Writers of Pro Football Prospectus 2008

Most Recent FO Features


» 2017 Play-Action Offense

Party like it's 2012! Play-action came roaring back to the forefront of the league last season. From traditional play fakes to RPO shenanigans, play-action was the hot sensation sweeping the nation.

17 Apr 2015

2014 Adjusted Interceptions

by Vincent Verhei

Jay Cutler and Philip Rivers were co-leaders in interceptions last season, but nobody threw more passes that should have been intercepted than Andrew Luck. Shaun Hill threw more interceptions than he probably should have, while no quarterback had better luck on bad throws than Drew Stanton.

Confused? Welcome to the murky world of interceptions, which are determined by lucky bounces and bad circumstances about as often as they are by player skill. That's why we developed adjusted interceptions. Here's the basic idea, which we first introduced four years ago:

  • We add in plays where the quarterback only escaped an interception because the defender couldn't hold onto the ball (dropped interceptions, which we've been tracking in game charting since 2007). Starting in 2012, we also included plays where the defender didn't straight-out drop the ball but instead had it knocked out of his hands by an offensive receiver (a "defensed interception"). These are listed as "Drop/Def INT" in the table at the end of this page.
  • We subtract plays where the interception (or a dropped interception) is tipped to the defender by a receiver who should have caught the pass. These are listed as "Tip INT" and "Tip AND Drop INT" in the table.
  • We subtract Hail Mary interceptions as well as interceptions thrown in desperation on fourth down in the final 2:00 of a game. These are listed as "HM/End Q4" in the table. We also bent the rules and included an interception thrown by Chicago's Jimmy Clausen against Detroit in Week 16 that came on fourth-and-10 with 2:02 to go in the game and the Bears trailing 20-14.

Obviously, there's going to be a little bit of subjectivity here, but our game charters do their best. If the defender has to dive for a ball only to have it bounce off his fingertips, that's not a dropped interception.

Of those three categories, it's dropped interceptions that explain most of the difference between actual interceptions and adjusted interceptions. Jay Cutler led the league with just three "Hail Mary" interceptions, and Matt Ryan and Philip Rivers were the only other players with more than one. It's similar for interceptions (and dropped picks) that bounced off a receiver's hands; Teddy Bridgewater led the league with three, and Shaun Hill and Derek Carr were the only other quarterbacks with more than one.

Interceptions are notoriously hard to forecast from year to year, because there's so much random chance and statistical noise involved. Historically, we've found that adjusted interceptions have a higher year-to-year correlation than standard interception totals and are a better predictor of future interceptions. For the entire period from 2007 through 2013, the correlation coefficient for adjusted interception rate (0.30) is twice as high as the correlation for actual interception rate (0.15). In each set of years except one (2011-2012), adjusted interception rate was a better predictor of interceptions the next year than actual interception rate.

As we mentioned in the intro, Jay Cutler and Philip Rivers were co-leaders in interceptions last season, with 18 apiece. It was Andrew Luck, though, who finished first in adjusted interceptions with 23. Cutler was second, but then you have to go through six other players (Andy Dalton, Blake Bortles, Eli Manning, Brian Hoyer, Drew Brees, and Peyton Manning) before you get to Rivers, whose overall numbers skewed high because opponents had an annoying habit of not dropping his passes.

Both total and adjusted interceptions, obviously, are strongly correlated with total pass attempts, which is why rate stats can be more telling here. Or can they? Aaron Rodgers had the lowest interception rate in the NFL last year, and was virtually tied with Alex Smith (Smith was actually better by a tiny degree) for the adjusted interception rate lead as well. At the other end of the spectrum, Jake Locker had the highest interception rate and the highest adjusted interception rate in all of football. So for all our fancy math, the names at the extremes don't change much.

Contrasting the two rate stats, though, can be illuminating. Take the Rams' Shaun Hill. His raw interception rate of 3.1 percent was just 30th out of 42 qualifying quarterbacks, but his adjusted interception rate of 2.2 percent was fourth. Hill, Teddy Bridgewater, and Josh McCown were the only qualifying quarterbacks with fewer adjusted interceptions than raw interceptions. Bridgewater's ranking improved from 29th to 15th in the move from raw rate to adjusted rate; McCown's, from 40th to ... uh, 31st. Well, he's still Josh McCown, after all. Other big movers include Mark Sanchez (37th to 26th), EJ Manuel (18th to seventh), and Robert Griffin (26th to 17th).

Four quarterbacks dropped at least ten spots in the rankings when we move from raw interception rate to adjusted rate: Ryan Fitzpatrick (22nd to 33rd), Brian Hoyer (27th to 39th), Michael Vick (eighth to 34th), and Drew Stanton (12th to 41st). And now we must discuss the unbelievably lucky season Mr. Stanton enjoyed for Arizona. He threw only five actual interceptions last year, but 13 adjusted, an eight-pick difference that was the biggest in the league. Stanton and Vick were the only qualifiers last year who were more likely to see defenders drop their passes than actually intercept them. Eight of Stanton's passes resulted in dropped interceptions; only Brian Hoyer had more. Now remember this: Stanton only threw 241 passes (not including pass interference calls) last season, 31st in the NFL and less than half of Drew Brees' league-high 653. Think about how ridiculous it is that somebody with so few passing plays still finished so high in any counting stat, even in something obscure like dropped interceptions. Put it this way: in 241 passes, Stanton would have needed to average better than 20 yards per throw to finish second in passing yardage. To finish second in completions, he would have needed to connect on 172 percent of his throws, which would be impossible unless he started throwing two balls at once, and would even then still be very impressive. Coming into last season, Stanton had nine career interceptions in 187 passes, a rate of 4.8 percent that is much more indicative of his true passing abilities than his one-in-a-million 2014 campaign.

Adjusted interceptions for defenses are a little weird, because you're basically rewarding teams for dropping interceptions. It's worth noting, though, that Jacksonville was last in the league with six interceptions, and they only dropped two others. That's eight interceptable passes in 16 games. 26 defenses had at least twice as many. Cleveland (13) and Cincinnati (10) were first and second in dropped interceptions. Perhaps it's the water in Ohio? Of course, they were also tied for second in total interceptions behind San Francisco, which had nine dropped picks itself. In other words, the key to dropping a lot of picks is being good enough to break up tons of passes in the first place. For individuals, Malcolm Jenkins (PHI), Perrish Cox (SF), Joe Haden (CLE), and Kendrick Lewis (HOU) were tied for the league with three dropped picks each. Atlanta's Desmond Trufant had four would-be INTs, but two of them were characterized by charters as "defensed" by the offensive player.

The following table lists adjusted interceptions for all quarterbacks with at least 200 pass attempts. Remember that most quarterbacks will have more adjusted interceptions than actual interceptions. The leaguewide interception rate is 2.60 percent, and the leaguewide adjusted interception rate is 3.30 percent. Therefore the blue "difference" column represents the difference between the two rates compared to that -0.70 percent average, so that positive always means "threw fewer picks than expected" and negative means "threw more picks than expected."

End Q4
Drop INT
Pass Att
(no DPI)
INT Rate
Adj Rate
(vs. Avg)
Andrew Luck IND 16 0 8 1 0 23 616 2.6% 3.7% 0.4%
Jay Cutler CHI 18 3 7 0 0 22 562 3.2% 3.9% 0.0%
Andy Dalton CIN 17 0 4 0 0 21 481 3.5% 4.4% 0.1%
Eli Manning NYG 14 0 7 0 1 20 600 2.3% 3.3% 0.3%
Blake Bortles JAC 17 1 5 0 1 20 474 3.6% 4.2% -0.1%
Brian Hoyer CLE 13 1 9 1 0 20 440 3.0% 4.5% 0.9%
Drew Brees NO 17 0 2 0 0 19 653 2.6% 2.9% -0.4%
Peyton Manning DEN 15 0 5 1 0 19 595 2.5% 3.2% 0.0%
Matt Ryan ATL 14 2 6 0 0 18 623 2.2% 2.9% -0.1%
Philip Rivers SD 18 2 2 0 0 18 570 3.2% 3.2% -0.7%
End Q4
Drop INT
Pass Att
(no DPI)
INT Rate
Adj Rate
(vs. Avg)
Derek Carr OAK 12 1 7 1 1 16 593 2.0% 2.7% 0.0%
Ryan Tannehill MIA 12 1 6 1 0 16 590 2.0% 2.7% 0.0%
Joe Flacco BAL 12 0 4 0 0 16 549 2.2% 2.9% 0.0%
Matt Stafford DET 12 0 4 1 0 15 600 2.0% 2.5% -0.2%
Geno Smith NYJ 13 0 2 0 0 15 366 3.6% 4.1% -0.2%
Ben Roethlisberger PIT 9 0 5 0 0 14 600 1.5% 2.3% 0.1%
Cam Newton CAR 12 1 4 1 0 14 446 2.7% 3.1% -0.2%
Josh McCown TB 14 1 1 1 0 13 323 4.3% 4.0% -1.0%
Ryan Fitzpatrick HOU 8 0 5 0 0 13 311 2.6% 4.2% 0.9%
Drew Stanton ARI 5 0 8 0 0 13 241 2.1% 5.4% 2.6%
Colin Kaepernick SF 10 0 2 0 0 12 479 2.1% 2.5% -0.3%
End Q4
Drop INT
Pass Att
(no DPI)
INT Rate
Adj Rate
(vs. Avg)
Russell Wilson SEA 7 0 6 1 0 12 452 1.5% 2.7% 0.4%
Kyle Orton BUF 10 0 2 0 0 12 445 2.2% 2.7% -0.2%
Tony Romo DAL 9 0 3 0 0 12 433 2.1% 2.8% 0.0%
Nick Foles PHI 10 1 3 0 0 12 313 3.2% 3.8% -0.1%
Austin Davis STL 9 0 4 1 0 12 282 3.2% 4.3% 0.4%
Tom Brady NE 9 0 2 0 0 11 581 1.5% 1.9% -0.4%
Teddy Bridgewater MIN 12 1 3 3 0 11 397 3.0% 2.8% -0.9%
Mark Sanchez PHI 11 0 1 1 0 11 309 3.6% 3.6% -0.7%
Kirk Cousins WAS 9 0 0 0 0 9 204 4.4% 4.4% -0.7%
Zach Mettenberger TEN 7 0 2 0 0 9 179 3.9% 5.0% 0.4%
End Q4
Drop INT
Pass Att
(no DPI)
INT Rate
Adj Rate
(vs. Avg)
Jake Locker TEN 7 0 2 0 0 9 147 4.8% 6.1% 0.7%
Mike Glennon TB 6 0 2 0 0 8 202 3.0% 4.0% 0.3%
Aaron Rodgers GB 5 0 3 1 0 7 520 1.0% 1.3% -0.3%
Alex Smith KC 6 0 0 0 0 6 466 1.3% 1.3% -0.7%
Robert Griffin WAS 6 0 1 1 0 6 214 2.8% 2.8% -0.7%
Shaun Hill STL 7 1 1 2 0 5 229 3.1% 2.2% -1.6%
Carson Palmer ARI 3 0 2 0 0 5 225 1.3% 2.2% 0.2%
Michael Vick NYJ 2 0 3 0 0 5 119 1.7% 4.2% 1.8%
Charlie Whitehurst TEN 2 0 2 0 0 4 182 1.1% 2.2% 0.4%
Colt McCoy WAS 3 0 1 0 0 4 128 2.3% 3.1% 0.1%
EJ Manuel BUF 3 0 0 0 0 3 130 2.3% 2.3% -0.7%

Click here for a look at the adjusted interception numbers from 2013.

Posted by: Vincent Verhei on 17 Apr 2015

24 comments, Last at 21 Apr 2015, 12:45am by tuluse


by theslothook :: Fri, 04/17/2015 - 4:35pm

Adjusting for down, distance, depth of pass, and pressure will likely shed even more light on this.

Having watched Luck, he's a total gunslinger. Yes, he throws longer passes a lot, but he's very favre esque in so many ways.

by Eddo :: Fri, 04/17/2015 - 4:54pm

"Starting in 2012, we also included plays where the defender didn't straight-out drop the ball but instead had it knocked out of his hands by an offensive receiver (a "defensed interception")."

What's the reasoning for this? If a QB throws an incompletion where the WR would have caught it were it not knocked away, I don't see the argument for adjusting that to be a completion.


"... Rivers, whose overall numbers skewed high because opponents had an annoying habit of not dropping his passes."

How do dropped interceptions correlate year-to-year? Rivers is considered to have a weaker arm than someone like Aaron Rodgers, so it follows that his passes should be easier to catch if they come to you.

by jfsh :: Fri, 04/17/2015 - 6:16pm

"What's the reasoning for this? If a QB throws an incompletion where the WR would have caught it were it not knocked away, I don't see the argument for adjusting that to be a completion."

I think they're referring to a situation where the defender would have had an interception, except that the offensive player managed to defend the pass.

by Scott Kacsmar :: Fri, 04/17/2015 - 7:41pm

Right. We also started tracking drops on offense in the same manner (defender knocked the ball away). So the whole Dropped/Defensed concept is the player dropped the ball because the opponent made him. The QB's throw is the same regardless.

by Eddo :: Sun, 04/19/2015 - 4:30pm

Thanks for the clarification.

by tuluse :: Fri, 04/17/2015 - 9:50pm

So that Aaron Rodgers fellow, not bad at this whole football thing.

by theslothook :: Sat, 04/18/2015 - 12:25am

Clearly they blew it letting that Flynn fella go.

by TomC :: Mon, 04/20/2015 - 2:42pm

And it's not like those were all dump-offs to running backs. In fact, I just checked PFR, and Rodgers was 3rd in the league in YPC. The guys around him were Hoyer(?!), Stanton, Fitzpatrick, and Luck, all among the "leaders" in adjusted INT%. Frickin' Packers, grumble, grumble.

by JoeyHarringtonsPiano :: Mon, 04/20/2015 - 3:07pm

Hoyer being among the YPC leaders is not a huge surprise. The Browns offense was based on power running with an occasional deep shot. Hoyer completed plenty of medium to long passes, but had a low completion %. The power running stuff was working pretty well until Alex Mack got hurt, then the the running game, and subsequently, Hoyer, fell apart.

by Vincent Verhei :: Mon, 04/20/2015 - 3:56pm

Deepest Avg. Pass, 2014 (min. 200 plays)

1. Drew Stanton, 11.4
2. Mike Glennon, 11.1
3. Brian Hoyer, 10.4
4. Josh McCown, 10.1
5. Cam Newton, 9.9

Shortest Avg. Pass, 2014 (min. 200 plays)

1. Alex Smith, 5.9
2. Robert Griffin, 7.0
3. Blake Bortles, 7.1
4. Kyle Orton, 7.1
5. Jay Cutler, 7.5

by duh :: Mon, 04/20/2015 - 5:54pm

I'm surprised to see Cutler on the 'short' list, it sure isn't how I think of him. I'm guessing it is a case of throwing short to try and 'take advantage' of Forte?

by Vincent Verhei :: Mon, 04/20/2015 - 6:13pm

Well, Forte did set a running back record with 102 catches last year, and his average catch came 0.4 yards past the line of scrimmage. The average throw to Forte was a little deeper at 0.6, but either way... yeah, that'll do it.

by duh :: Tue, 04/21/2015 - 12:21am

Vince, thanks for answering, it was supposed to end with a period not a dang question mark. ;-)

by TomC :: Mon, 04/20/2015 - 9:52pm

Marshall and Jeffery were hurt all year, and nobody was getting open downfield. Half the time Cutler was dumping it to Forte, half the time he was forcing it to someone else. Thus the crap YPC and the crap INT %. It was ugly.

by tuluse :: Tue, 04/21/2015 - 12:45am

The line had injury problems too, so no slow developing plays.

by Will Allen :: Sat, 04/18/2015 - 10:52am

It really is promising to have a rookie qb, playing on team with a so-so o-line (with terrible left tackle performance), a so-so defense, no superstar receivers, and without their best running back, come in with the 15th best adjusted int rate. The Vikings absolutely have to make their top priority to be giving Bridgewater a comfortbale environment to further develop. That means retaining Peterson, unless somebody offers so much that they can immediately make a big gain in o-line performance.

by ammek :: Sat, 04/18/2015 - 12:31pm

Ken Whisenhunt: quarterback guru.

Interesting numbers, as ever, though I wish the tables were drop down sortable. The blue column is confusing; if I understand correctly, it refers to percentage points rather than straightforward %, doesn't it? In any case, it's hard to compute because the figures are pretty much bunched together, apart from ridiculous luckmeister Drew Stanton.

by tuluse :: Sat, 04/18/2015 - 1:18pm

Here's the table copy-pasted into a Google sheets chart. https://docs.google.com/spreadsheets/d/1gzjqgLECVITghGCkrx2bWSGILoVJXejE...

Sort to your heart's content.

by Vincent Verhei :: Sat, 04/18/2015 - 4:07pm

The formula for that column is:

(Adusted Int. Percentage) - (Int. Percentage) -((League Avg. Adjusted Int. Percentage)-(League Avg. Int. Percentage)).

The fact that most players end up bunched together is telling in an of itself. This is really only telling for a handful of players at either end (Stanton at one, Shaun Hill at the other). Everyone else could rise or fall dramatically in the tables if just one play had ended up differently.

by ammek :: Mon, 04/20/2015 - 4:53am

Thanks, both.

I take Vincent's point about bunching. An alternative way of presenting the final column would be to make that first minus in the formula into a divide. That unbunches the numbers a bit. Stanton's 5.4% adjusted interception rate, compared with an expected 2.8%, is 93% higher. For every pick he threw, a QB with average luck would have thrown almost two, and Shaun Hill would have thrown three-and-a-half.

by herewegobrownie... :: Sat, 04/18/2015 - 12:37pm

So now we see why PFF gave Stanton a red (poor) rating - I thought it was too harsh, but it was based on the idea that he "should've" been worse than he was.

by commissionerleaf :: Sun, 04/19/2015 - 7:47pm

The takeaway is that avoiding interceptions is more luck than skill, right?

by ChrisS :: Mon, 04/20/2015 - 1:27pm

I'd say it is more accurate to say that having interceptions dropped is more luck than skill. The coefficient of variation (std dev relative to average) for drop% is twice that of int%, showing that drops are far more random than interceptions.

by RickD :: Mon, 04/20/2015 - 2:36pm

It may be luck, but it's certainly not Luck.