Since we’re in the midst of the off-season, between the two major team-construction events, I thought it’d be a good time to explain my personal philosophy on what it takes to win a Super Bowl.  The goal is to try to dig down to the most basic attributes a team must have, and in the process examine clichés like “defense wins championships” or “you need an elite QB”.

Here’s where I am currently:

To win the Super Bowl:

– You need a lot of good players.

– You need some of those players to play great (greatly?) in the playoffs.

– You need to get a few lucky bounces along the way.

That’s my current theory at least, I’ll update it as I find new data that conflicts.  Today I’ll give my thoughts on the QB position and then examine the whole Offense vs. Defense wins championships thing.

Overall, I’d like to identify a set of necessary conditions for winning the Super Bowl.  Today is the first step in that endeavor.

First I’ll attack the QB position:

You do not need an elite QB to win the Super Bowl.  It certainly helps, but the fact is you just need a “good enough” QB.  The entire “elite” debate is ridiculous, but especially at the QB position.  When you use the term “elite”, you’re alluding to a very exclusive group.  Is the top 10% fair?

If so, that means the number of “elite” starting QBs in the NFL is between 3 and 4.  So congratulations Patriots, Packers, and Broncos; one of you obviously won the Super Bowl last year.

The fact is, the term “elite” is useless since there’s no standard definition, and even if there were it wouldn’t mean anything significant.  Teams need a QB that is “good enough”.  So what does my list look like?

Good Enough – Brady, Manning (both), Rodgers, Brees, Ryan, Roethlisberger, Flacco, Schaub, Romo (i hate him but yes), Rivers, Kaepernick, Luck

That’s my list, although I’m sure there is some disagreement.  Also, this is not to say a QB not on that list CAN’T win the SB, just that, at this moment, I don’t believe they’ve proven themselves good enough.

Joe Flacco is not an “elite” QB, by any realistic definition.  However, that doesn’t matter, he still won the Super Bowl.  Colin Kaepernick came damn close.

The problem for Eagles fans is that neither Vick nor Foles looks good enough this year.  It’s possible Foles could get there, but extremely unlikely this season.  I’m pessimistic about Vick’s chances.

Does Defense Win Championships?

Well, what do you think?  Now consider:

– Over the last 10 seasons, 5 SB winning teams ranked in the top 7.5% for points allowed per game.  Put differently, the top 24 out of the last 320 team seasons contained half of the Super Bowl winning teams.

This is, perhaps, the most surprised I have ever been by a stat I’ve looked up while writing.  In any case, it would certainly appear as though having a truly great defense goes a long way.

What about offense?

Based upon the above stat, I would expect to find a similar result (i.e. be great at something, either offense or defense).  Looking at the same sample:

– Just 1 SB winning team (New Orleans ’09) in the last ten years placed in the top 7.5% of points scored, or 92.5 percentile.

If we look at the top 10% of offenses over the last 10 years, just one more SB champ makes the cut, the 2004 Patriots.

Put together, it seems like having a “great” defense is preferable to having a “great” offense.  That definitely has to change the overall theory a bit.

What about the reverse perspective?

– The lowest ranked offense to win the Super Bowl in the last ten years was the 2008 Steelers, who averaged just 21.7 ppg, good enough for exactly 160 out of 320, or the 50th percentile.

The 2003 Patriots are just above them, having averaged 21.8 ppg.  Interesting to note that the Patriots (now offensive juggernauts), won their first title with a fairly weak offense.

As you might have expected, both the ’03 Patriots and the ’08 Steelers had historically great defenses.   Both placed in the top 10 of the last 10 years.

– The lowest ranked Defense to win the SB in the last ten years was the ’11 NY Giants, who allowed 25 ppg (while scoring just 24.6, leaving them the only SB winning team with a negative point differential, I really hate the Giants).

That places them 265th out of 320 teams, or 17th percentile.

– Among SB winners of the past 10 years, the 2007 Giants are the lowest ranked offense that did NOT have a corresponding historically great defense.  They scored 23.3 ppg, ranking the team 120th out of 320, or the 62.5 percentile.

I feel like we need some charts, so here are two, one showing points allowed vs. Wins and the other showing Points Scored vs. Wins.  Note this is just regular season data.

Both illustrate strong correlations.  The values:

– Points Scored to Wins:  .768

– Points Allowed to Wins: -.669

Have I confused you yet?

If not, good for you, because I’ve certainly confused myself.  Let me attempt to summarize what we’ve seen here today:

– A GREAT defense goes a long way, but is NOT a necessary condition to winning a SB.  Neither is a GOOD defense, as evidenced by the ’11 Giants, ’06 Colts, and ’07 Giants (all placed in the bottom 50% yet still won the SB).

– A GREAT offense does not go as far, but a GOOD offense does seem to be a necessary condition for winning a SB.  No offense ranked in the bottom 62.5% for the last ten years has won the SB.

What does it all mean?

While teams should focus on being good everywhere, it appears as though Offense should take slight preference, as the only way to win without a very good offense is to have a historically GREAT defense, which is obviously harder to put together.

Or you can be like the NY Giants and just ride a wave of lucky bounces…twice.

Yesterday we did the defense, so naturally today we will do the offense.  However, this is a little bit less informative due to the injuries the team sustained last season.

I’m going to skip the original chart I made because it was based on perceived ability (not performance), and included all of the injured starters.  Using PFF’s grades, this is how the offense looked last season.

A couple of notes, then I’ll highlight a few areas.  Due to the injuries, there were a number of players that received playing time.  I used whichever player took the most snaps as the rule for inclusion.  Watkins and Scott played nearly the same amount, but they ranked similarly so the color wouldn’t change.  There were also a few players who were on the edge as to which color they deserved (for instance Shady was the last spot in the top third).  I could have done the mixed colors again, but I wanted it to be as clear as possible.

Takeaways-

– Both Jeremy Maclin and D-Jax played well below their potential.  Perhaps not surprising given the QB situation and the state of the O-Line, but it remains a cautionary note going into next season.  We know Maclin has the talent, but we can’t pretend he’s a lock to realize that next year.

– Jason Peters will hopefully come back as green.

– I know everyone loves Jason Kelce, but PFF graded him close to the bottom for his rookie year.  I think he’ll be good, but this is another position at which fans seem to be buying the upside without realizing there’s a reasonably significant chance it doesn’t work out.  For the record, his rating in 2011 would have placed Kelce LAST in 2012, below Dallas Reynolds.

– The right side of the line is tough to project.  If the Eagles draft a RT at #4 and shift Herremans back to RG, then both of those spots could go green.  Keep that in mind as we get closer to the draft.

Drafting a starting RT could provide immediate improvement at TWO positions.

– Although I don’t need to mention this, having your QB grade out red is bad.  Of the 27 QBs that played more than 50% of their team’s snaps, just 4 ranked worse than Vick (Ponder, Fitzpatrick, Sanchez, Weeden).

– It will be interesting to see how James Casey and Brent Celek split time.  I know Kelly says he will use a lot of 2 TE sets, and Casey can line up almost anywhere, but it wouldn’t surprise me at all to see Casey supplant Celek as the starting TE.

– Finally, PFF’s ratings shouldn’t be taken as gospel. So feel free to disagree with the colors above, just note that the source is Pro Football Focus, not me.

That’s all for today.  Working on some bigger stuff that takes time, but the results should be fun.

Now that we’re through free agency, I think it’s time to take another look at the defensive line-up.  Previously, I put together an illustration of the Eagles 2012 starting defensive line-up, with each position color-coded according to performance/ability.

Today, I’ll put the same illustration together, this time with a the scheme/roster additions reflected.  I think doing this visually makes the comparison much easier to see.  Additionally, I’ll go back an update last season’s using PFF’s ratings in an attempt to take some of the subjectivity out of the ratings.

This is an attempt to make roster evaluation as simplistic and easy as possible.  Red = Bad, Green = Good.

Please note that this (the color-coding) is obviously subjective (hence the PFF adjustment in the second chart).  My guess is there are a few players for whom there will be disagreement over the corresponding color rating.

Here is the original graphic of last year’s defense, not updated with PFF:

For reference, Green = above average starter, Yellow = league average starter, Red = Below league average.  If I could go back and change anything, it’d be the CB ratings.  I was far too generous to both Nnamdi and DRC.  Using pure performance (not potential), they both should have been red.  For those paying attention, that means the ENTIRE Eagles secondary would be rated below league average (which sounds about right in hindsight).

Now let’s use PFF’s ratings.  If a player was in the top third of the league (starters), they’ll be green, middle third = yellow, bottom = red.  This chart will be a reflection of ONLY total performance, so no adjustments made for inconsistency (see multi-colored ratings above).

Note that Nate Allen and Akeem Jordan have replaced Anderson and Chaney.  At the time I did the last one, Anderson and Chaney were the starters.   However, over the whole season Jordan and Allen played more snaps.

Looks like the original ratings were pretty good, except for the previously mentioned CBs.  Additionally, while Kendricks had runs of both very good and bad play, his overall rating was very poor.

As you can see, there was far too much red and not nearly enough green last year.  It comes as no surprise then, that to date, most roster changes have been made on defense.

Now comes the hard part, trying to project next year’s team.  First a couple of rules:  Injury-risk will not be taken into account.  This is obviously a major concern, but I don’t see any good way to incorporate it into the graphic.  Therefore, I’ll projected the “on paper” defense, then highlight the risks afterwards.  Also, since we don’t know the exact defensive alignment Kelly will use, I’ll stick with a basic 3-4.

Keep in mind that the actual physical alignment will be different from what’s shown above. However, I think right now this is as good a guess as any in terms of personnel.

What jumps out?

– Phillips, Chung, and Fletcher can all be considered “injury risks”.  That’s a big chunk of the secondary (as well as the source of major improvement over last year).  If any or all of those guys go down, the Eagles are back in the red.

– The DL has gone from a strength (relatively speaking) to a weakness.  Cox is a keeper and should anchor the line (I don’t see him having an issue transitioning to the 3-4, regardless of whether he ends up as a 3-tech or 5-tech.)  However, Spooky is just a body and can’t be counted on for even league-average play.  Also, the opposing DE spot is a glaring hole on the roster (hence all the Star/Floyd draft talk).  I think there’s a good chance they address this with one of those prospects.  It’s also possible Vinny Curry puts on some weight and takes a shot at this spot.  In any case, as of this moment, it has to be viewed as a weakness.

– As I mentioned after their signings, the Williams/Fletcher combo is going to take some getting used to (Fletcher may not even win the job).  The hope here is for league-average play, nothing more.

Overall, the D looks a lot different from last year (personnel-wise).  However, it’s clear we still need a couple of impact players (i.e. green).  Maybe Barwin rebounds and becomes one of them.  Maybe Cole or Graham transitions better than projected.  Maybe Kendricks flourishes at his new spot and finds a consistent game.  Hopefully the team can use the draft to upgrade from red to green either at NT or DE.  That’s a lot of unknowns.

Keep in mind that it’s easier to go from bad (red) to average (yellow) than it is to go from average to good (green).  It’s clear that Roseman and Kelly looked at the second chart above (the PFF adjusted one) and tried to address every red mark (weakness).  On paper, they’ve been successful.

Even so, as of this moment, the Defense is still pretty far from one we can safely project as “good”.  Though after last season, I’ll be happy with just average.

There’s a lot of disagreement about how much of a role luck plays in turnovers, and fumbles in particular.  I’ve addressed this in more detail before (search fumbles for the posts), but today I just want to throw out some quick data to give a better illustration of the overall points I’ve been trying to make.

First, we can mostly agree that fumble recovery % is luck-dependent.  Not all fumbles are created equal, and where they happen can effect which team recovers.  However, I think it’s fair to say that, in general, teams should recover around 50%.

Last year the Eagles recovered 35% (according to Teamrankings).  Buffalo was worst in the league at 30.6%.  Washington led the league at 67.39%.

I previously looked at the persistence of fumble recovery rate and, as expected, found no meaningful correlation.  While that’s not enough to prove recovery rates are entirely luck, it is enough to show that the Eagles are very likely to improve on that measure next year.

Fumble Differential

We’ve talked about TO Differential, but now I’ll look at pure fumble differential (gained-lost).

– The 2012 Eagles’ fumble differential was -17, or 2nd worst in the last 10 years.  Obviously, the NFL average is 0.

– There is no evidence of persistence in fumble differential.  That means, despite how bad the Eagles were with fumbles last year (-17), that has no predictive power for what that measure will be next year.

– Something to remember here is that the differential also includes fumbles forced and recovered.  The Eagles gained just 5 last year.  The NFL average is 11 over the past 10 years.  The lowest over that time was 3, not far off from the Eagles’ 5.

So let’s back away from the luck aspect for the moment (that takes a lot longer to address, particularly regarding “fumble-prone” players).  What I’m basically say is this:

– The Eagles are very likely to fumble less next year.

– The Eagles are very likely to force more fumbles next year.

– The Eagles are also likely to recover a greater % of all fumbles next year.

Regardless of where you stand on luck vs. skill, I’d be surprised if many people disagreed with any of those three statements after looking at the data.

Put those together and the Eagles are very likely to have a significantly improved fumble differential, which obviously means the team is likely to have an improved overall TO differential.

Note: Nothing is guaranteed here, we are just talking about probability.  Just because the team was unlucky last year doesn’t mean it can’t ALSO be unlucky next year.  That’s possible.

The key here is that there appears to be extremely LITTLE room for the team to get worse (because of how bad they were), hence they will likely get better.

There’s also the issue of interceptions, which I haven’t really addressed.

I’ll try to tackle interceptions sometime soon, but I wanted to be very clear about the overall ideas here.  Even if you believe fumbles are not inherently luck-dependent, the Eagles are still very likely to improve next year.

I want to make sure we’re not losing the forest for the trees here.  I’ll tackle the luck vs. skill issue in more detail later and provide further evidence for why I believe fumbles are heavily luck-dependant.

BTW, this post is the result of a very thoughtful email from someone who raised a few issues with my previous post.  As I’ve said before, I make no representations for being perfect, so critiques are both welcome and necessary.  Please email me at  eaglesrewind@gmail.com if you have any issues with posts I put up.  I’d rather have the chance to fix mistakes or explain assertions than have individual monuments to my stupidity preserved for posterity.

NOTE: After today’s post (and recent analyses), I believe it’s time to change the Blog subtitle.  Hard to believe, but I think I’m now the most positive Eagles blogger I know of. Thank god for data and objectivity.

Today let’s go back to the turnover issue.  As most of you know, the Eagles were terrible last year when it came to “taking care of the ball”.  The team had a turnover differential of -24, which is tied for LAST place over the previous 10 seasons.

I repeat, LAST over the previous 10 seasons. (The Chiefs of last year were equally bad).

With that comes the previous analysis I did showing the correlation between TO Differential and Wins.  The value is a very significant .64.  So turnover differential is really important and the Eagles were historically bad last year by that measure, hence a 4 win season.

Now let’s take it one step further and see if we can learn anything about what might happen THIS year.

I did a similar analysis, but this time focused on the CHANGE in both turnovers and WINS from one year to the next.  Given the .64 value above, teams making improvements to TO differential year-over-year should see a corresponding improvement in Wins (or losses is TO differential gets worse.)

Here’s the chart, with Win Change on the X and TO Differential change on the Y:

The correlation value is .63, about what we would expect given the first order Wins/TO Differential correlation value.

The chart is a little tough to read, so let me make it simple with an example:  Let’s say your team’s TO Differential FELL by 10 this coming season.  To find the corresponding win change, find -10 along the Y axis and then move left or right until you hit the line.  Here, a decline in TO differential of 10 equates to an approximate 4 win decline.  HOWEVER, notice the spread is relatively large; this means we need to recognize that our level of precision (confidence) here isn’t very big.

So what does this mean?

Well let’s go back to the 2012 Eagles.  The team’s TO Differential was -24.  As I’ve discussed before, this was heavily influenced by incredibly bad luck with fumbles (in both their existence and the team’s recovery rate.)  I expect both measures to improve dramatically next season,  purely because last year’s were so unlikely.

Let’s say the Eagles move from the 99% in TO differential (negative perspective) to just the 90% (Using the last 320 team seasons/10 league years as data).  That equates to a TO differential of -13.  That’s still really bad (hence the 90%), so we’re not really doing the Eagles any favors here.

However, because last year’s measure was so bad, moving to -13 is still an IMPROVEMENT of 11 turnovers.

Here’s where it gets fun.  Going back to our chart, an improvement of 11 in TO differential equates to an approximate gain of 5 wins.

Presto chango, 4-12 last year just became 9-7 next year.  

Unfortunately, there’s that pesky little precision problem, which means a 5 win gain is far from certain.  However, the overall point remains: by just being BAD instead of HORRIBLE when it comes to TOs, the Eagles can achieve a much better record.

It must be noted though, that the Eagles achieving 4 wins last year with a -24 TO differential is a minor miracle in itself.  The team’s “overperformance” last year means there’s probably a little less room for adding wins due purely to positive mean regression next year.

Now let’s get really crazy (prepare your happy face):

Over the last 9 seasons (not including this past year since we don’t have subsequent season data), 17 teams registered a TO Differential of -15 or worse.

The average change in TO differential the following season?

+18.4

Conclusion:

a) The Eagles are likely to improve their TO performance substantially next year.

b) Positive improvements in TO differential are STRONGLY correlated with positive Win changes.

A + B = The Eagles have a significant chance (nothing is certain) to be A LOT better than you think next year.

Surprise news from the Eagles yesterday: the team signed a punter (Jones) and cut McBriar.  For most casual fans, this move probably doesn’t register as anything other than replacing one guy most people don’t know with another.

However, it has the potential to make a big impact, and put in context with the other moves the team has made, shows a clear recognition of one of the team’s biggest weaknesses last year:  special teams.

Though it flies beneath the radar of most casual fans, special teams is indeed important, and the Eagles were terrible in this area last season.

Let’s look at some stats:

– The Eagles ranked 28th in the league in Net Yards per Successful Punt last year with 37.3.

– The 49ers and Saints were tied for 1st overall with 43.2, meaning these teams, on average, gained SIX more yards per punt than the Eagles did.  While most of the league didn’t perform nearly as well, this is an indication of just how big a role the punting game can play (6 yards per punt difference is HUGE in case anyone didn’t realize that).

– According to Football Outsiders, the Eagles ranked 24th in the league last year in overall special teams.  Only 2 teams ranked in the bottom third of the league made the playoffs (Redskins and Texans).

– According to Football Outsiders, the Eagles ranked LAST in the league in punting last year.

– The Eagles were also UNLUCKY, with the fourth largest special teams disadvantage due to factors outside the team’s control (opposing field goals, kickoff distance, punt distance).

So what?

Looking at the above stats gives us an appreciation for how weak the ST unit was last year.  That last stat is interesting because it fits what I’ve been saying/showing for a long time:

The 2012 Eagles were a mediocre team with VERY bad luck.

While the offense and defense are more important, it’s nice to see Roseman and Kelly taking improvements everywhere they can find them.  There is no guarantee that Donnie Jones will be any better than McBriar was or that the coverage will be any better either.  However, it’s likely that both areas will improve, due to the low bar set by last year’s team.

The upshot is field position.  I’ve covered this before, but the Eagles ranked last, by a long shot, in relative field position last year (-6.67 yards).  Improvements to special teams will help this measure.  Consequently, the offense will have fewer yards to travel to get in scoring range and the defense will have more yards behind them.

Field position is an aspect of the game that is tough to focus on in real-time.  6.67 yards does not sound like much, but remember that is an AVERAGE PER DRIVE.  Last year, every Eagles offensive drive started, on average, almost 7 yards farther from the end zone then the opponent’s corresponding drive.

While everyone is focusing on the defensive scheme and roster changes while getting excited about the offense, arguably the easiest area for improvement is special teams.

The signings of Jones, Benn, Chung, Phillips, etc… will go a long way in that area.  I don’t expect fans to get excited over things like kick coverage and net punting average, but improvements here will show up in the box score by making it easier for the Eagles to score and harder for opponents to do the same.

Let’s step away from the draft and free agency for a moment, since we’ve reached a lull in the action there.  Instead, let’s talk about expectations in general, specifically how team performance this past season should factor into our predictions for next year.

I’ve noticed quite a few people are “down” on the Eagles for this coming season.  Nevermind that the draft has yet to happen and free agency is not yet finalized; generally people do not believe the Eagles can be a playoff team next year.  This shouldn’t come as news to anyone, but the real question is why?

I believe the answer is composed of two related parts: A) anchoring, and B) misunderstanding of the overall NFL competitive landscape.

Anchoring

Anchoring is a fairly common problem than many people don’t fully appreciate.  It’s a specific type of cognitive bias that results in poor evaluations/predictions from almost all of us.  Basically, it’s relying too heavily on the first piece of evidence.  A common example is any game that involves guessing how many pieces of something are in a jar (candy usually).  If the guesses are public, then the overall range of guesses will usually center around the first person’s guess.  Despite the fact that everyone is looking at the same jar (has the same info), the initial guess, by virtue of being first, tends to “anchor” all subsequent guesses, regardless of its accuracy.

While football records are obviously a bit different, I submit that the most recent season’s record fulfills a similar function as the first guess in the example above.  Jumping to the Eagles: last season the team won just 4 games.  Consequently, predictions for next season  are all made in reference to that number (which is why predicting the playoffs for the Eagles next year isn’t common).

Despite the fact that last year’s team roster/scheme/coaching staff/etc… bears little resemblance to the one the Eagles will play next season with, the record remains a major factor.

So the obvious question for us is: how useful is last season’s record in predicting this coming season’s?

Taking the last ten seasons for each team, I plotted season records against the same team’s record from the following year.  Here’s the chart:

There definitely appears to be correlation (as we’d expect), but the value is just .30 (weak/moderate).  Additionally, I’d argue that within this sample are two teams that have an inordinately large impact on the overall correlation values by virtue of their remarkable consistency (the Patriots and Colts obviously).  While I’m usually hesitant to play with the sample, in this case I think it’s worth looking at the results without these teams.  In essence, due mainly to the Brady/Manning effect, I think they are anomalous and a poor representation of the usual state of the league.  More specifically, I think they are useless when it comes to drawing conclusions about the Eagles this year.

When we remove them, the correlation value drops to .22.  Still real but fairly weak.

So far, we’ve got reasonable evidence that the Eagles record this year doesn’t mean much when it comes to next year’s performance.  I’d actually argue that it has almost NO SIGNIFICANCE, due to the overall weak correlation combined with the dramatic makeover the team has undergone.

Average Change

We can also look at the average change in record from year to year.  Using the same sample, I found the absolute value of the difference in wins from one season to the next.

The verdict?  On average, NFL team’s # of wins change by 3.07.  With just 16 games in a season, a 3 win difference in either direction is BIG.  Also, that’s AVERAGE, which means there were a LOT of changes greater than 3.

In fact, the median value was 3, which means half of the season-to-season changes were GREATER than 3 wins/losses.

Upshot

This obviously has large implications for the Eagles, and more specifically Eagles fans.  A change of 4 wins (slightly above average) puts the Eagles in playoff contention (or a winless season).

To be clear: All fans are free to COMPLETELY IGNORE last season’s record, for it is a very weak indicator for how the team will perform this coming year.

At the moment, I’m fairly bullish on the Eagles chances for next year, though there is still a long way to go.  In any case, the fact that the team sucked last year really doesn’t matter, so by all means, forget it ever happened.

P.S. you’re welcome.