The Dallas Cowboys had some injury problems last year, to say the least. This article on Adjusted Games Lost for 2013, from Football Outsiders, is an attempt to come up with a weighted value for the games a team loses to injury. I have a link below to an article that explains the methodology, but mainly it attempts to account for the impact of players missing games. It counts a starter missing a game as more significant than a backup, and it also tries to factor in the effect of players at less than full strength (those reported as doubtful, questionable, or probably on the weekly injury report who still play).
When I saw the results, they were not quite what I expected. Football Outsiders showed the Cowboys with a total AGL in 2013 of 67.9. (You want a low number). That placed them 17th in the league, which is of course pretty much where the Cowboys wound up overall in the NFL. The league average for 2013 was 70.2. The place the Cowboys wound up seemed, well, more average than I expected. By contrast, in 2012, Dallas had a seasonal AGL of 86.5, which was 28th, almost at the bottom of the league.
FO also breaks the numbers down into offense/defense and unit. There, it shows some results that fall more in line with what I anticipated. The defense accounted for 51.4 AGL, which is 75.7% of the team's overall figure, and puts the defense 29th in the league in this ranking. That again tracks very closely to how the defense performed as a whole.
The numbers on the defensive line reinforce our perceptions even more. Dallas' AGL here was 26.0, 31st in the league, and exceeded only by the Chicago Bears. Much of da Bears' number was attributable to Henry Melton's ACL injury, which is not a comforting thought for Cowboys fans. But this falls right in line with the frantic shuffling of bodies in and out of the line all season. That unit accounted for 38.3% of all AGL for the team, despite representing only 18.2% of the total positions on the roster.
Outside of that, Dallas only stood out in two other areas (FO only showed the top and bottom 8 teams for most subunits). Tight end had no AGL at all, which is a nice bit of data as Jason Witten is not getting any younger. And quarterback had a 1.1 AGL. That was 19th in the league, but as FO observed, the bad luck for the Cowboys was that it was mostly due to Tony Romo missing the last game of the season. The overall AGL for the offense was 16.6, sixth best in the league. Balance that with the defense, and once again, 8-8 seems to fall right in line with what you would expect if this has any validity as a predictor of team performance.
So the perceptions of the Cowboys were somewhat borne out by looking a little deeper into the numbers. The defensive line took the brunt of the injuries, the defense as a whole was much more beat up than the offense, and that correlates pretty well with the production of the respective units in 2013.
But that is not quite the whole story. There is another factor, and it not only changes the picture for the Cowboys, I believe, but also points out that there are times you have to be careful with using statistical analysis.
Because things weren't quite what I had expected, I dug a little into the way FO comes up with the numbers. The whole idea of AGL goes back to an article called, appropriately, "The Injury Effect". This discusses the first time league-wide AGL numbers were compiled, back in 2007, and is based on work done by Bill Barnwell in analyzing injury impact on the New England Patriots. This is the only place I can find a discussion of the assumptions and parameters for figuring AGL. It includes this:
First, the data: The NFL injury reports, from 2001 to 2006, as compiled by Football Outsiders intern Chris Povirk. Injuries were weighted with a simple metric that measured the effect of an injury: A player who was listed as "Out" or placed on IR was scored with four points for each week he was in that role; a player who was listed as "Doubtful" three, "Questionable" two, and "Probable" one.
Following that, we compared the injuries to what we determined were the team's 22 offensive and defensive starters heading into the season, its "projected" starters. For example, in 2006, although Tony Romo ended up being a huge part of the Cowboys' success, Drew Bledsoe was the expected starter going into the season. We focused specifically on their injury rates, since the majority of a team's salary cap, training camp, and focus is placed into those starters. This differs from last year's research, which looked at a team's 53-man roster equally. (Emphasis added.)
There is a possible disconnect for the Cowboys here. I have not been able to figure out if or how this accounts for players who are out for the year before the season ever starts. I have asked, but haven't gotten a response from FO yet. There are some obvious questions here:
- How was Jay Ratliff counted? Was he a starter (which was the projection) in their calculations, and how many games did he count?
- What about Tyrone Crawford? He was quite likely to have been the fill-in at LDE instead of George Selvie. It seems his absence should be reflected in the numbers.
- There was some expectation that Matt Johnson was going to claim one of the starting safety spots. He is almost certainly not accounted for in this system.
My guess is that Ratliff did count as a starter, and he and Anthony Spencer were almost all of the 26.0 AGL on the defensive line. But I am not sure about that. And I still think that this evaluation does not account at all for players that don't make it out of training camp. This happens to a lot of teams, of course, but once again, in 2013 Dallas was hit hard, especially on defense and along the defensive line. The FO numbers show that injuries hurt the team very badly there, and that was reflected in the final record. The offense had a good year, but when you average a good result with a bad result, you get mediocre, which is pretty much the descriptor for how Dallas has wound up the past three years.
The health of the team is perhaps more important than we think. This indicates that getting the Cowboys into the top quarter or so of teams healthwise would more than likely correlate to a playoff berth, based on the past couple of years.