Writers of Pro Football Prospectus 2008

27 Oct 2005

To Blitz or Not to Blitz

Over at PROTRADE's Insight section, Gabriel Desjardins takes a look at the question that we've been wanting to answer here for some time: does blitzing work? (PROTRADE has Stats Inc. data I don't have, in case you are wondering why we've never answered this question.) The answers seem pretty obvious: blitzing on first down leads to more sacks and interceptions and fewer rushing yards, but more first downs and more passing yards, in particular yards after catch. Looking at third downs, the evidence seems to show that blitzing helps on short-yardage downs and hurts on longer-yardage downs.

(Late add: it turns out that there was some confusion about methodology, cleared up in the comments below.)

Posted by: Aaron Schatz on 27 Oct 2005

16 comments, Last at 30 Oct 2005, 1:31pm by Ryan Mc

Comments

1
by pawnking (not verified) :: Thu, 10/27/2005 - 5:36pm

Aaron, can you do a similar article using DVOA? I found myself frustrated reading this article. It was a really good idea, and there was some attempt at differentiating between plays and field position, but I had problems with it. First, the one you alluded to above. Second, the line Long passes are actually more likely when an offense is blitzed, which implies that quarterbacks don't just look for any nearby open man when they're under pressure. Is "Long passes" the total yards gained on a play? Isn't it possible that a short pass becomes a long gain because there are fewer defensive players to tackle the receiver after a blitz? Desjardins doesn't consider this. Third, I would greatly prefer such analysis based on the success or failure of plays with or without a blitz, which are not clearly defined in this article.

I think the realevant stats should be:

1) how often would a blitz prevent a successful play?
2) On average, how succesful are plays from blitzes in terms of yards per play? In terms of TDs per play?
3) How does field position change the above variables?
4) Are some teams consistetly better at blitzing than others?
5) What does all of this mean to Greg Easterbrook?

2
by JonL (not verified) :: Thu, 10/27/2005 - 5:48pm

This may be similar to no. 1, but it might be noteworthy to see the ratio of passes thrown away/intentional grounding penalties vs. simple incompletions caused by blitzing.

I thought the examination of TD rate based on field position was interesting, although I'm assuming the % at 55 and 80 yards out are statistical outliers.

3
by Yuri (not verified) :: Thu, 10/27/2005 - 6:02pm

I don't want to re-state the obvious, but like in choosing your financial portfolio, or in playing poker, the key is variety. Let's say you studied film and determined that vs. team X's offense blitzing on every down is the best strategy (they can't handle the blitz!). Once you do it a couple of times--nit only will prepare the offense for what to expect but also make them BETTER in reacting. (similar to tennis, if your opponents keep hitting to your weak backhand, it will improve).

The counter-argument to this is that 1) There are a huge number of situational splits, just like DVOA does it, by down, field position, yards to go, score etc. You strategy ned not be "blitz every time" but
"blitz every time in situation X,Y,Z and not blitz in A,B,C"--sort of an "advanced rule of thumb" which is not immediately obvious
2) In a similar sutuations the offense will come out with diferent personnel and/or audible to a different play
3) There are a huge number of blitzes that can be used by position or number of players blitzing.

Ultimately yes, I support the calls for the Blitz strategy analysis using DVOA. But once it produces "rules of thumb" when to blitz against whom they will be known, and therefore not useful :)

4
by stan (not verified) :: Thu, 10/27/2005 - 6:05pm

I haven't even read it yet, but the point of defense is to get the ball back. If I get burned for first downs the first two times I blitz, but the 3d one gets a sack which the offense can't overcome and punts -- defense has won and blitzing worked.

The relevant factor is which strategy gets the ball back.

5
by MJK (not verified) :: Thu, 10/27/2005 - 6:10pm

I would imagine the problem with trying to use DVOA for to see if blitzing works is that Aaron doesn't have data that says when a team blitzed? Is that correct? I know Aaron can tell us whether any given play by any given offense was a "success" or not, but is "the defense blitzed" versus "the defense only rushed 4" in the data that Aaron has access to?

6
by MJK (not verified) :: Thu, 10/27/2005 - 6:18pm

Another problem is the question "does blitzing work?" I'm sure you could probably predict the probability of "getting the ball back" for a case when you blitz on every down (or even on every passing down) versus the probability when you NEVER blitz. But blitzing is like running a draw or a screen-it's most effective when the defense doesn't expect it. Therefore, the most effective strategy is probably to blitz sometime and not others--this probably has the "always blitz" and "never blitz" strategies beat (this is a major problem I have with Easterbrook's outright condemnation of the blitz).

The problem is, if you define success based on whether you get the ball back, they way stan suggests (which I think is a really good idea), predicting how likely a "sometimes blitz" strategy is to work gets really difficult (visions of game theory and Nash equilibria flash through my mind), unless the blitzing strategy is essentially to use dice or a dartboard.

I mean, it's easy to say "blitzing gives up X yards per play versus non-blitzing that gives up Y". But it's a lot harder to say that "your probability of getting the ball back is P if you use this blitzing strategy, whereas it is Q if you use this one".

7
by James (not verified) :: Thu, 10/27/2005 - 6:56pm

This article didn't answer any questions that I had.

I think about blitzing alot and this is what I've always thought.

1. Like the guy above defense should get the ball back
2. You want the offense to get in a 3rd in long before they have good enough field position to pin your team deep.
3. Get pressure from the qb can't see and cause a fumble if no. 3 fails.

side note- why do blitzers go after the quarterback as if the only way to make the play fail is to knock him out

How about gathering yourself and at least causing a bad pass. Why not treat the qb like an open three point shooter that you are allowed to foul. Make his throw inaccurate first(this causes picks with enough bad throws) and if you get to him just grab him. Once a guy is within 3 yards of a qb its inexcusable for a pass to be completed. Whatever you do if you miss you gave up a big play because a good qb is now out of the pocket with recievers who have had 3-5 seconds to beat coverage.

8
by bobman (not verified) :: Thu, 10/27/2005 - 7:13pm

Isn't this a situation-specific issue, like running or passing? Do you run and expect it to work on the top run D? Not really; You might run to keep balance and free up passing chances. But at the end of the day, you look at the stats and say "we got only 88 yards rushing after all that effort, man, what a failure."

Same with blitzing--do you blitz aginst top-flight QBs like Brady or Peyton? Not unless you want to get burned. Do you take on top OLs, or a RB who has a good reputation for picking up the blitz? Could just bite you. But do you blitz Carr in Houston, Orton in Chicago, Smith in SF? (and maybe Simms in TB this week... we'll see.) Yes, yes, and yes. And you will succeed.

Aaron's caution on the methodology seems sound too--hard to make an assessment when your control group is not really a control group.

9
by Gabe Desjardins (not verified) :: Thu, 10/27/2005 - 8:22pm

hi guys - there was a small editing hiccup on the article that led to some confusion, but it's fixed now.

The two situational blitz/no blitz comparisons in the article are:

a) 1st and 10
b) 3rd and varying yardage

In each case, the control group is "no blitz" in the same situation.

I don't compare 3rd down to 1st down (that was the sentence that was unclear in the original version.)

To answer one of many questions: "Is “Long passes� the total yards gained on a play? Isn’t it possible that a short pass becomes a long gain because there are fewer defensive players to tackle the receiver after a blitz?"

Yards/Pass is just how far away the receiver is when he catches the ball. YAC is yards after catch. I broke the play into both components -when the defense blitzes, YAC goes up for 3rd and 10 situations.

I didn't get a chance to look at the success or NES rates for these plays, but that would make a good follow-up article.

Ben Alamar has an article that discusses QB performance against the blitz:

http://www.protrade.com/insight/InsightArticleAlamarQBsUnderPressure.htm...

10
by Gabe Desjardins (not verified) :: Thu, 10/27/2005 - 8:24pm

shoot. Some of my comment got eaten by HTML tags.

Yards/Pass is just how far away the receiver is when he catches the ball. YAC is yards after catch. I broke the play into both components - when the defense blitzes, YAC goes up for 3rd and less than 10, while the actual pass length goes up for 3rd and more than 10 situations.

11
by rk (not verified) :: Thu, 10/27/2005 - 9:31pm

RE: 7
I understand your general statement but I think it's a little ridiculous to say that a QB shouldn't complete a pass if a guy is 3 yards from him. A lot of times the pocket doesn't even have a 3 yard radius. Your analogy to basketball makes some sense, but if a shooter can't make a shot when he has 9 feet of open space, he'll never make a shot.

12
by Michael David Smith :: Thu, 10/27/2005 - 9:38pm

In general, I agree with those who say the best thing about the blitz is the way the offense doesn't know when it's coming. I don't think fans pay enough attention to that aspect of strategy -- keeping your opponent guessing. I was watching a game in a bar one time when a team successfully ran a fake punt. I heard another guy in the bar say, in all seriousness, "That fake punt worked really well. They should run it more often."

13
by pawnking (not verified) :: Fri, 10/28/2005 - 10:29am

I have to say that one of my favorite things about this site, and generally the links it provides, is that you respond to feedback. It makes me feel like a small part of the process of enjoying football more. Gabe, thank you for addressing some of these issues we've raised.

14
by stan (not verified) :: Fri, 10/28/2005 - 11:23am

Another problem with analyzing whether a blitz "works" is the fact that blitzing is often used to stop the run.

If you blitz to stop run (because you are weak against the run with a base front) and the offense passes successfully, do we "blame" the blitz call or the basic weakness against the run?

This is the fundamental analytical problem. I once had a defense which had a quality secondary and horrible defensive line. We blitzed 80% of the time. It was the only way we could stop the run. Once in a while we got burned by the pass. If someone looked at our stats and concluded we were better against the run than the pass, their understanding would have been exactly backward.

The best way to minimize the damage an offense can do to you is to protect your weak links and ask your strongest players to step up. Often the strongest players will appear to be the ones getting beat, but that is only because they aren't getting any help.

You have to understand WHY.

Look at McNair and Culpepper. When McNair (in 2003) and Culpepper (in 2004) had all day to throw and never had to worry about hot reads, etc., they put up great numbers. When their pass pro degraded, they looked like chumps. Did they get worse? No, their teammates did (and they really weren't that good to begin with). You gotta know WHY things are happening.

15
by Dennis (not verified) :: Fri, 10/28/2005 - 12:25pm

So did someone send this link to TMQ?

16
by Ryan Mc (not verified) :: Sun, 10/30/2005 - 1:31pm

I'd like to see stats on how often QBs get sacked after executing a play action fake on an obvious passing down. That's a play call that drives me crazy, because who seriously thinks you're going to run?

I yesterday's Georgia-Florida game, Ga faced 4th and 10 late in the game when trailing and had to go for it. They had the QB execute an elaborate play fake before setting to throw. Absolutely nobody on defense was confused that it might be a run in that situation and all it meant was that for the first three seconds of the play the QB had his back to the action, wasn't watching his receivers run their routes and had no idea if there was a blitz on or not. By the time he set to throw there was a guy right in his face and the ball simply clunked to the ground incomplete.

A fake only works if you fake the thing that the defense actually thinks you're going to do.