BREAKING: Quarterbacks, And How This Affects Them Comment Count

Ace June 4th, 2014 at 3:56 PM

As it turns out, taking this kind of beating has a long-lasting effect. [Fuller]

We all know that a well-timed hit on a passing quarterback, whether or not it results in a sack, makes it far less likely that pass will be completed, not to mention more likely to fall into the hands of a defender. It stands to reason, as well, that the cumulative effect of multiple hits on a quarterback will eventually affect his performance even when he's not taking hits.

Devin Gardner's 2013 season stands as a testament to this hypothesis. After the MSU seven-sack breaking point, there were several occasions when it seemed like he simply didn't have the juice to make certain throws that he didn't have any trouble making before his body was demolished piece by piece.

A recent article from the excellent site Pro Football Focus helpfully quantifies the effect of such punishment on a quarterback's accuracy and interception rate, using a data sample of every NFL throw from the past six years. As a quarterback takes more hits over the course of a game, their accuracy predictably plummets—on every throw, not just the ones when they're eating a defender:

After every sack or hit the quarterback takes, their Accuracy Percentage decreases by an average of a half of a percent. While that might not seem like much, there also isn’t much difference between the best and worst quarterbacks in the league. Based on the graph, an above average quarterback after five hits or sacks performs as well as an average quarterback with no hits or sacks. Once that above average quarterback has been sacked or hit 10 times, they play as well as a below average quarterback who hasn’t been sacked or hit.

The article also shows that interception rates increase substantially as the hits pile up, and more experienced quarterbacks are much better at mitigating these effects than rookie signal-callers—something to keep in mind when considering the relative merits of Gardner and Shane Morris.

When looking at Michigan's 2013 season, there's no question Gardner was a victim of this phenomenon. There's also evidence that he's better than most at handling the heat:

  Att. Comp. % YPA TD% INT%
2 sacks or fewer 152 94 61.8 10.1 7.9 3.9
3+ sacks 193 114 59.1 7.6 4.7 2.6

The heroic Ohio State (3 sacks) performance skews the numbers, though in fairness, so did Indiana's defense on the other. Even so, Gardner performed markedly better when not under constant pressure, especially when it came to producing yards and points.

The oddball interception rates can be chalked up to the concerted effort by Gardner to make fewer risky plays as the season went along. The desperation throwaways that resulted in stuff like the Stephon Tuitt pick-six turned into sacks intelligently taken as the season wore on, to the benefit of the turnover margin and serious detriment to Gardner's health and ability to make big plays on the ground or through the air.

So here's the hopeful part. Imagine a world in which Michigan has a running game that can move the ball forwards, forcing defenses to respect the run instead of pinning their ears back and going full-bore for Gardner's chest. Imagine a coherent Michigan offense that finds a way to counter the constant opponent blitzes. Imagine a full season of an offense directed by an NFL-level talent who doesn't end half his games resembling a coal miner.

This could very well be Michigan's reality in 2014. If it is, expect Devin Gardner to do big things.



June 4th, 2014 at 4:16 PM ^



June 4th, 2014 at 4:26 PM ^

Pump the brakes. From where cometh this sudden burst of optimism? Not used to it.

Love love love the idea of "coherent offense." Seems like a genius idea. We should try it.


June 4th, 2014 at 4:29 PM ^

I was just looking at Gardner's B1G numbers from last season vs Braxton Miller's and considering the offenses they were in, I'd say Garder comes out pretty favorably.

Gardner(8 games):  8.85 ypa, 150/244 (61.5%), 14 TD 3 INT, 2154 yards.

Miller(9 games):  7.98 ypa, 129/207 (62%), 20 TD 4 INT, 1652 yards

Considering Gardner had no running game and poor interior OL that's pretty damn good.  Miller meanwhile had an extremely dangerous running game, 4 seniors on the OL, and often had receivers so wide open due to running threat that even his terrible throws couldn't miss them.  If Michigan can have even a decent running game, Gardner and the offense can be scary good.  


June 4th, 2014 at 4:32 PM ^

This is what I learned about Devin Gardner from watching him in 2013

  1. He is a warrior.
  2. He is confident, but sometimes a bit too reckless
  3. He is not so talented, that he can carry an entire offense all by himself.
  4. He is a warrior.
  5. What else?  Oh yeah...he is a f*#king warrior!

Michigan is capable of good things with him under center.


June 4th, 2014 at 4:53 PM ^

If I correctly understand the nature of the data presented, then the relationship between sacks and accuracy might be spurious e.g.,QBs are likely to be sacked more by a team with a good defense--eg good DBs may cause the receivers to be covered and increasing the chances of a coverage sack.   Also, QBs playing vs such teams are more likely to pass inaccurately or be intercepted e.g., again because the DBs are good.

These relations will induce an apparent correlation between the number of sacks and the accuracy/interception rates even if there is no direct relation.  As the authors note, further statistical analysis is needed (eg examining the correlation between sacks and accruracy after controlling for the quality of the defense).

I do not doubt that there will still be some correlation--the argument about the effect of hits is very plausible.  However, the correlation may be much smaller than it appears.


June 4th, 2014 at 5:28 PM ^

Yep.  What we would really like to know is whether within each game the QB's performance decreases as the hits rack up.  As you noted, the snaps that occur after a high numbers of hits are likely also disproportionately snaps taken by bad offenses or against excellent defenses, so it is impossible to know whether the performance decrease is due to the hits or the competition.  But if you examined how much performance increases or decreases across a game as hits accumulate, you would control for the level of competition.

Even if you did that, it would be hard to know whether the changes were due to the direct effects of the QB being hit, or other effects.  Performance could go down because the QB is flustered, but it could also go down because the play-calling changes after the QB is hit, or because other variables change based on the hits.


June 4th, 2014 at 6:25 PM ^

to look at performance for an individual qb within a single game (before pooling the results across different ones).   Someone could get such data by looking at a game log.  Clearly, it would be a lot more work but maybe worth it.  The topic is very important.


June 4th, 2014 at 10:23 PM ^

Read it again. Ace is talking about the course of a season. The data presented in the graph and sited article represent in game data ie. a QB's accuracy decreases over the course of a game based on the number of hits taken.

I would be more interested to know if this only shows hits in the pocket, or represents hits taken on scrambles and designed QB runs as well.


June 4th, 2014 at 11:00 PM ^

I don't think that's right.  The NFL data, at least as I read them, are just presenting the accuracy rates (or whatever other dependant variable) for snaps taken after X hits.  So, for all snaps taken after 10 hits, the accuracy rate is ~67.5%.  The problem, is, better defenses and worse offenses are probably overrepresented as the number of hits get higher, because better defenses get more hits and worse offenses are hit more often.  And so there's no way to tell whether the hits are driving the effect or the quality of teams is driving the effect.


June 4th, 2014 at 11:22 PM ^

Yes, but the point is it's within a single game, not over the course of a season.

And I'm not sure you're right here. QB A threw 10 passes against defense A and completed 7. On those attempts he was hit (cutoff X) times. On his next 10 passes against team A he completed 6. On those attempts he reache 10 hits. On his next 10 passes against defense A, he completed 5. He's playing the same defense and with the same offense. Quality should have nothing to do with it because defense A likely didn't suddenly get better over the course of 4 quarters.


June 4th, 2014 at 6:21 PM ^

Are those Gardner stats divided BETWEEN games (by opponent, depending on sack number) or WITHIN games (prior to the 3rd sack vs after)? (i.e., Were the OSU game stats included in both?)

If the former, splitting up the sample by bad defenses (more likely to have few sacks) and good defenses (more likely to have many sacks), presenting some serious correlation issues (as another commenter brought up).  QBs do worse against good defenses isn't especially insightful. (This affects the data even WITHIN games, due to sample size, but not as much as if you just differentiate by opponent.)

If the latter, the assertion that Gardner played better when not under constant pressure is spurious, unless you assume he had less pressure during the earlier part of the game (before the third sack) than the later part of it (after). (e.g., If, hypothetically, the 3rd sack happens at the half, you're counting the same opponent/game equally in both categories and therefore can't assume that the pressure was any different in the first half or the second half.)

The study is focused on how QBs change within games.  To me, Gardner was more affected between them, as the hits took their toll over the course of the season he changed how he played.

I would certainly agree that if Michigan has a strong run game this year, Gardner will put up big stats.  I just don't see how a new coordinator paired with subtracting Lewan, Schofield, Gallon and Toussaint would create said run game, but IN NUSS WE HOPE!


June 4th, 2014 at 6:25 PM ^

QBs should be allowed and even encouraged to get the ball out to another offensive player irrespective of where a target recipient is situated relative to the line of scrimmage.  Bounce that baby off the carpet (but not your mother's carpet), forward pass, or even nib it soccer-style and cease with the counterproductive forward throw bias.  Abolish the the scrimmage line except for lining up.  Tear down your rulz, Mr. Gorbachev NCAA.  A thought, perhaps a bad one.


June 4th, 2014 at 6:50 PM ^

I totally agree that DG is a good QB, especially if the running game can move the ball forwards, and I totally believe the talent is there, but I don't see it happening in 2014.  Looking at the o-line there's only one guy coming back that was a positive contributer on the line...Glasgow. Everyone else by my record was a liability last year.  I am optimistic that Kalis and Mags with another year of experience, age, and training will be solid, but I'm not so sure on the other guard and tackle position regardless of who plays there this year....they still seem another year away.  So, if we have one good guy, two servicable guys, and two liabilities I don't see much optimism for a quality running game, but that's just me.  So, why should we be optimistic that the running game will get going?

gustave ferbert

June 4th, 2014 at 6:47 PM ^

I would be interested to see if there are any quarterbacks in the sample that broke the trend and got MORE accurate as the game went on and he took more hits.  I'm sure Brady did it! 

Marley Nowell

June 4th, 2014 at 7:34 PM ^

This title scared the living shit out of me.  I thought Devin was hurt again and out for the season or something.  DON"T USE ALL CAPS UNLESS IT IS SUPER IMPORTANT

Gentleman Squirrels

June 4th, 2014 at 11:34 PM ^

Langford took till game 4 of last season to really get going. Its not just about the running game getting going though. Its also about how well and how fast the o-line gells into a coherent unit. If they can do that, I like our chances of having a good running game with Green and Smith (and Hayes and the rest of the RB rotation).


June 5th, 2014 at 10:45 AM ^

I almost thought I was on the wrong site, between Ty Isaac and this it almost seems as though us fans still have a reason to live suddenly or something.

I wish that Gardner could be The Wolverine, with adamantium skeleton and super fast healing.  Then I would be really confident about the season.  I hope we can blow out some of the early teams/carpets and give Shane time, to both rest Gardner and have a backup with experience.


June 5th, 2014 at 1:00 PM ^

Most NFL teams average 30-40 pass plays per game. So a 5% drop in accuracy amounts to < 2 fewer completions per game. At the NFL average of 10-12 yards/completion, that's not much impact.

What other effects were considered? Quality of defense affecting hits/game at the edges seems to matter. Plus, it usually takes most of a game to get to 8+ hits on a QB - what's the plot of accuracy per quarter look like? At some point you'd think fatigue / end of game heroic attempts would drive down accuracy at the end of games anyway, regardless of hits.