≡ Menu



In September, I started a post by asking you to make this assumption:

Assume that it is within a quarterback’s control as to whether or not he throws a completed pass on any given pass attempt. However, if he throws an incomplete pass, then he has no control over whether or not that pass is intercepted.

If that assumption is true, that would mean all incomplete pass attempts could be labeled as “passes in play” for the defense to intercept. Therefore, a quarterback’s average number of “Picks On Passes In Play” (or POPIP) — that is, the number of interceptions per incomplete pass he throws — is out of his control.

After doing the legwork to test that assumption, I reached two conclusions. One, interception rate is just really random, and predicting it is a fool’s errand. Two, using a normalized INT rate — essentially replacing a quarterback’s number of interceptions per incomplete pass with the league average number of interceptions per incomplete pass — was a slightly better predictor of future INT rate than actual INT rate. It’s not a slam dunk, but there is some merit to using POPIP, because completion percentage, on average, is a better predictor of future INT rate than current INT rate.

So, why am I bringing this up today, at the start of Super Bowl week? Take a look at where Sunday’s starting quarterbacks ranked this year in POPIP (playoff statistics included, minimum 250 pass attempts):

RkPlayerTmCmpAttCmp%IntINT RTPOPIP
1Tom BradyNWE45573162.2101.43.6
2Joe FlaccoRAV36862459101.63.9
3Colin KaepernickSFO16927062.641.54
4Robert Griffin IIIWAS2684126561.54.2
5Aaron RodgersGNB42062467.391.44.4
6Nick FolesPHI16126560.851.94.8
7Ben RoethlisbergerPIT28444963.381.84.8
8Blaine GabbertJAX16227858.362.25.2
9Matthew StaffordDET43572759.8172.35.8
10Sam BradfordRAM32855159.5132.45.8
11Cam NewtonCAR28048557.7122.55.9
12Andrew LuckCLT36768153.9192.86.1
13Matt SchaubHTX41363365.2142.26.4
14Carson PalmerRAI34556561.1142.56.4
15Ryan TannehillMIA28248458.3132.76.4
16Christian PonderMIN30048362.1122.56.6
17Peyton ManningDEN42862668.4132.16.6
18Russell WilsonSEA29145564112.46.7
19Josh FreemanTAM30655854.81736.7
20Michael VickPHI20435158.1102.86.8
21Eli ManningNYG32153659.9152.87
22Drew BreesNOR42267063192.87.7
23Brandon WeedenCLE29751757.4173.37.7
24Chad HenneJAX16630853.9113.67.7
25Jay CutlerCHI25543458.8143.27.8
26Matt RyanATL47669268.8172.57.9
27Andy DaltonCIN34355861.51737.9
28Philip RiversSDG33852764.1152.87.9
29Jake LockerOTI17731456.4113.58
30Ryan FitzpatrickBUF30650560.6163.28
31Tony RomoDAL42564865.6192.98.5
32Mark SanchezNYJ24645354.31848.7
33Matt CasselKAN16127758.1124.310.3

Everything about New England’s offense was an outlier this year, so I’m never surprised to see Brady at the top of any list. But what does it mean that Flacco and Kaepernick come in at 2nd and 3rd in (fewest) interceptions per incomplete pass? Well, the initial reaction is that perhaps both quarterbacks were really lucky, particularly Flacco. Kaepernick ranked 11th in completion percentage and finished 4th in interception rate, thanks to his great POPIP. Meanwhile, Flacco ranked 21st in completion percentage and finished 5th in interception rate! That’s a particularly drastic split, although not necessarily one that places him in great company.

Things get even more interesting when you consider the types of throws each quarterback is making. In my second POPIP post, Mike Clay helped me quantify the relationship between interception rate and the length or depth of a throw. Look at the NFL averages below:

LOS or Behind120081351.1%1023185.2%

On throws behind the line of scrimmage, completion percentage is very high and interception rate is very low; as pass attempts get progressively farther away from the line of scrimmage, the interception rate increases and the completion percentage decreases.

That all makes sense, until we get to this next bit of information, courtesy of the ESPN Stats and Information department:1

Including the postseason, Joe Flacco and Colin Kaepernick are #3 and #4 respectively in average length of throw. Flacco’s average pass travels 9.93 yards in the air, Kaepernick’s 9.66. Jay Cutler and Andrew Luck are #1 and #2.

So Joe Flacco throws a ton of deep passes. And his completion percentage is pretty low. No issues there. The weird part is that his interception rate is extremely low. It would be pretty simple for me to end this post with the line, “Joe Flacco has had a really low interception rate this year, despite being both inaccurate and a risky passer (as measured by his average length of throw). You can imagine how sustainable that is.”

Except here’s the weird thing: Flacco’s had a really good interception rate most of his career.2 It is, of course, possible that Flacco has simply been Wyatt Earp, and his POPIP will regress to the league average in the future. But that’s not the only possibility.

Commenter Red brought up a good point in the original POPIP post:

Could it be that accuracy and decision making are two independent components of QB play? Brett Favre is probably the best example of this. He had several seasons in which he completed over 65% of his passes, yet still had awful INT % in those years (2003 and 2008 come to mind). Favre was always known as a great thrower, but a poor decision maker, which is backed up by the discrepancy between his completion % and INT/INC %.

I suppose that would make Flacco a good decision maker but an inaccurate quarterback? I’m not sure if those are the elements that lead to the fifth best interception rate in the NFL, but it’s worth pondering. Roman Gabriel, Doug Williams, Neil O’Donnell, Neil Lomax, Roger Staubach, and Donovan McNabb threw far fewer interceptions than you would predict based on their completion percentages. Subjectively, Flacco doesn’t really remind me of a McNabb or an O’Donnell, a player who works well within the system and doesn’t take many chances. Joe Montana, Mark Brunell, Ken Anderson and Neil Lomax also had pretty low POPIP ratios, too, although I’m not sure if there’s a common thread among those quarterbacks, either.

So, is Flacco lucky, or is he the type of quarterback who will always have a high average depth of throw, a low completion percentage, and a low interception rate? And if so, why is that the case (perhaps because he’s usually playing in favorable game scripts)? My intuition is that he’s lucky, although he’s been lucky for quite a few years now. If that’s the case, you never know when Lady Luck will turn on you, and Super Bowl Sunday would be as likely a day as any other. But it’s worth pondering whether there is a legitimate reason Flacco’s been able to keep his interception rate so low.

[Update: I e-mailed with Mike Clay after publishing this post, and Mike showed me some numbers indicating that Flacco historically has had a really low interception rate on deep passes. In the second footnote to this post, I noted that only 2 of his 148 passes that were labeled “deep” were intercepted. After Mike pointed out his historical data, that made me take another look at the 2012 numbers. I think deep means greater than or equal to 15 yards, so based on the aDOT/INT table, you would expect 8-12 interceptions on so many deep throws. Considering that Flacco threw only two, and has a history of not throwing interceptions on deep passes, perhaps there’s something there.

On the other hand, I hesitate to give him too much credit, because I would think “trying not to throw interceptions when you go deep” isn’t a skill that only Joe Flacco has acquired. Part of it may also be related to the game script idea hinted at above, in that more of Flacco’s deep passes may come when his team is winning and he doesn’t need to be as risky. All quarterbacks throw more interceptions in losses, but he threw 6 INTs in 5 losses (ignoring the CIN game) and only 4 in 10 wins (13, if you count the post-season). Essentially, it may be that Flacco is throwing passes like he’s losing but since he’s winning, he can be more conservative with them. I’m also not sure how that impacts how we would predict his future interception rate: if his team is trailing, he may revert to the league average in that metric, but if his team continues to be a Super Bowl contender, maybe it won’t.]

  1. Via e-mail, ESPN also added this: Including the postseason, Kaepernick has the best completion percentage (52.6%), yards per attempt (17.0) and QBR (99.7) on throws 20+ yards downfield. []
  2. I’m not really sure what to do with this information, but it probably isn’t irrelevant. Flacco threw ten interceptions this year, and I checked the official play-by-play logs to see how the passes were labeled: two were “short left”, four were “short right”, two were “short middle”, one was “deep right” and one was “deep middle.” He threw 157, 215, 102, 49, 36, and 63 passes in those categories, respectively, with two passes not described. So 31% of his passes were deep this year, but those accounted for only 20% of his interceptions. []
  • Brian

    Do POPIP rates correlate from year to year? Are Flacco’s seasonal POPIP rates consistently good?

    • Chase Stuart

      Flacco’s POPIP rates have to be good, because he’s always been inaccurate and he’s always thrown interceptions. He ranked 20th in interceptions/incomplete pass in 2008, 10th in 2009, 8th in 2010, and then 5th in 2011 (before ranking 2nd in 2012).

      What I’ve found is all correlations when it comes to INT rates are really random. It’s tough to even come up with a number, because since INT rates have been constantly decreasing, picking the right cut-off rate is hard to figure out (and the answers vary pretty significantly based on the cut-off you choose). I think that trying to predict future INT rate from past INT rate is just about impossible; trying to predict future INT rate from past CMP% is slightly better.

  • “Deep” in the play by play means the pass was caught more than 15 yards from the line of scrimmage.

  • Gus A.

    As a Steeler fan, I’ve seen a lot of Flacco and I don’t think his low INT % is a fluke. My observation is that he rarely underthrows the deep ball and this is where more INTs will happen.

    • Chase Stuart

      If only Rahim Moore had watched as much tape!

    • Kit

      I agree–with what you said about Flacco, not with being a Steelers fan, that’s a terrible idea. Flacco has one of the strongest arms and when he throws deep it always used to go 5 yards farther than receiver or defender. Now that he’s got the timing w Torrey worked out he’s completing a lot more deep throws, but the ones he misses are still always overthrown. In past seasons one of the biggest reasons he got picked was because he seemed to be deciding where he was going to throw the ball before the snap; consequently he threw into coverage a lot more. INT’s will always be impossible to predict though, because no one can account or all the picks that defenders should have but drop, or all the balls that a receiver mishandles, resulting in picks.

  • Ben Stuplisberger

    I can’t remember where this was discussed before, but an interesting explanation for the low completion %/low interception rate combination was that inaccurate passers are more difficult for the defense to make plays on, because the path of the passes were less predictable than from an accurate passer. I think McNabb was used as an example. Otherwise, this problem is interesting and perplexing. I’m stumped.

  • Red

    Ben, that’s been my theory for years. McNabb is indeed a good example, as he was always known for throwing dirtballs at his receiver’s feet, where no one, including the defender, had a chance of catching it.

    However, the most glaring contrast is the last two seasons of quarterbacking from the Denver Broncos. In 2011, Tim Tebow completed a historically awful 46.5% of his passes; in 2012, Peyton Manning completed 68.6% of his passes. Yet, they had nearly identical interception rates (2.2 % for Tebow, 1.9% for Manning). Tebow’s POPIP was 4.1%, while Manning’s was 6.0%.

    Peyton makes better decisions and can read a defense far better than Tebow, AND Peyton throws shorter passes that are statistically proven to yield fewer INT’s. Almost inconceivably, Tebow still has a better POPIP. The only rational explanation is that Tebow’s passes are SO inaccurate that nobody on the field has a chance to make a play on them. From watching the games, I know this is true, as many of Tebow’s passes landed 15+ yards away from anybody, and often landed out of bounds or even in the stands. Is throwing the ball so terribly that it can’t be intercepted really a skill? I don’t think so.

  • Red

    Sorry to double post, but I just thought of an analogy from a different sport: bowling. Think of a strike as a completion, and a split as an interception. In order to leave a split, you must hit the headpin. In order to throw a strike, you also must hit the head pin. Therefore, in order to consistently throw strikes, a bowler must take the risk of hitting the headpin and possibly leaving a split.

    Conversely, think of a terrible bowler who regularly misses his target by a foot or more, and even tosses a few shots in the gutter. His strike % would be very low, but his split % would also be low, precisely because his shots were so far off. However, the number of strikes sacrificed by this “avoid splits strategy” would be so counterproductive to that bowler’s score that you would never call it a true skill. Substitute completions for strikes and interceptions for splits and the analogy still holds.

  • Flacco had a very high “Poor Throw %” (overthrows, underthrows, passes thrown wide) in 2012. I wonder if his poor passes are so poor they cannot be caught by anyone, hence few INTs.

  • Richie

    What if Flacco makes extra effort to only throw balls where defenders can’t reach them? Let’s say he throws deep to a receiver, but always tries to throw the ball to the side of the receiver opposite where the defender is, and maybe he even tries extra hard to do this. These would be balls that are difficult for receivers to catch (or even get to) and impossible to intercept. This could lead to low completion and interception rates.

    So often I think back to the last couple years of Dan Marino’s career. I think that his head never really understood what his arm was no longer able to do. I think he would often try to thread the needle to a receiver, and it would get picked because he didn’t have the same ability to thread the needle he once did.

    What if Flacco is just making less of an effort to throw these risky passes?

  • Ravensfanintx

    Been watching Joe since UD, so let me shed a little insight.
    Go back to his senior year in college..48 tds, 4 picks.
    His MO is to throw it away, far away, if nothing is there.
    Rarely will. He try and ore a throw.
    Lastly, since the addition of 82 last year, who outruns very one, joes arm is so long that he just chucks it out there and let Torrey. Goget it. If it falls, it falls.
    Lastly, a point on the mans psyche.
    He doesn’t. Care about stats or anything but wins.

  • Bryan

    The observation that one of the responders made about Joe Flacco and his deep passes rarely being underthrown is true of Colin Kaepernick as well. My feelings on this issue is that there are three different reasons why any given pass play could result in an interception. It could be a poor decision by the quarterback, it could be an inaccurate throw by the quarterback, or it could be simply bad luck (short term variance) where a pass is deflected in some nature into the hands of a defender.

    I really think the key here is the fact that we have in play Kaepernick’s predecessor, Alex Smith, who also excelled by throwing an extremely low number of interceptions. While the sample size isn’t large enough to tell with any absolute certainty, I believe the system has a lot to do with these low number of interceptions that Flacco, Kaepernick, and Smith have thrown. And could it be any coincidence that these two teams are almost exact mirror images of each other as far as strengths and weaknesses go? And that they are coached by brothers with very similar offensive philosophies?

  • Mon

    “Assume that …if [a quarterback] throws an incomplete pass, then he has no control over whether or not that pass is intercepted.”

    Your whole argument rests on the validity of your premises (as all arguments do). Your premise is false. This one, at least. So, the rest of your article doesn’t matter in the slightest. I kept reading since I’m a numbers junkie and it’s 2:55am, however:

    1) Sometimes Quarterbacks put the ball where no one can get it so they don’t get sacked or throw an interception. I don’t just mean balls thrown away (which is the most obvious example where your premise is false), but also balls thrown in the dirt, or thrown to open space.
    2) Sometimes Quarterbacks put the ball where only their receiver can get it and here is where it gets tricky. “Assume that quarterbacks have control over completions” — they don’t. Why? Drops. A quarterback could exercise every iota of his control to completely harmony and precision making the most perfect of calculated risks throughout the duration of a play as if Deep Blue himself were going through the QB read progressions/timing/reading of defenses and a laser-guided robot could deliver the ball perfectly to a receiver and…he could still drop it. This affects incompletion rating. But here’s an interesting paradox.

    It also affects, sometimes, interception percentage. A QB throws a perfect pass, the receiver drops or rather “props” the ball up, and a defender intercepts it. This happens more often than one might think, and it happens enough to influence metrics.

    Meanwhile, the important part is that your argument is rubbish because your premises are invalid. No point.