Writers of Pro Football Prospectus 2008

26 Feb 2011

Quarterbacks And Sack Times

J.J. Cooper, a blogger for NFL Fanhouse, watched every sack of the NFL season with a stopwatch and counted whether the sack came within three seconds of the snap. While it's a little simplistic to say that all three-plus-second sacks are the quarterback's fault, and that all sacks shorter than that can be blamed on the line, the results make intuitive sense -- the Manning Brothers took just one "long" sack each all year, while Ben Roethlisberger and Michael Vick combined for 39. The king, though, was Joe Flacco, with 25 "long" sacks and only 15 "short" ones.

A few other observations:

* On average, about 35 percent of sacks are "long," so the Flaccos and Roethlisbergers and Vicks of the world, with the vast majority of their sacks lasting longer than three seconds, really are outliers.

* Jimmy Clausen: Five "long" sacks, 29 "short" sacks. It's not all his fault.

* Didn't Aaron Rodgers have a rep for holding the ball too long as a young player? He seems to have gotten over it somewhat -- 10 "long" sacks, 21 "short" sacks.

Posted by: Vince Verhei on 26 Feb 2011

38 comments, Last at 18 May 2011, 7:11pm by jerry guy

Comments

1
by Matt Aquiline (not verified) :: Sat, 02/26/2011 - 2:53pm

As you pointed out, these seem a bit simplistic. Wouldn't a huge factor be the number of attempts? FOr instance, Big Ben and Vick have about the same under and over 3 second sack stats. But if Vick dropped back say 100 more times than Ben (he played 4 more games and Andy Reid refuses to run the ball), and gained 300 more yards of scrambles on busted passing plays, or conversely, had 4 more fumbles on busted passing plays and 2 more interceptions, on those 100 more drop backs, then that chart is deceiving in showing them as having about the same success at avoiding sacks when in reality they may have did it at much different rates. Or am I missing something?

2
by battlered90 (not verified) :: Sat, 02/26/2011 - 2:59pm

Right on. It would be fairly easy to divide all stats by attempts to get a better sense of whats going on here. Should also set an attempts threshold to get rid of irrelevant attempts. I don't particularly care that Leon Washington's one sack was longer than three seconds. Still the results regarding Clausen are interesting. And the Manning's sack rate is unbelievably low esp given the publicity given to the Colts problems at Oline. When Peyton bails on a play and flops on the ground to avoid taking a hit that is registered as a sack right?

4
by Matt Aquiline (not verified) :: Sat, 02/26/2011 - 3:17pm

The problem is, it's not easy. WHat are we counting? Sacks per attempt which lasted more than 3 seconds? He'd have to time every passing play and see if each passer passed the ball in less than 3 seconds. when is the pass over, when it leaves his hand I guess. And how about scrambles? I guess those would all be over 3 seconds, but if one was on the borderline somehow (Vick realizing his tackle has whiffed again and bails immediately), does the clock stop when the QB starts running, or when he reaches the line of scrimmage? breaks the plane of the LOS or completely crosses (he can legally throw a forward pass if any part of his body is still behind the LOS, but the ball has "advanced" on his scramble to wherever it is, if I understand the rules). So while the chart is good, it is too simplistic, but if we want it to actually tell us great stuff, we're demanding a $#$@load of work from the poor guy, I think. I'd write this on the guy's own blog, but I tried to post my comment twice and it hasn't gone up, so apparently I'm computer illiterate

8
by tuluse :: Sat, 02/26/2011 - 3:46pm

For tackles of QBs behind the LOS, my understanding is that it's up to the scorers discretion whether it is a sack or a rushing attempt.

22
by Aaron Brook's Good Twin (not verified) :: Sun, 02/27/2011 - 8:16pm

A good clue is if a lineman is now 5 yards downfield -- it's either a run, or you are watching the Lions.

11
by Jerry :: Sat, 02/26/2011 - 4:42pm

You have the raw data, even for guys who were sacked once, so you can divide by attempts yourself if you think it would be useful. You can even post those results here if you think they're interesting.

3
by Anonymous1 (not verified) :: Sat, 02/26/2011 - 3:16pm

This data would have been better presented in percentage terms.

Cutler's 37% long holds are right around average, but the table is arranged to make him look like one of the worst offenders. Whereas someone like Freeman looks good but over half come from long holds.

5
by Matt Aquiline (not verified) :: Sat, 02/26/2011 - 3:20pm

Agreed, sorting through the data was hard on the eyes

6
by AnonymousA (not verified) :: Sat, 02/26/2011 - 3:29pm

This is a huge problem with semi-pro sports statistics (FootballOutsiders, BaseballProspectus, etc.) -- they tend to use overly simplistic measures to make things "clearer". What you really want is a bar graph of times with narrow bins -- that will both show the mean/max/reliability visually, it will also show the distribution. As it is, we don't have know if all players have the same distribution, so a comparison has very little meaning. The fact that there are a total of TWO bins (<3s, >3s) is a pretty crushing flaw for any real analysis.

12
by Vince Verhei :: Sat, 02/26/2011 - 5:13pm

Just to clarify: These are not Football Outsiders' stats. We just linked to them.

14
by The Ninjalectual :: Sun, 02/27/2011 - 2:41am

I hate to lump "AnonymousA" in with a category, but it seems like every few posts these days someone has an epistemological crisis. It's like a bunch of people all just realized individually, "statistics are inherently flawed... therefore THESE statistics are inherently flawed... and we can't ever KNOW anything! If we can't ever KNOW anything then I have to tell everyone: Hey everybody! These statistics aren't perfect!"

I know that every writer at FO has at one time or another clarified that stats are just a tool to use when analyzing what you see, but you can't just SAY that, everybody's gotta figure it out on their own.

33
by AnonymousA (not verified) :: Mon, 02/28/2011 - 1:58pm

This is a complete mis-characterization of my comment. The problem is not that "statistics are inherently flawed" -- they're actually inherently perfect. All (correctly used!) stats are provably derived from a set of assumptions.

The problem with semi-pro sports stats is that the statistical formulae are often employed when the assumptions aren't honored. My favorite example of this is the use of averages. The arithmetic average of data points drawn from a Gaussian distribution converges to the mean/maximum of that distribution. The arithmetic average of data points drawn from a Cauchy distribution...does not converge. Google the distributions -- they are visually VERY similar, yet the result of averaging your data from one is incredibly useful, and the result of averaging your data from the other is a nigh-meaningless number.

So what statistical process was employed by the author of this post (which is not FO, though FO has made similar mistakes)?

1) Bin the data. What's the justification for this? Why only 2 bins? Why was 3 seconds chosen?
2) Sort QBs by the mode of this binned data.

So we don't know:

1) Is the distribution the same for all QBs/lines/defenses? Even if we apply the same statistic to all QBs/situations, if the distributions are different, the numbers are incomparable.

2) What distribution(s) are we dealing with? Are they multi-modal? If they are, this analysis is an obvious joke. If they aren't, is the mode (of this binning) meaningful?

3) What is the reliability of this analysis? Computing confidence intervals with only two bins is not particularly worthwhile, but even if there was good data available, and it was consistently from the same unimodal distribution, and a proper mean analysis was done on it, the confidence intervals could be so wide that the analysis is meaningless.

As to my personal motivation: this is not a harangue or a down-my-nose too-good-for-this-stuff critique. It's an attempt to point out where we are, to help motivate people to do better. I wish I had the time to really brush up on stats, watch every game of the last 10 years and come up with some good analyses. Maybe some day I will. But in the mean time, the best I can do is point out areas others can do better. Not much of a contribution, but it's all I can offer right now. That's a far cry from "stats are meaningless! We can't ever know anything!" I hope that difference comes through in all my posts.

36
by Aaron Brooks' Good Twin (not verified) :: Tue, 03/01/2011 - 4:25pm

Do you have a legitimate basis for asserting that QB sack times are forced resonance-based (Cauchy) as opposed to randomly distributed (Gaussian) as a fundamental phenomenon, or are you just harshing on the use of mean as a statistical variable?

17
by Rabbit :: Sun, 02/27/2011 - 12:56pm

This data deserves to be incorporated into your offensive line ratings, I'll say. Any offensive line ratings, actually, that uses straight sacks as input.

18
by Rabbit :: Sun, 02/27/2011 - 12:57pm

For that matter, into defensive back's ratings, as well.

9
by tuluse :: Sat, 02/26/2011 - 3:57pm

Easy enough to do, I made a google doc with the QBs sorted by over 3 second sack percentage descending.

https://spreadsheets.google.com/ccc?key=0ArVjaWQMmwD7dHZqb1ZjcGVyTGxXYlp...

15
by dcaslin :: Sun, 02/27/2011 - 10:45am

Thank you for doing what the 100 paid sportswriters that linked to this article couldn't be bothered to do (FO excepted, they have their own stats to spend time on)

19
by dcaslin :: Sun, 02/27/2011 - 1:03pm

Ok, I just realized you didn't add in attempts, so I added that in as well and shared here (I removed most of the low sack numbers, b/c I'm lazy): https://spreadsheets.google.com/ccc?key=0AlpZoLilujCfdFJiNnFja3pjdndOZ1V....

Top offenders w/ long sacks per attempt:
Ben Roethlisberger 5.14%
Joe Flacco 5.11%
Michael Vick 5.11%
Colt McCoy 4.95%
Jason Campbell 4.86%
Troy Smith 4.83%
Jay Cutler 4.40%
(Kolb is next w/ 3.70%, then Cassell w/ 3.11%)

20
by Intropy :: Sun, 02/27/2011 - 3:02pm

Interesting numbers. I'm glad someone is going the work to grab them and others are using it for some further analysis. I think perhaps time at which pressure is made might be better than time of sack. It's certainly a lot more difficult to figure out what counts as pressure. But if a quarterback is hit 1 second into a play, stays on his feet, scrambles, then gives up the sack at the 6 second mark, that's on the line. If a quarterback is hit 1 second into the play, stays on his feet, scrambles, and makes a play down field, that too is on the line.

21
by tuluse :: Sun, 02/27/2011 - 4:05pm

I think this muddies things a bit because it punishes QBs who played behind really terrible lines (Jay Cutler). Whereas Matt Cassell was not sacked very often, but over 50% of them were after 3 seconds.

29
by Kal :: Mon, 02/28/2011 - 11:34am

I don't think it punishes Cutler; I think it makes it clear that some of the sack issues were because he waited too long to get rid of it. Now, he ALSO got sacked a bunch before 3 seconds, which means that he knew he was going to get hit and waited and waited anyway.

I suspect some of that is a product of a Martzian offense where you have to wait a while for long developing plays, but some of it is on Cutler.

34
by tuluse :: Mon, 02/28/2011 - 3:01pm

What about this hypothetical:

Frank Omiyale is caught flat footed, and his defender loops around him easily. Cutler sees him, and starts rolling to his right. J'Marcus Webb engages initial, but his defender eventually gets past him too, right after the 3 second mark, he catches Cutler and brings him down.

Now yes, Cutler could have thrown the ball away more and taken less sacks, but his percentage of sacks over 3 seconds compared to total sacks is almost exactly average.

25
by Bobman :: Mon, 02/28/2011 - 12:45am

The percentage stats indicate that Philip Rivers is the most average NFL QB, at least in this regard. His long rate, short rate and overall sack rate are all wihtin a gnat's ass of each average (based on the 33 QBs in dcaslin's table). I would have assumed an above-average long sack rate based on my perceptions of SD's passing game. Then again, all those injuries may have changed their game to a more balanced long/short attack....

Almost as average across the board, in terms of sack rate: Sam Bradford. Nobody else is within 1.00% of the average on two, or even all three measures, but those two guys.

7
by bubqr :: Sat, 02/26/2011 - 3:46pm

Adjuste per attempts, number of blockers and rushers, and you might have a very useful stat.

10
by takeitdown :: Sat, 02/26/2011 - 3:59pm

I'm a big fan of time to pressure stats, though it requires looking at every passing play. There are a number of QBs who seldom get over 3 seconds in the pocket (let's say pocket breaks down in under 3 seconds 80% of the time), and thus normally get rid of the ball at 2.5 seconds. They may have to hold it for 3+ due to down and distance, time of game, etc., and get hammered. Even though they get rid of the ball 95% of the time in under 3 seconds, they will show a decent percentage of 3+ second sacks.

On the other side, you have a QB whose line gives him 4 seconds most times (pocket breaks down in less than 3 seconds 20% of the time). If he generally throws at 3.5 to 4 seconds, he could end up with very similar stats, even more less than 3 second sacks (as he's used to a 4 second clock) but clearly, in general, "has time."

I'd like to see FO, or a similar group, do such a study. If the OL necessitates only 3 step drops and quick throws, they shouldn't get credit for few sacks, obviously, or for being quality, as they're hamstringing the entire offense.

The biggest key seems to be what clock the QB can establish (a 2.5 second clock means quick reads, short throws, flies while a 4 second clock means options for double moves, deep ins and comebacks, etc.)

13
by Dennis :: Sat, 02/26/2011 - 9:17pm

The flip side to the long sacks are how many positive plays the QB makes holding onto the ball for a long time. Guys like Vick and Rothleisberger seem to make a lot of positive plays in situations where other QBs might have thrown the ball away.

What you really need to look at it are the total results when a QB holds on to the ball for over three seconds - attempts, completions, ints, sacks, and scrambles.

23
by Bobman :: Mon, 02/28/2011 - 12:28am

Ugh, good point. One reason I love Peyton Manning is that he is quick to decode a lost cause and dump the ball out of bounds or find his RB, or if the RB is behind the LOS and the LBs are charging, 18 often throws it to the ground at the guy's feet, to avoid an INT or negative yardage play. This results in few "long" sacks but probably more "zero" plays (as opposed to negative plays like picks and sacks. Don't get snarky about picks, folks--half of his unusually high number, for him, happened in a 3-game span.) So Peyton is generally safe and thatis reflected in his low long sack total.

And one reason I really don't like facing Roethlisberger is the very point you bring up--sure he holds it forever, but some of those times he pulls a rabbit out of his ass, along with a deck of cards, a boquet of flowers, a top hat.... and then a sack. (And a very puzzled proctologist!) There's value to those rabbits, and it might outweigh the negative values of the sacks.

So a complete measure of whatever we're getting at here would also have to take into account the safe "zero" plays the QB executes in avoiding a sack, as well as the great plays and Favrian brain farts he commits in holding long and trying to avoid a sack at the last moment.

30
by Independent George :: Mon, 02/28/2011 - 11:49am

The thing is, the big plays can only happen if he's on the field to attempt them; it doesn't matter if you win one game on a heroic play if you're lost for three more in the aftermath.

Neither Manning has missed a start in a combined 19.5 consecutive seasons; a lot of that is luck, but a lot is also the ability to throw the damned ball quickly. Meanwhile, Vick and Roethlisberger have battled injuries, and it's only going to get worse as they age.

32
by Will Allen :: Mon, 02/28/2011 - 1:04pm

Yep, a willingness to get clobbered with regularity is, in most instances, not consistent with maximizing the number of wins in a season. A good quarterback's first responsibility is to stay on the field.

37
by Anonymous Coward (not verified) :: Wed, 03/02/2011 - 1:37pm

Spoken like someone who never played football.

Anyway a quaterback's first responsiblity is to not fumble the snap.

35
by Scott C :: Tue, 03/01/2011 - 2:34pm

Even that isn't good enough.

A QB that turns a 2.5 second sack into a 3.5 second sack isn't hurting his team.

This stat penalizes QB's who manage to buy time by being more elusive when their OL breaks down early.

This whole thing is nearly useless other than to confirm what we already knew with our own eyes:
Manning is really good at throwing the ball away before getting sacked. Rothlisberger tries to buy time avoiding defenders. Rivers usually gets the ball away, but will sometimes try and buy time and is too slow to do so consistently and take a sack.

Nothing new here.

16
by Andy (not verified) :: Sun, 02/27/2011 - 12:13pm

Though you might also need to know the relationship between pass length and time of play "development" (ie: a lot more work), it would have also been useful to incorporate how far these QBs generally passed down field. It's a lot quicker for Tom Brady to toss it to Welker 3 yards past the LOS, than it is for Philip Rivers to bomb it to Malcom Floyd down the field. 85% of the attempts @ Welker were withing 10yds of the LOS, while only 23% of those at Floyd were. Given this, I agree with AA that it would have been useful to have more than 2 catagories.

24
by Bobman :: Mon, 02/28/2011 - 12:30am

While short yardage is indicative, it's no guarantee. A 25 yards in the air pass is of course long and takes time, but a 4 yard pass to Welker could just as easily be the result of two seconds or five seconds of holding the ball--just nothing deep was available so WW comes back to the ball, or he crosses and re-crosses the field as the check-down guy and finally gets the pass once Brady has exhausted his other deeper targets.

26
by RickD :: Mon, 02/28/2011 - 1:36am

Fascinating to see the Manning brothers together at the bottom of the scale (or top, depending on your perspective).

27
by herm :: Mon, 02/28/2011 - 2:33am

Fascinating, perhaps, from a biological standpoint, but hardly suprising- Peyton has been known to be great at avoiding the rush for ages now and Eli has been similarly impressive over the last few years.

28
by justanothersteve :: Mon, 02/28/2011 - 8:23am

Interesting also that Eli's most famous pass is the David Tyree helmet catch where Eli pulls a Rothlisberger and somehow doesn't get sacked after holding it well beyond 3 seconds before chucking it downfield.

31
by Spielman :: Mon, 02/28/2011 - 12:27pm

Not by choice, though. Eli was forced by the pressure to step up at just past the two second mark, and got grabbed so that he wasn't able to do that. He was *trying* to be a totally conventional Manning brother there, but was forced to freelance by the way he was grabbed.

38
by jerry guy (not verified) :: Wed, 05/18/2011 - 7:11pm

Excellent analysis; I believe we can do more work on this subject and is should lead to a more objective analysis of the game; it is interesting how much unjustified criticism clause took.
Furthermore we could brake the time frame into more segments e.g. 3.0-3.1, 3.1-3.2 etc and document success of the play