Stat Analysis
Advanced analytics on player and team performance

2018 Adjusted Games Lost: Part I

by Vincent Verhei

A handful of trends clearly stand out in the NFL injury data from 2018.

  • Injuries, as a whole, have been on the rise in recent years … or, at least, teams are reporting injuries more often.
  • This is particularly true for wide receivers, tight ends, offensive linemen, but most of all for defensive backs.
  • Most of the time, it's obvious from the injury report who will and who won't be playing on Sundays.

We have collected the data from the NFL's weekly injury reports for every season since 2002. This allows us to measure not just who played and who didn't, but who was able to play with injuries, even if those injuries meant the player was at something less than 100 percent. That's why we call this metric adjusted games lost -- in addition to players who missed games entirely, we also count those who hit the field after appearing on the injury report at an adjusted rate. Further, we track whether the injured player was a starter, a situational reserve, or simply bottom-of-the-roster fodder. Obviously, an injury to a starting tackle is more important than one to a guy who only plays on special teams.

We should note that we have changed our definition of "situational reserves" this year. In years past, we have used rather broad strokes to look for non-starters with specific roles -- third-down backs, third wide receivers, nickelbacks, etc. This year, we actually looked at each player's snap count to try to determine who was regularly seeing the field, including rotational defensive linemen, two-down linebackers, and change-up running backs. This is far more art than science, and it can be difficult to measure as player's roles change throughout the year, but we think it gives a more accurate measure of who is and is not an important role player.

As a result, we ended up with more AGL by situational reserves this season than ever before, a total of 455.4. That breaks the record of 438.7 set in 2014, which is the only other year this number broke 400, and it is partly why 2018 set a new record with 2,484.2 AGL across the league, just nosing past the 2481.2 figure of 2016. That record likely won't last for long, as injuries have steadily been climbing upward over time. From 2002 to 2007, league-wide AGL was below 1,600 every season. From 2008 to 2011, that number fell no lower than 1,600, but never hit 2,000. Since 2012, however, it has never fallen below 2,000. Initial reaction might be to tie the rise in injuries to the 2011 CBA that limited offseason practices, but in reality things have been much more gradual, and the rate of increase has not changed since that season.

If we limit things strictly to starters, injury levels have dropped slightly the last two years but are still higher than they were before 2016. That year set the record for AGL by starters at 2040.2. The last two years, the AGL totals for starters have been 1920.9 in 2017 and 1918.4 in 2018.

As a reminder, players are listed as either questionable (theoretically meaning they had a 50 percent chance of missing that week's game), doubtful (75 percent), or out (100 percent) on weekly injury reports. In prior years, there was also a probable (25 percent) designation, but the NFL did away with that in 2017. After two years of data, we now see that this was pointless. Teams simply leave that space blank for players who are probably going to play, and those players play at about the same rate that probable players did in the past.

Game Status Report Results, 2016-18
Game Status 2016 2017 2018
Players DNP% Players DNP% Players DNP%
Probable/Blank 1,527 4.6% 2,011 4.1% 2204 4.3%
Questionable 1,733 25.8% 1,426 33.0% 1204 32.1%
Doubtful 213 99.5% 170 97.6% 160 95.6%
Out 971 100.0% 852 100.0% 856 98.5%

(Yes, 13 players who were listed as out in the injury reports ended up playing in 2018. Four of them ended up starting: Indianapolis' Ryan Grant and Minnesota's Everson Griffen, both in Week 2, and Tampa Bay's Donovan Smith and Mike Evans, both in Week 9.)

We also see that the alleged percentile designations are largely meaningless. You can be very certain that probable/blank players will play, and that doubtful players won't. Questionable players end up playing about twice as often as not, though that data needs to be taken with a heaping helping of salt.

2018 Injury Reports: Blank and Questionable Players
Team Blanks Played Pct. Rk Questionable Played Pct. Rk
NYJ 154 151 98.1% 5 35 31 88.6% 2
HOU 150 146 97.3% 11 51 24 47.1% 29
WAS 118 106 89.8% 30 60 40 66.7% 16
MIA 116 113 97.4% 10 56 37 66.1% 20
OAK 107 102 95.3% 22 37 21 56.8% 26
IND 99 97 98.0% 6 59 48 81.4% 4
SEA 93 90 96.8% 15 43 25 58.1% 25
ATL 88 87 98.9% 3 6 4 66.7% 16
CLE 86 84 97.7% 7 39 30 76.9% 6
CAR 85 83 97.6% 8 19 16 84.2% 3
CIN 83 81 97.6% 9 16 9 56.3% 28
DEN 79 76 96.2% 17 17 10 58.8% 24
SF 76 68 89.5% 31 45 36 80.0% 5
MIN 74 72 97.3% 12 32 22 68.8% 15
ARI 70 63 90.0% 28 77 51 66.2% 19
JAX 68 66 97.1% 13 35 21 60.0% 23
KC 64 60 93.8% 24 28 17 60.7% 21
GB 61 60 98.4% 4 66 40 60.6% 22
NO 60 54 90.0% 28 9 3 33.3% 31
TEN 59 55 93.2% 25 27 11 40.7% 30
PHI 54 50 92.6% 26 33 24 72.7% 10
DAL 53 51 96.2% 16 48 36 75.0% 7
TB 51 49 96.1% 19 40 37 92.5% 1
NYG 46 44 95.7% 21 10 2 20.0% 32
PIT 35 35 100.0% 1 15 11 73.3% 9
BUF 33 32 97.0% 14 30 17 56.7% 27
LAC 33 31 93.9% 23 21 15 71.4% 12
NE 26 25 96.2% 18 85 59 69.4% 14
DET 25 24 96.0% 20 64 48 75.0% 7
CHI 25 25 100.0% 1 20 14 70.0% 13
LAR 21 18 85.7% 32 6 4 66.7% 16
BAL 12 11 91.7% 27 75 54 72.0% 11

What's most interesting in this table is not who played and who didn't, but how teams reported injuries differently. The Houston Texans had led the league in probable/blank designations in each of the past five seasons. Last year they fell all the way to … second place behind the New York Jets. The Jets themselves had finished second in this category four times in the past five years. The Ravens, meanwhile, only had a dozen blanks all season, and only one of them actually missed a game.

The data on questionable players is also interesting. For the fourth year in a row, no team had fewer questionable designations than the Atlanta Falcons, who had just six all season (the Rams also had just six). The Patriots led the league with 85 questionable players, fourteen times as many as Atlanta. There is also a huge variance from team to team in just how often questionable players ended up playing. The Buccaneers listed 40 players as questionable, but only three of them ended up missing that week's game. The Giants only listed 10 players as questionable, but only two of them actually played. In the end, the only thing truly questionable about the questionable designation is its reliability as useful information.

2018 AGL Results

2018 AGL Results
Team 2018 AGL Rk 2017 AGL Rk Dif
BAL 29.7 1 101.6 27 -71.9
BUF 33.8 2 44.5 9 -10.7
CHI 36.6 3 118.1 31 -81.5
LAR 39.6 4 15.6 1 +24.0
PIT 40.4 5 30.5 4 +9.9
NO 50.7 6 95.9 24 -45.2
CLE 51.1 7 52.7 12 +1.6
NYG 53.3 8 96.3 25 -43.0
KC 60.8 9 80.0 19 -19.2
SEA 65.7 10 88.2 22 -22.5
TEN 68.7 11 27.6 3 +41.1
MIN 73.9 12 47.7 11 +26.2
NYJ 74.6 13 47.4 10 +27.2
DEN 76.2 14 65.9 15 +10.3
DET 76.8 15 72.7 17 +4.1
ATL 77.9 16 16.3 2 +61.6
DAL 78.5 17 37.2 5 +41.3
NE 78.6 18 61.3 14 +17.3
OAK 81.7 19 44.2 8 +37.5
HOU 89.6 20 107.3 29 -17.7
GB 92.7 21 84.8 21 +7.9
MIA 95.3 22 115.5 30 -20.2
ARI 95.6 23 105.6 28 -10.0
WAS 99.3 24 121.0 32 -21.7
LAC 100.0 25 67.6 16 +32.4
CAR 103.8 26 42.9 7 +60.9
JAX 104.6 27 38.3 6 +66.3
CIN 104.9 28 80.8 20 +24.1
SF 105.5 29 91.6 23 +13.9
IND 116.5 30 101.2 26 +15.3
PHI 118.5 31 53.5 13 +65.0
TB 128.0 32 78.6 18 +49.4

The Baltimore Ravens got through 2018 with the least amount of damage, while the Tampa Bay Buccaneers suffered the most carnage. In the Ravens' case, this is a relative term -- they fail to make the top 100 healthiest teams on record, and just last year they would have finished fourth. Specific Ravens who battled injuries last year include offensive linemen Alex Lewis (6.9 AGL), James Hurst (6.3), and Ronnie Stanley (1.9); defensive backs Tony Jefferson (2.9) and Marlon Humphrey (2.9); and quarterback Joe Flacco (4.0 between being injured as the starter and officially returning as the backup to Lamar Jackson).

Tampa Bay, meanwhile, certainly had a bad season, but not among the bottom ten we have studied, and not close to the all-time record (158.8 by the 2016 Chicago Bears). We'll get into specific Buccaneers in a later column, but suffice to say that if you played defense for Tampa Bay in 2018, you are probably in pain as we speak.

Not surprisingly, a high AGL is a bad thing. The correlation between AGL and DVOA was -0.398 -- good teams, generally speaking, had fewer injuries than bad teams. Three of the top four, four of the top six, and six of the top 10 teams in AGL made the playoffs. Injuries weren't necessarily a death knell, however; the Eagles and Colts finished just ahead of the Bucs, and each of those teams still won a playoff game.

We've also included AGL data from 2017 so you can see which teams had the biggest improvements and declines. Six of the seven teams that made the playoffs in 2017 but missed them in 2018 did so in part because they suffered more injuries than they had a season prior; the Jaguars had the biggest decline in the league, and the Falcons, Panthers, and Titans also had a long list of ailments. Four of the teams that took their place in the postseason field -- the Bears, Ravens, Seahawks, and Texans -- had significantly better injury luck in 2018 than they had in 2017.

As noted, most of the Buccaneers' league-leading AGL came on the defensive side of the ball -- to a degree that we've never seen before.

Offensive and Defensive AGL, 2018
Team Offensive AGL Rk Defensive AGL Rk Dif
CLE 10.3 1 40.8 22 30.6
LAR 10.9 2 28.7 13 17.8
BUF 16.8 3 17.0 6 0.2
BAL 17.2 4 12.4 2 -4.8
TB 17.3 5 92.0 32 74.7
CHI 20.7 6 15.9 4 -4.8
PIT 25.7 7 14.8 3 -10.9
GB 28.4 8 64.3 30 35.9
NO 29.1 9 21.6 8 -7.5
TEN 31.2 10 37.5 19 6.2
SEA 32.8 11 32.9 14 0.1
ATL 33.3 12 44.6 25 11.3
NYG 34.0 13 19.2 7 -14.8
MIN 36.3 14 37.6 20 1.3
KC 36.5 15 24.3 9 -12.3
OAK 41.5 16 40.2 21 -1.3
DET 41.8 17 35.0 17 -6.7
NE 42.8 18 35.8 18 -7.0
DAL 44.1 19 34.5 16 -9.6
HOU 44.6 20 44.9 26 0.3
PHI 47.2 21 71.3 31 24.2
LAC 47.5 22 52.5 28 5.1
DEN 48.9 23 27.3 11 -21.7
NYJ 49.7 24 24.9 10 -24.8
CAR 51.3 25 52.5 27 1.2
IND 59.7 26 56.8 29 -2.8
CIN 61.1 27 43.8 23 -17.2
SF 61.3 28 44.2 24 -17.1
ARI 61.5 29 34.2 15 -27.3
MIA 67.0 30 28.3 12 -38.7
JAX 88.4 31 16.2 5 -72.2
WAS 88.9 32 10.3 1 -78.6

The Bucs' 92.0 AGL on defense absolutely shatters the prior record of 77.5 held by the 2014 Oakland Raiders. Tampa Bay had the most injuries at linebacker -- which is particularly astounding because they were a 4-3 defense and theoretically used linebackers 25 percent less often than several teams -- and the second-most in the secondary, and the Bucs were in the top ten for defensive line injuries as well. In related news, Tampa Bay finished dead last in defensive DVOA, and fired defensive coordinator Mike Smith midway through the season. Under the circumstances, it's amazing the gap between them and next-to-last Atlanta was so close. Then again, the Bucs improved to 26th in defensive DVOA over the second half of the season (subscription required), so Smith may have been part of the problem as well.

The good news for Tampa Bay is that the offense was quite healthy in 2018. (This includes injuries only and does not account for Jameis Winston's suspension.) The gap between Tampa Bay's defensive and offensive AGL is more than double that of any other team. Green Bay and Cleveland were next in line in that category.

No team had a bigger gap the other way than Washington, which had the most AGL on offense but the fewest on defense. They were followed closely by Jacksonville (next to last in offensive AGL, fifth in defensive AGL). Those two teams finished second and third on the list of most injured offenses on record, trailing only the 2016 Vikings.

Later this week we'll break things down by unit to just why teams like Tampa Bay, Washington, and Jacksonville finished as poorly as they did.

Comments

6 comments, Last at 13 Jun 2019, 5:11pm

1 Re: 2018 Adjusted Games Lost: Part I

As a reminder, players are listed as either questionable (theoretically meaning they had a 50 percent chance of missing that week's game), doubtful (75 percent), or out (100 percent) on weekly injury reports. In prior years, there was also a probable (25 percent) designation, but the NFL did away with that in 2017.

The NFL does NOT use those percentages, and hasn't for many years.

Recently...

Out = certain to not play.
Doubtful = unlikely to play.
Questionable = uncertain if player will play.

It also requires that players be listed if they missed or had curtailed practices due to injury, unless they are "certain" to play. I suspect these players are the origin of those pesky "Blank" statuses. A player who is deep on the depth chart might not play, even if he is fully recovered from his Tuesday practice muscle strain. Such a player would be listed, because to leave him off would be to risk violating the policy. But "Questionable" seems wrong if he is as likely to play as in any other week.

2 Re: 2018 Adjusted Games Lost: Part I

One of the biggest reason why I'm very bullish on the Eagles.
With more luck on the health side (bring back Chip smoothies) they should be considered a contender, especially given the outlook on offense and despite a terrible LB corps/questionable DE depth.

3 Re: 2018 Adjusted Games Lost: Part I

A useful metric but still fairly broad brush. The Pats, for example, were relatively less healthy last year than in 2018 but come the playoffs, they had key contributors ready to roll (Flowers, Edelman, Gronk, etc). It seems like teams should be as cautious as possible in the regular season with irreplaceable guys to provide a better chance of having them coming December. Of course, you have to make the playoffs too.
I wonder if it would be useful to weight positions in some way. Losing a top ten QB, for example, would be far more damaging to a team than an injury at any other position on the roster.

4 Re: 2018 Adjusted Games Lost: Part I

Not surprisingly, a high AGL is a bad thing. The correlation between AGL and DVOA was -0.398 -- good teams, generally speaking, had fewer injuries than bad teams.

Was this consistent from start to finish? I would assume bad teams would be more likely to sit players with borderline injuries than good teams. This isn't a difference in philosophy or view on injury, just a change in risk/reward between teams that are say 8-6 vs a team that is 3-11.

6 Re: 2018 Adjusted Games Lost: Part I MATH ERROR

I found a dependency in the numbers you present in your 2 charts. TB's total offense + defense AGL should be 109.3, but you have it as 128. That change would make them 3rd worse. I've double checked the other numbers and don't see any other issues. Of course I'm only going off of what was presented. Feel free to delete this comment after the fix has been made.

I do however want to thank you for posting this (and part 2) I appreciate your analysis and it leads to great/intelligent conversations. Keep up the amazing work.