 25.6 | 31 | 26.5 | 23 | 25.0 | 32 | 25.2 | 32 |
TB | 26.5 | 23 | 27.5 | 7 | 25.7 | 30 | 26.0 | 18 |
TEN | 27.0 | 12 | 27.0 | 14 | 27.0 | 10 | 26.7 | 2 |
WAS | 26.8 | 18 | 27.1 | 13 | 26.9 | 14 | 25.9 | 21 |
AVG | 26.7 | 26.9 | 26.7 | 26.0 |
How did the Broncos finish with the second-youngest defense? DeMarcus Ware (32) was the only player older than 28 to play any snaps. Denver also had the second-oldest offense, giving the Broncos the biggest age differential in the league between its offense and defense.
Saints: Can They March Out of Cap Hell?
New Orleans was the oldest team in the league last year thanks to the game's oldest offense. Sure, it has been a historic offense since Sean Payton met Drew Brees, but things slipped a little and a case of cap hell is threatening to drop the offense deeper in 2015.
The Saints traded Jimmy Graham in a blockbuster move to Seattle, which might mean more snaps for 35-year-old Ben Watson. That's probably going to have to be fixed in the draft. Pro Bowl center Max Unger was also part of that deal, which makes the Saints seven years younger at center after 14 starts from the departed Jonathan Goodwin (36) last year. The Saints also said goodbye to "screen legend" Pierre Thomas (30) and traded Ben Grubbs (31) to Kansas City.
So the team has dumped some veteran contracts, though they also traded 23-year-old receiver Kenny Stills to Miami. The core trio of Saints that has paced this offense since 2006 (Brees, Jahri Evans and Marques Colston) remains, but the Saints are going to look a bit different in 2015.
Steelers: Still Old, But Not the Oldest Defense
Arguably the craziest SWA stat we've had so far is that Pittsburgh had the oldest defense for seven consecutive seasons (2007-2013). That streak is over after the Colts and Bengals edged out third-place Pittsburgh in 2014. We thought the Steelers would be even younger, but Cortez Allen's benching and injury-plagued seasons for Jarvis Jones and Ryan Shazier nixed that. The team even brought James Harrison and Brett Keisel back, but only the 37-year-old Harrison should return for 2015. Troy Polamalu and Ike Taylor recently retired after 12-year careers in Pittsburgh, so a revamped secondary should lead the team to its youngest defense under Mike Tomlin.
With defensive coordinator Dick LeBeau also gone, this really is the end of an era in Pittsburgh, at least for the defense. We may never see a streak in SWA like this again, because this was certainly done by design. Since Kevin Colbert took over as general manager in 2000, no team has been able to find and keep as many defensive stalwarts as the Steelers. Since 2000, the Steelers have had eight defenders play at least 150 regular-season games for them. The rest of the NFL has had just 21 such defenders in that time.
Re-checked and Lance Briggs was the only omission from the first table. So it's 29 total players. pic.twitter.com/oratYNL5Xw
— Scott Kacsmar (@FO_ScottKacsmar) April 14, 2015
Colts: Drafting Andrew Luck Was Ryan Grigson's "Easy Button"
Yes, that was a Staples reference. Indianapolis' path to becoming the oldest defense is the complete opposite of what the Steelers did. The Colts got there due to not finding players in the draft and having to load up the roster with veteran free agents and trades. Only 15.3 percent of Indianapolis' defensive snaps last year were from players the Colts drafted. By comparison, a team like Oakland gets to 16.7 percent on just Khalil Mack and Miles Burris alone. If we include undrafted free agents that only the Colts have had, that number is still just 27.5 percent.
[ad placeholder 3]
The Colts finished with the oldest defense despite the unit's best player, a 33-year-old Robert Mathis, missing the entire season due to suspension and injury. Their top three players in defensive snaps are all coming back in 2015 and will be at least 30 years old: Mike Adams (34), D'Qwell Jackson (32) and Greg Toler (30). This offseason the Colts added pass rusher Trent Coles (33) and safety Dwight Lowery (29), though their roles are to be determined.
Ryan Grigson has drafted one defender (Bjoern Werner) in the first four rounds of the 2012-14 drafts compared to eight offensive players. If Grigson does not start finding defenders, the Colts will continue spending on other team's players and filling out the roster with undrafted talent.
Only the Bengals had a roster with a larger age differential in favor of defense than the Colts. That could be changing this season with the Colts adding Frank Gore (32) and Andre Johnson (34) on offense, though we are still talking about short-term fixes. The lack of young building blocks on defense is a major problem.
Given what Grigson has built with Andrew Luck making a comparatively meager quarterback salary, it's scary to think what the team might look like when Luck is the highest-paid player in NFL history.
Growing Older, Growing Up?
If we compare SWA from 2014 to 2013, we can see which teams aged the most or went into the biggest youth revolt.
The Jets ranked 27th in offensive DVOA in 2013, but Geno Smith was only a rookie. He started 13 games last year and the Jets added Eric Decker, Jace Amaro, Chris Johnson and eventually traded for Percy Harvin. The Jets again ranked 27th in offensive DVOA while gaining a league-high 1.1 SWA (only team above 0.7).
Jacksonville went in the other direction. After ranking 32nd in offensive DVOA, the team dumped Maurice Jones-Drew for younger running backs, drafted Blake Bortles and surrounded him with three young receivers (Marqise Lee, Allen Robinson and Allen Hurns). The offensive line also started multiple rookies, leading to a league-high decrease of 1.9 in SWA. The end result was a ranking of 31st in DVOA (32nd in passing). The offense helped make the Jaguars the youngest team in the league, and adding a 27-year-old Julius Thomas should only help maintain that. Marcedes Lewis (31) hadn't shined since the 2010 season. We have SWA data back to 2006 and the Jaguars are the youngest offense (24.7) on record.
[ad placeholder 4]
Switching to defense, the Vikings had the biggest youth movement (-1.2 SWA) last year, saying goodbye to Jared Allen (32) and hello to rookie Anthony Barr. Cornerback Xavier Rhodes also became a 16-game starter in his second year. Mike Zimmer figured he could turn a young defense around in his first year, though the Vikings only climbed from 27th in defensive DVOA to 23rd in 2014. It is however a process and the Vikings should be moving in the right direction with players like Barr, Rhodes and Sharrif Floyd.
Cleveland gained the most experience on defense (+0.9 SWA) thanks in part to players like Karlos Dansby (33) and Donte Whitner (29; made the Pro Bowl). While Mike Pettine's defense struggled early, by season's end the Browns improved from 24th to 11th in DVOA (eighth in weighted DVOA) and will have to carry the team in 2015.
Again, it's not necessarily good or bad make these types of changes. Sometimes "aging" is just a roster sticking with a working core, and we know everyone only ages in one direction. One year after the San Diego defense had the biggest SWA decline (-1.9), the Chargers had the second-largest rise in SWA (+0.9) this season. They let those players grow together.
The Washington Tradition Ends
We have not forgotten special teams. The strongest correlation for SWA to DVOA in 2013 was on special teams. It was also negative (-0.35), but that correlation also shrunk close to zero in 2014 (-0.07).
Chip Kelly's Eagles had the oldest special teams, but finished first overall in special teams DVOA. Kelly's squad only ranked 25th in 2013 when the unit ranked 19th in SWA. I wouldn't say he has this down to a science yet.
For eight years in a row (2006-2013) the Redskins ranked among the four oldest special teams in the league. That finally changed in Jay Gruden's first year with a ranking of 21st, though some credit probably belongs to new special teams coordinator Ben Kotwica, an Iraq war vet. This story by Mike Jones of the Washington Post summarizes some of Kotwica's methods to get guys motivated for his "special forces." It also touches on some of the craziness during that abysmal 2013 performance, including ex-coordinator Keith Burns challenging a player to fight him.
Despite the efforts to get younger in Washington, the special teams still ranked 29th in DVOA with a negative rating in each of the five elements we measure. Just goes to show you can change the culture, but you can't magically will Kai Forbath to kick more touchbacks.
Comments
26 comments, Last at 24 Apr 2015, 5:43pm
#10 by MJK // Apr 22, 2015 - 11:41pm
Well, yes and no. It does suggest that a lot of their current teams are only a few years into the league, meaning they are still on rookie deals. Given that they are talented, means that they likely have a fair number of talented players about to become free agents and be in line for big deals--meaning that they are likely to suffer attrition.
#22 by Temo // Apr 24, 2015 - 6:51am
This, so much. I feel like every time people look at the age of teams and try to project ahead, they always end up overestimating.
And it's because the promise of youth is contingent on A) Players developing, B) Players staying healthy, C) Being retained through free agency.
But of course, players peaks, development, health, and contract demands are so variable and subject to so many variables it's almost useless to say "team is young, will continue to be good for a while".
#11 by ammek // Apr 23, 2015 - 3:46am
I remember during the 1997 playoffs the announcers marveling at how young the Packers' offense was. Favre was 28; Dorsey Levens, coming off a 1,400-yard season, was 27; they had two 1,000-yard receivers in Robert Brooks (27) and Antonio Freeman (25), and a pro bowl tight end Mark Chmura (28); and four of the starting linemen were under 27, not including the previous year's top pick John Michels (24). The announcers were convinced this offense was about to become a dynasty.
Three years later that crop of young talent had almost disappeared. Only Favre, Freeman and Ross Verba would start more than five games for the Pack in 2000 (and Verba had been moved inside to guard from LT where he had played in 1997). Chmura was out of football because he's a jerk; the starting guards were salary cap casualties; Brooks, Levens and RT Earl Dotson all had their careers torpedoed by injury; and Michels as well as third-string WR Derrick Mayes (aged 23 in 1997) and backup RB Travis Jervey (25) were busts. Instead, the 2000 Packers' offense would be relying on players such as RB Ahman Green (23), rookie tackle Chad Clifton (24) and WR Donald Driver (25).
It's a dangerous game, predicting the future.
#12 by dank067 // Apr 23, 2015 - 8:58am
Likewise, the 2010 Packers defense that was coming off a Super Bowl and consecutive 2nd place finishes in DVOA featured Clay Matthews (24), Nick Collins (27), BJ Raji (24), Tramon Williams (27), Sam Shields (23), AJ Hawk (26) and Desmond Bishop (26) all playing at a high level. (I guess Hawk was just ok.) It sure seemed to me they were set for a long time at basically every position group.
With a few exceptions (Henderson and Freeman in '95, Driver), GB really whiffed at drafting the offensive skill positions from 93-99. Easy to see how quickly they thinned out there when injury and attrition set in. They still managed to find plenty of good O-linemen though, even Flanagan, Rivera and Wahle were all drafted 96-98.
#2 by Led // Apr 22, 2015 - 1:42pm
Is there data on consistency of SWA from year to year? In addition to uncertainty about a connection between age and performance, if teams' SWA and rankings are relatively fluid from year to year then I'd be less likely to draw any conclusions from the numbers for particular years.
#3 by Scott Kacsmar // Apr 22, 2015 - 3:14pm
I just grabbed a file on 2006-2011 that I didn't have last night for this. The year-to-year correlation for total team SWA was the highest its ever been this year at 0.74. The lowest we have was 0.46 in 2007-08. For just rankings, 2013-14 correlation was 0.66.
Teams have been trending younger
League-wide SWA
2006: 27.2
2007: 27.2
2008: 27.2
2009: 27.1
2010: 27.0
2011: 26.9
2012: 26.8
2013: 26.7
2014: 26.7
#4 by jklps // Apr 22, 2015 - 5:07pm
"Despite the efforts to get younger in Washington, the special teams still ranked 29th in DVOA with a negative rating in each of the five elements we measure. "
29th is a big jump from 32nd that had historically bad numbers!
That is about all we can cheer about for our DC team.
#5 by Karl Cuba // Apr 22, 2015 - 8:35pm
Would it be possible to run these results with some measure of spread such as standard deviation so a couple of very old players don't overly skew the results.
And what is the evidence that an aging kicker or punter is a large problem? (Obviously a query from a niner fan with a really old kicker and an aging punter)
#8 by jtr // Apr 22, 2015 - 11:16pm
I initially thought the same thing, then realized they must be running SWA for the whole coverage and return teams rather than just the kicker and punter for special teams. It's a lot easier to see why a team would struggle with an aging returner or gunner than punter.
On a related note, I would be curious to see if low SWA on special teams has any correlation with offensive or defensive DVOA in subsequent seasons. In many cases, special teamers are starters-in-waiting who contribute as gunners, returners, etc while developing at their future positions. So possibly a team with young special teams in year n would be well set up in year n+1 or n+2 as those players mature.
#13 by Pottsville Mar… // Apr 23, 2015 - 9:25am
I don't understand how these can possibly be relevant when they're so bunched together. For instance, the Saints are the oldest team in the league at 27.3 - but if their average age was just half a year younger, they'd be younger than average (Washington is 18th at 26.8).
I suppose there's some value in noting the extreme outliers, like how Jacksonville is the youngest team in the league at 25.4, but it doesn't seem like fractions of a year would have much importance when discussing team-building strategies and problems.
#14 by jtr // Apr 23, 2015 - 10:44am
That's a very good point. It's also very sensitive to the presence of just one old player. As an experiment, I tried replacing Drew Brees with a younger qb. I replaced 1/22nd of their snaps (since Brees was responsible for 1/11th of their offensive snaps) with 26 instead of 36 years old. And it put them at 26.8 years old, which is right around average. So a 36 year old QB can raise an average aged team to the oldest team in the league all by himself; this does not speak highly of the utility of these numbers.
I'm not sure what a good way to mitigate the impact of a single old player on the ratings is; perhaps setting a cap on QB age in these ratings, or using the average of the QB's age and 30. Or maybe even dropping QBs from the average entirely. After all, QBs don't age like other positions, where each year over 30 usually means a slight dropoff for a QB but a dramatic dropoff at most other positions.
#15 by TacticalSledgehammer // Apr 23, 2015 - 10:50am
Doesn't that open a whole can of worms though? After all, you can argue that running backs drop off faster than recievers and cornerbacks, who drop off faster than linebackers, who drop off faster than offensive and defensive linemen, etc. Maybe separating into units rather than teams is the way to go - its probably worse to have a bunch of 30+ players at your skill positions than on the o-line.
#16 by jtr // Apr 23, 2015 - 11:14am
Splitting into units is a good idea. It's certainly true that a team with a 32 year old guard and a 25 year old running is better equipped for success than a team with a 25 year old OG and a 32 year old RB, but (all other things equal) those two teams would have the same SWA.
I guess the real takeaway is that the age of a football team is a much more complex issue than can really be effectively boiled down into one number.
#18 by Thomas_beardown // Apr 23, 2015 - 12:55pm
I would actually adjust age based on position instead. Give each player a score based on how old they are compared to league average for that position and possibly based on historical data for "peak" for that position.
#19 by jtr // Apr 23, 2015 - 1:30pm
I like that idea. I think Z-score (standard deviations above or below the average) for the player based on position would do nicely. Multiply each player's Z-score by the fraction of team snaps that he played, then add the whole thing together to get a total Z-score for a unit or team. This would scale player age appropriately to his position and give a much more meaningful number than we have with SWA.
#25 by Scott Kacsmar // Apr 24, 2015 - 4:37pm
We could certainly do something like that for units, but again I ask, what's a good or bad age for each position? I'm not sure how we'd tie the ages into performance here, especially for all the positions (OL and defense) we don't have an individual metric like DVOA for.
Tampa Bay and New Orleans both primarily used a 35-year-old QB last season, but Drew Brees is much better than Josh McCown. The consensus-best OL in the league (Dallas) had four players in the 23-26 range. The best receiver in the NFL last year may have been a 22-year-old rookie. The oldest player in the league, 42-year-old Adam Vinatieri, was named first-team All-Pro. If you draft a rookie, he's going to be in that 21-23 range and if he's a high pick, he should play a lot. That can skew the numbers, but that's just a matter of the team playing its premium pick at hopefully a key position.
I think with age we can draw conclusions like players don't peak before 24, or workhorse RBs start declining around 28-29, or that WRs not named Jerry Rice aren't very helpful after 35, but I'm not sold we can ever tie age and performance together in big-picture format. Game's still about good players.
#21 by wiesengrund // Apr 24, 2015 - 3:48am
I was wondeing how SWA correlates to AGL? Taking the example of the Falcons, that their Off and Def were so young last year was definetly not an effect of Dimitroff drafting well (he didn't and the picks didn't make up the bulk of the young snaps the team took), but more an effect of an avalanche of injuries, forcing more street-FAs and UDFAs to play snaps than if the team would have stayed healthy. The fact that the bottom half of NFL rosters are most likely comprised of UDFAs and young, cheap roster-filler players, might lead to a correlation between SWA and AGL, so I was wondering if you ever ran those numbers?
#24 by Scott Kacsmar // Apr 24, 2015 - 4:23pm
Definitely thought about looking at it compared to AGL.
Correlation between SWA and AGL
2006: -0.09
2007: 0.00
2008: -0.03
2009: -0.27
2010: -0.14
2011: -0.14
2012: -0.12
2013: 0.19
2014: 0.18
So usually not much there. Did used to be negative, which would support the idea that as injuries pile up, teams are forced to throw younger talent out there. But I guess that's not always the case, as evident by the Steelers bringing 36-year-old James Harrison back last season. Sometimes injury is the only way an old veteran can still see the field a lot.
SWA captures exactly what a team put on the field since it's snap-based, but that's definitely not a perfect representation of the intended team due to injuries and suspensions. That's where AGL comes in, where the intended starters and key subs are credited for missing time. So while I understand there should be a little relation between the two stats, they really are tracking different things philosophically (have vs. want).