Why Chip Kelly failed in the NFL and what it means for UCLA

Chip Kelly’s return to the college ranks is exciting — the sport is better when better coaches are involved. But there are a lot of questions regarding what a Kelly offense looks like after his time in the pros and how his philosophy has shifted in the six years since he left Eugene. From my 2018 UCLA preview:

He went 46-7 as Head Duck, and now he’s at a school with a much higher recruiting ceiling. It appears he is integrating some of the basic concepts — “It’s a lot of read option, dish-and-dunk kind of stuff” — and he’s breaking in some of the sports science techniques (GPS devices and whatnot) that he’s become known for.

I’m curious about the philosophical changes Kelly’s offense seemed to undergo in the NFL, just how much his new college offense will look like his old one, and most of all, whether he can do what no UCLA coach since Terry Donahue has done: win consistently.

Kelly’s NFL tenure started out like a house afire.

His 2013 Eagles ranked third in offensive DVOA and improved from 4-12 to 10-6. They pushed tempo as far as you can in the NFL (they averaged one snap every 23.9 seconds in 2013, then one every 22.2 in 2014, both tops in the league). They scored 30-plus points eight times, and they lost in the playoffs only via last-second field goal.

Photo by Rich Schultz /Getty Images

Even into his second year, as the magic began to fade (they ranked 13th in offensive DVOA), the Eagles went 10-6.

By 2015, the magic was gone. Philadelphia fell to 26th in offensive DVOA and 7-9 overall, and Kelly was fired. He took over a horrid roster in San Francisco in 2016, went 2-14, and became a TV commentator for 2017.

As Kelly was preparing for his lone year in San Francisco, Smart Football’s Chris B. Brown penned the definitive take on Kelly’s devolution. The gist:

• Defenses grew more sophisticated in handling tempo.
• NFL officials don’t let you go as quickly as Kelly wanted.
• Having the QB run is terrifying for a pro team that’s invested millions of dollars in the position. An injury could wreck your season even more than it could in college.
• Without that run threat and with immobile QBs like Mark Sanchez and Sam Bradford, Kelly’s play-calling was predictable.

And perhaps his preferred philosophy (which included the elements of what you might call an optimal college football offense) wasn’t as suited in a pro environment that features more adaptable coaching and fewer talent advantages. As the Washington Post’s Mark Maske noted, Kelly was given far too much control of personnel decisions in Philadelphia. GM Kelly perhaps screwed head coach Kelly out of success.

Still, I would add one more thing to the list of factors against him in Philadelphia and San Francisco.

Close your eyes and picture Kelly’s Oregon offense.

Maybe you’re flashing back to that Statue of Liberty against Michigan. But more likely, it’s some dam-bursting, 70-yard explosion. Kelly’s Ducks wore you out with tempo and then sliced you vertically.

If there’s a singular difference between college and pro football, it’s that big plays are a lot harder to come by in the NFL. A lot harder.

Football is a “game of inches.” We hear that on broadcasts, from coaches, and out of our own mouths. We understand how tight the margins are in this sport.

But as games and seasons pass, the gray area disappears. The narrative says you either won, or you lost. You succeeded, or you failed.

Want to see how blurry the lines are between success and failure in football, though?

Below is a chart that features two pieces of data:

1. The distribution of gains for Kelly’s unstoppable Oregon offenses.
2. The distribution of gains for Kelly’s failed NFL offenses.

The data is presented cumulatively, so that by the time you get to the far right, you’re accounting for 100 percent of plays. As you can see, there was very little difference here between Kelly’s college and pro offenses.

Broken out into yardage ranges, that data looks like this:

There are but meager differences.

• Kelly’s NFL teams suffered more zero-yard plays — completion rates were 62 percent at each level, but Kelly did call more passes at the NFL level, which meant more incompletions — but also suffered fewer negative plays: 9 percent vs. 11 percent.
• For any offense, a large percentage of your plays gain between one and six yards. At both levels, Kelly’s offenses did this about 35 percent of the time.
• Twenty-two percent of his Oregon plays gained between seven and 14 yards, while 20 percent did in the NFL.
• One in 27 plays (3.7 percent) gained 30-plus yards for the Ducks. Only one in 38 did (2.6 percent) in the NFL.

That last one doesn’t sound like much of a difference, but considering differences in tempo (his Oregon offenses averaged 75.2 plays per game, and his NFL offenses averaged 66.9), that’s 2.8 explosions per game versus 1.7, one more easy score per game.

A lack of explosive big plays might’ve made all the difference.

We know there is less scoring in NFL games, but we don’t necessarily think about why.

College offenses are about as efficient as pro offenses, on a down-by-down basis. The average success rate for college offenses from 2009-17 was 42.1 percent, and in the pros, it’s 41.2 percent.

As much as anything, it’s the explosive plays that define the difference between college and pro.

A coach once told me that when watching Kelly’s Philadelphia offenses, one or two 19-yarders per game would make him think, “That would have gone to the house at Oregon.”

In college since 2009, 2.7 percent of plays gained at least 30 yards. In the NFL, it was 2.4 percent.

Combined with differences in tempo (college games averaged 72.7 plays per game to NFL’s 63.1), that means two big gashes per game to the NFL’s 1.5. That, plus slight differences in efficiency and turnovers (1.55 per game in college to 1.49 in the pros), accounts for most of the difference in scoring (26.4 points per game vs. 22.5).

These margins are tiny. But they add up.

Think of it this way: the college-vs.-NFL difference of 0.9 percentage points of success rate over 66.9 plays per game means 0.6 fewer successful plays per game. The average successful play gains about 12 yards in both college and pro, so losing about half a successful play means gaining about seven yards fewer per game.

Meanwhile, removing half of a single big play per game could subtract those seven yards plus another 25 or more yards from a box score. Missing 30 yards from a single play could drop your per-play average over the course of a game by about 0.5 yards.

For an offense like Kelly’s, so defined by big plays at Oregon, you can see how costly this difference could be. Removing a single huge gain can change a game much more than taking away a couple chain-movers, for obvious reasons.

Granted, things worked brilliantly for a while.

His first year in Philly produced a higher big-play rate than at Oregon — 9.8 percent of plays gained at least 20 yards compared to 7.7 percent at UO.

But that shrank quickly, to 6.8 percent in 2014, 6.1 in 2015, and 4.7 in 2016.

According to data in the Football Outsiders Almanac, the extremes in Kelly’s philosophy were evident each year, even with the 49ers:

• Kelly’s NFL offenses ranked first in percentage of snaps out of the shotgun/pistol every year: 86 percent in 2013-14, 94 percent in 2015, and 99 percent with SF in 2016.
• They lined up in a single-back formation at least 92 percent of the time each year — the most in the NFL from 2013-15 and second most in 2016.
• They did attempt balance on first downs, but they ran more on second-and-long than anyone (at least 41 percent of the time each year, ranking no lower than third), almost always ran in power situations (ranking second or higher three of four times), and attempted as much play-action as anyone (fourth or higher three of four times).

Combined with tempo, this all sounds like what Kelly wanted to do at Oregon.

The genius of his run-heavy system at Oregon was that you could create big plays without big risk. The Ducks’ tempo made defenses scatterbrained, and backs like Jeremiah Johnson, LaMichael James, Kenjon Barner, and De’Anthony Thomas — guys speedy enough to combine for six yards per carry in the pros — made them pay. And when they overcompensated, quarterbacks Marcus Mariota, Darron Thomas, and Jeremiah Masoli would keep the ball.

Photo by Jonathan Ferrey/Getty Images

In the NFL, his run game was always at least above average in explosiveness, but “explosive” means something different there, particularly on the ground. Big NFL plays almost all come through the air.

You eventually have to be able to throw, so after Kelly’s initial NFL surge, his philosophy became less and less of a fit.

He’s now back in his college comfort zone, though.

Through recruiting, you can build more permanent talent and speed advantages here, and you can run the ball with great effect, especially if you are less afraid of getting your QB hit.

If Kelly’s got anything in his first year at UCLA, it’s options at QB.

Former Michigan starter/statue Wilton Speight is available, in case NFL experience scared Kelly away from running the QB as much. Kelly certainly tried to sell Speight on his NFL offense.

Kelly pitched Speight on a system like the one the coach ran in the NFL, where he turned to pocket-passers such as Nick Foles, Sam Bradford and Mark Sanchez. Foles was named to the 2013 Pro Bowl while leading the league in yards per pass attempt and passer rating. Sanchez, the former USC star, personally vouched for Kelly’s offense as the two quarterbacks trained together.

Still, your opponent’s math changes if it has to account for a mobile quarterback. At Oregon, Kelly had guys who could throw (especially Mariota) and punish defenses for over-pursuing against a hand-off.

One would assume a tie would go to a more mobile guy like freshman Dorian Thompson-Robinson or a split-the-difference QB in Devon Modster, who can run but also completed 65 percent of his passes filling in for Josh Rosen. Betting odds favor Modster.

Photo by Harry How/Getty Images

As Brown wrote, Kelly’s NFL legacy will end up more like a Hal Mumme than a Bill Walsh.

The NFL’s Kelly was an innovator who paved the way for others to innovate and succeed more than he did.

As the new coach of the San Francisco 49ers, the man who was at one time football’s leading innovator seeks redemption in the heart of Silicon Valley, America’s current cradle of disruptive innovation, a fitting landing spot given that it appears Kelly is seemingly hurtling toward being the next victim of the “Innovator’s Curse.” [...]

The second idea behind the Innovator’s Curse is that, having once innovated, it’s increasingly difficult for the innovator to continue innovating. To use Silicon Valley examples, there are countless IBMs, Xeroxes, and Yahoos: one-time disruptors whose cultures and ideas ossified and who eventually became the disrupted.

The NFL indeed evolved toward Kelly’s vision. Teams are lining up in the shotgun more and using more one-back, three-receiver sets. The Super Bowl, won by the Eagles, was full of college offense. Plus, research suggests play-action is the way to go if you’re looking for easy yards. And Kelly made defenses improve at adjusting to tempo.

There are worse legacies. And at only 54 years old, Kelly’s got a time to add another chapter in Westwood.