Let's Party Like It's 1999

In 2012, according to Football Outsiders, the Eagles ranked 25th in offensive DVOA, 26th in defensive DVOA, and 23rd in special teams DVOA. That's remarkably bad. Some people, like Brent Cohen, have projected a big turnaround for the team this year that leans heavily on wishful thinking, Chip Kelly instant results, and large regressions to the mean.

I wanted to try to figure out what the chances of such a rebound season are, so I went through the Football Outsider archives and looked at every squad that, like the 2012 Eagles, finished 20th or worse in both offensive and defensive DVOA. Taking the overall adjusted DVOA for all 113 teams, I recorded how much it changed the next season.

What I found is that most of those teams did, unsurprisingly, improve. In fact, 79.6% of all teams in this group had better DVOA scores in year N+1, and 57.5% improved by at least 10 percentage points. The average delta was +11.9%.

This makes sense. Teams sink to the bottom of the heap for all sorts of reasons, including poor coaching, quarterback play, defensive failures, injuries, turnovers, and more. The Eagles had all of that. Just keeping the offensive line healthy, improving the defensive tackle rate, and recovering a few more fumbles should help this team get better. But the 2012 Eagles ranked 28th in overall team DVOA, at -22.4%. A +11.9% bump is average among the set I examined, but that only gets you to -10.5% -- good enough for a mere 25th place and only 6-7 mean wins on Football Outsiders' 2013 projections.

So if this year's Eagles improve at the average of their predecessor bad teams, we're still looking at a subpar squad. It's possible they improve more, though. 26.5% of all teams I looked at increased their DVOA by at least 20 percentage points. Beyond that, only 8.8% of teams jumped by 30 percentage points or more. That means there's approximately a 1/4 chance that the Eagles get back to what is essentially league-average form and less than 1/10 chance the team approaches the land of the top 10-15 NFL teams.

You can find hope or despair in these numbers. A (minority) chance to rebound to league average is appealing to root for on its own, although there's almost as large of a chance that the team's performance actually declines.

My takeaway is that improvement is likely, but a big swing isn't -- and that shouldn't be surprising. This is a long term rebuilding process, as Jeff Lurie indicated in his interview this Sunday with Zach Berman:

"I just think it's not in wins and losses... It's absolutely instilling a culture in the program that he brings to it, a sense of preparedness, a pride in being the best you can be for the fans and the team, and winning every day - winning the day, each day. And whatever happens, happens."

The Eagles probably won't win tonight in Kelly's debut. Nor are they likely to win many games the rest of the season. Lurie knows what's important is establishing a foundation for success going forward, the way Andy Reid did in 1999.

I don't remember much about the 1998 season (seeing as I was only 10 at the time), but the year the wheels came off the Ray Rhodes express was about as pretty as 2012. The 3-13 record reflected a horrible quarterback trio and 25th-or-worst DVOA ranks in all three phases.

Reid wasn't going to turn that mess around immediately. In 1999 he improved the team's overall DVOA +23.6% (from -34.6% to -11%). That's in the 23rd percentile of best turnarounds in the aforementioned group. The result was only a modest increase in wins, a 5-11 record -- but it laid the building blocks for what was to come.

With the important exception that the team doesn't have a quarterback of the future yet, 2013 is shaping up to be 1999 all over again, or at least I hope it does. Because wins matter much less today than that the losses fuel wins tomorrow.

Geno Smith and Expected Draft Value

I hope everyone is reading Brent Cohen's work over at Eagles Rewind. ​He's playing with stats and coming to interesting conclusions, especially with his PVM (Positional Value Multiplier) prospect rankings that add a discerning adjustment to the standard draft preparation.

​Brent's goal is to take the consensus draft prospect rankings and apply two main modes of adjustment. First is a risk factor that attempts to quantify how reliable different positions are. He shows that offensive tackles, for example, have a bust rate of only 26%, compared to 50% for quarterbacks and 66% for running backs. One can quibble with the metric (you might adjust for likelihood of Pro Bowl appearances instead of failure to start), but it's a good one to have. The second main adjustment is for positional value, i.e. how much one player is worth relative to another at a different position. Brent uses salary cap data to show a clear hierarchy, from centers and guards at the bottom to defensive ends and quarterbacks (far and away) at the top. Again, you can quibble with the exact numbers, but it's quite promising stuff.

The PVM rankings that emerge are interesting to look at. Certain players receive bumps -- especially quarterbacks. Others fall slightly, like guards and centers. The final result is not a radical departure from the original consensus, but that's not Brent's goal.

It is mine, though. I look at the numbers pulled, especially on salary cap figures, and it's obvious to me that there's a huge disparity in value between different positions. Brent's cap percentages are at right, and you can see that while the middle group is close, there are certain spots that vastly outweigh others. As I described above, the PVM rankings make an adjustment along these lines, but it doesn't come close to matching the discrepancies in value. In expected value terms, if you have two similarly-talented prospects -- one a center and one a defensive end, you should take the defensive end every time. In fact, if you take it literally, the center would have to be three times as good of a prospect as the end to warrant consideration. 

Maybe this is too extreme of an adjustment for your tastes. If you want something closer to the consensus prospect values, Brent's PVM should serve you well. But I think this more accurately represents the reality on the ground. Chance Warmack may become a Pro Bowl guard -- many scouts rate him as the best player in the draft. But there's a strong case to be made, based on this data, that he would have to be 1.5 times better than the next best cornerback, 1.8 times better than the next best defensive end, or 2.7 times (!!!) better than the next best quarterback.

Plenty of caveats apply here, including the risk factor that I mentioned above. And you may either disagree with the salary cap distinctions entirely or have a more refined model based on it. I also haven't run my own projections for using this data to provide an updated draft list.

But if you look at the draft through this qualitative lens, it's easy to make an affirmative case for a quarterback, even Geno Smith. At this point I don't expect the Eagles to draft him. Most scouts consider Smith the best quarterback prospect, but also suggest that he wouldn't be so in previous years. Average rankings seem to consider him about 20th overall. That doesn't seem like the type of player you want to use your top five pick on.

However, I've said before that you always take the quarterback if you think he's your guy and the data backs that up. A good starting quarterback is worth nearly 50% more than the next most valuable position, according to salary cap numbers (and one might argue that's understating it). It's importance towers over other concerns. That means that even if Smith has a worse chance than average to become a good starter, his expected value from that smaller probability is still high -- likely higher than many players who will be drafted ahead of him. Smith may not be the safe pick, but if the goal is to maximize value, he may be the right one.

​Photo from Getty.