Possibly the closest Super Bowl matchup in history also poses the question: how much does it mean when certain aspects of an NFL team improve dramatically in the second half of the season?
08 Mar 2013
by Danny Tuccitto
Apparently, I've picked up whatever bug it was that made Aaron have to communicate via grunts (translated by Rivers) and scribbles for 90 percent of the MIT Sloan conference, so apologies in advance for this being shorter than you might expect. Promise I'll make it up to everyone in the second post, which will be arriving in this internet space next Tuesday.
With that bit of groveling out of the way, it's time to once again present Football Outsiders' Adjusted Games Lost (AGL) totals for the recently completed season. For any new readers out there, AGL is our metric that quantifies how much teams were affected by injuries, and is based on two fundamental ideas: (1) Injuries to starters, injury replacements, and important situational reserves impact a team more than injuries to bench warmers; and (2) Injury effects should be adjusted for whether or not a player was out, doubtful, questionable, or probable. Although this should go without saying, higher AGLs are worse than lower AGLs.
Below are the AGL totals and rankings for 2012. For comparison, I've also included the 2011 totals.
|Team||2012 AGL||2012 Rank||2011 AGL||2011 Rank|
|Team||2012 AGL||2012 Rank||2011 AGL||2011 Rank|
The average team AGL this past season was 64.5, which was up slightly from 2011 (60.0), and therefore qualifies as the highest leaguewide average in our database. Put differently, an average NFL team these days suffers the injury equivalent of four important players missing the entire season. That's over one full season more than what the expectation was five years ago (45.5). However we interpret it, the bottom line is that 2012 continued an upward trend we discussed in last year's piece.
Another leaguewide trend that continued last season was the year-to-year AGL correlations, for which we're still trying to come up with a suitable explanation. From 2008 to 2012, team AGLs have been more consistent than they were from 2002 to 2007. During that earlier period, previous year AGL explained less than one percent of the variation in current year AGL; since then, it explains about 10 percent on average. Of course, this means that 90 percent of it doesn't carry over from one season to the next, so we shouldn't get carried away by that result. It does, however, seem reasonable to conclude that AGL totals have gotten more predictable of late.
Something else readers might find interesting from a predictive perspective is how well total AGL has correlated with wins and total DVOA over time. Here's a table for you:
|Year||r(Wins)||r(DVOA)||r(Win Diff)||r(DVOA Diff)|
To situate ourselves, the first column of correlations is current-year AGL with current-year wins, and the second column is current-year AGL with current-year total DVOA. The final two columns involve correlations between year-to-year AGL change, year-to-year win change, and year-to-year total DVOA change.
Starting at the bottom of the table, we see that for the 2002-2012 period, the change in AGL from year to year does a better job of explaining why a team won (or lost) more games compared to the previous season than current-year AGL does explaining win totals in the current year. The same goes for the relationship between AGL and DVOA.
In that context, 2012 turned out to be a year in which injuries played a much larger statistical role than normal, especially with respect to how teams performed relative to 2011; case in point: the St. Louis Rams. Last year, we mused about reasons for St. Louis' consistently awful AGL, which averaged a ranking of 28th from 2007 to 2011. Enter Jeff Fisher, who had previously been associated with a Tennessee Titans team that was consistently healthy, ranking third on average from 2007 to his departure in 2010. Could there really be a Jeff Fisher effect? I highly doubt it, primarily because, in addition to hiring Fisher, St. Louis purged their roster last offseason. I'd guess that, as a general rule, if a team gets rid of a bunch of injury-prone guys, it's not going to be as injury-plagued.
A few other teams seemed to benefit from year-to-year AGL improvements. The San Francisco 49ers avoided injuries for the second straight season under Jim Harbaugh, which might help explain why they weren't bitten by a projected regression to the mean in 2012 -- despite all statistical indicators to the contrary. San Francisco's primary NFC West rival for the foreseeable future, Seattle, rode a wave of improved health (27th AGL in 2011, fourth in 2012) to four additional wins and the second-largest DVOA improvement in the NFL (39.8 percentage points). Admittedly, Russell Wilson had something to do with it, too.
On the other end of the spectrum, several teams saw decreases in performance this past season track alongside changes in AGL. Green Bay was the most injury-riddled team in the NFL last season -- only the sixth 100+ AGL in our database by the way -- and that drop from 16th to 32nd coincided with four fewer wins. Green Bay's division-mate Detroit suffered a similar fate, as they won six fewer games in 2012, thanks in some part to a 20-spot drop in the AGL rankings. In the AFC, the New York Jets fell from fifth to 21st in AGL, and also saw their total DVOA drop 31.4 percentage points from 2011 to 2012.
There were a few trends worth mentioning related to the shenanigans of injury reporting. First, as we detailed last year, "questionable" is about as detached from the reality of injury reporting guidelines as "offensive pass interference" is detached from the NFL rulebook. Continuing a trend going back to 2008, 69.3 percent of players listed as questionable on the injury report actually played that week, which was nearly 20 percent more often than the 50 percent participation rate questionable is supposed to represent.
Speaking of questionable, the second bit of shenanigans we noticed was Bill Belichick's shift in diversion tactics this season. You'll recall that, for much of the previous decade, Tom Brady was listed as probable with a shoulder injury. Well, in 2012, Belichick didn't list him one time, instead opting for a game whereby seemingly the entire New England Patriots roster is questionable. Excluding New England, the average number of questionable designations for an NFL team over the course of the 2012 season was 21.6, with the Arizona Cardinals having the second most at 45. Bill Belichick listed a player as questionable 138 times.
One person in the league who refuses to be outdone by Belichick, however, is Rex Ryan, who apparently seemed content filling the "probable" void vacated by his arch-nemesis. In 2012, the New York Jets led the NFL with 164 probable designations when the average of the other 31 teams was 56.6.
To close, here's one final injury reporting anomaly -- to put it mildly -- that comes courtesy of the practice participation reports teams are now required to provide several times leading up to their games. If we only look at players who contribute to AGL (starters, injury replacements, and important situational reserves), and focus on those listed as doubtful on the injury report, we find the following. If the player did not participate in the final practice of the week, he had a 0.8 percent probability of playing. However, not a single doubtful player who fully participated in practice played that week. And what's more peculiar is that, if the doubtful player was "limited" in practice, he played that week 13.3 percent of the time.
That's all for now. Next Tuesday, I'll detail 2012 AGLs for team units.
25 comments, Last at 21 Mar 2013, 9:36am by Dean