Building a winning NFL team isn't like any other construction project.
NFL executives have a budget, a crew, equipment and raw materials—but the wrecking balls start swinging the instant they're done; all they can do is hope more of their building survives 17 weeks of destruction than any of the other 31 buildings on the same block.
A year after they started, they salvage what they can from the wreckage and start all over again.
In today's NFL, the salary cap is stubbornly flat, making budgets tight around the league. There's only so much cap money to go around, and the average annual value of big-money contracts seems to go up with each new contract extension.
NFL executives can't just try to get good players at every spot on the roster; market prices and the salary cap force them to approach roster-building with a clear philosophy. Which positions will be strengths? Which positions will have depth? If any positions have strength and depth, where will they skimp?
Surveying today's NFL landscape, is there an ideal roster-building philosophy? Go big on a quarterback? Build a smothering defense? Hoard dominant playmakers?
Is there a way to invest your team's resources that will give you an advantage over everyone else?
Offense vs. Defense vs. Both
The first question to answer is whether it's best to invest more heavily in offense, defense or strive to be balanced.
Obviously, if we're waving magic wands, it's best to have a great offense and a great defense, but we're not waving magic wands—we're writing checks. Let's take a look at the 2012 playoff teams and see how they built their mousetraps.
The surest sign of strength in the NFL is points differential: The total points a team's offense has scored, minus the total points its defense has allowed. In 2012, 11 of the 12 playoff teams were among the top 13 teams in points differential.
Only the No. 6 Chicago Bears and No. 9 New York Giants finished in the top 13 in points differential and missed out on the playoffs; only the No. 21 Indianapolis Colts were outside of the top 13 and made the postseason.
Organized from left to right, we see the Patriots with the NFL's highest-scoring offense and No. 9 scoring defense. Not surprisingly, they had the NFL's best scoring differential. Second best in scoring differential were the Denver Broncos, who had the No. 2 offense and No. 4 defense.
Then the pendulum swings back toward defense, with the Seahawks, 49ers and Falcons finishing third through fifth in scoring differential. Those three teams all had top-five scoring defenses, but they finished ninth, 11th and seventh in scoring offense, respectively.
The Washington Redskins were an interesting case. They finished with the No. 4 scoring offense at 27.2 points per game, but 22nd in scoring defense, allowing 24.2 points per game. The 49ers had the highest disparity the other way, with the No. 2 defense and No. 11 offense.
Outside of the Redskins defense and both units from the Minnesota Vikings and Indianapolis Colts, every playoff team's offense and defense finished in the top 12 of the league. Here's the table feeding that chart.
|SRS||O PPG||O Rank||D PPG||D Rank||Diff||Diff Rank|
|New England Patriots||12.8||34.8||1st||-20.7||9th||14.1||1st|
|San Francisco 49ers||10.2||24.8||11th||-17.1||2nd||7.8||4th|
|Green Bay Packers||7.3||27.1||5th||-21||11th||6.1||7th|
Pro Football Reference
Unfortunately for NFL executives, it looks like the answer is "both." Even teams with exceptional offenses or defenses are at a big disadvantage if the other unit isn't at least above average.
How Do You Get Both?
Leaving $103 million for the rest of the pot sounds like plenty, but signing one player to a $20 million contract drops the average salary available per roster spot from $2.32 million to $1.98 million. That's a $340,000 gap across 52 roster spots in a league where the minimum rookie salary is $405,000.
Many fans love the idea of forgoing the big quarterback and spending that money bolstering the running game and defense. When the 2000 Baltimore Ravens won a Super Bowl with unheralded quarterback Trent Dilfer at the helm, they proved this model could work.
The problem: In order to pull that off, the Ravens didn't only need one of the best scoring defenses of all time, allowing just 10.3 points per game, they also needed a top-five running game to muster a scoring offense that ranked 14th in the NFL.
In order to make up for the drag a replacement-level quarterback puts on a team, the rest of the team has to be truly exceptional.
The quarterback market is extremely efficient. As this chart shows, only Joe Flacco's monster deal (born of a negotiation where he had unprecedented leverage) and Tom Brady's intentional discount vary from a tight correlation between performance and pay.
There's a reason teams are spending so much on top quarterbacks; they do wonders for the rest of the team. Avoiding interceptions is the biggest part of maximizing a team's turnover margin, and turnover margin is a huge driver of success, on both sides of the ball. Offenses that avoid turning it over turn more drives into points, and defenses that take the ball away stop other teams from scoring.
Just look at the correlation between points differential and turnover differential.
For those of you who aren't stats geeks, that R-squared value is what's called the "coefficient of determination," and it goes from zero (a quarterback's thrown interceptions have no effect on point differential) to one (thrown interceptions are the sole factor in determining point differential).
An R-squared of 0.4683 is a very, very strong correlation, and it's about twice as strong as the still-significant R-squared of defensive interceptions to point differential (0.2354).
Fumbles are part of turnovers too, but it's well established that fumble recovery rates are all but completely random, and forced fumbles not much less so. Football Outsiders even publishes an annual summary of fumble luck and highlights extremely unlucky and lucky teams due for a regression to the mean.
Investing money in defense hoping to snag more turnovers won't bear as much win-loss fruit as investing money in a quarterback who'll make plays without turning the ball over much.
Protecting Your Investment?
Let's say you have your franchise quarterback. (In which case, congratulations! He'll make you look smart for as long as your checks clear.) Is it better to maximize his performance by protecting him with quality pass-blocking or surrounding him with weapons?
It's awfully hard to isolate quality quarterback play from quality receiver play. It's also hard to isolate quality quarterback play from quality offensive line play. Pro Football Focus team offense stats (subscription required) can get us close, though.
Let's regress 2012's team-adjusted net yards per attempt (a great one-number passing effectiveness stat that incorporates touchdowns, sacks and interceptions) against Pro Football Focus team "Pass" grades (incorporating the play of the quarterback and all pass-catchers).
As with the correlation between thrown interceptions and point differential, we see a huge coefficient of determination: 0.4775. If you have good quarterback play and receivers who get open and make plays, you will move the ball effectively, with more touchdowns and fewer interceptions.
Not exactly rocket science, but it confirms that PFF's subjective grading system correlates strongly with statistical reality.
Now let's look at the same adjusted net yards per attempt, this time regressed against PFF's team pass-block grades.
Again, we have a significant coefficient of determination (0.197), but not nearly as huge as the connection between passing game production and quality quarterback/receiver play.
This makes intuitive sense.
The Carolina Panthers have the fourth-best PFF team pass-block grade (subscription required) so far in 2013, but they aren't getting great receiver play, and quarterback Cam Newton got off to a slow start. The Detroit Lions are getting surprisingly solid play from their patchwork offensive line—but have invested huge money and draft picks getting Stafford weapons.
Both quarterbacks are former No. 1 overall draft picks of surpassing physical talent a year apart in age, and they are experienced enough to have ironed out any rookie luck, good or bad.
So far this season, though, Newton's adjusted net yards per attempt is 6.25, ranked 15th in the NFL; Stafford's is 7.30, ranked seventh.
Of course, the best way to maximize your cap dollars is to get talent at less cost than it's worth.
The Lions rolled with depth veterans and rookie Larry Warford at three of the five offensive line spots. Per PFF, they're getting better pass-blocking (subscription required) than teams like the St. Louis Rams and Chicago Bears, who spent big free-agent money on veteran linemen.
The word "Moneyball" has jumped the shark, according to The Washington Post's Lydia DePillis.
She may have a point when it comes to folks slapping the word "Moneyball" for any kind of statistical optimization. Its original use still holds water, though: Sports teams can beat their competition by getting worthwhile production from players they didn't pay free-market price for.
As the Washington Redskins found out between the time owner Dan Snyder bought the team and current general manager Bruce Allen took over, there isn't a direct correlation between money spent and on-field production.
Follow the Blueprint
The Chiefs are the toast of the town, a worst-to-first revival that's as unexpected as it is spectacular. How did they do it?
Head coach Andy Reid inherited a stout running game and relentless pass defense, gave up a whole lot (likely two second-round picks) to get a quarterback whose best quality is that he doesn't throw interceptions, and added veteran pass-catchers like receiver Donnie Avery and tight end Anthony Fasano.
Avery and Fasano haven't panned out like Reid may have hoped. By refusing to fork over huge money to extend veteran left tackle Branden Albert, though, and drafting rookie Eric Fisher to eventually replace him on the cheap, Reid and the Chiefs are following our statistical blueprint to a "T."
It's hard to argue with the results.