Thursday, December 30, 2010

Examining QB "Yards after Catch"


by Adam Tarr

One stat that remains frustratingly difficult to find is Yards After the Catch, or YAC. Unless you're reading a gamebook from a Buffalo Bills home game, it's rare that you will find any recorded information about a game that details how much distance each pass travelled through the air, and how much distance was covered by the receiver after the catch. This information is significant, because as Brian has mentioned in the past, the evidence suggests that receivers deserve the lion's share of the credit for YAC.

Season-by-season YAC numbers for receivers can be found for all active receivers on a few sites, but YAC numbers organized by quarterback are even harder to find. The best I've found is stats.com, which lists the top 20 QBs by YAC in each conference for each season from 2006 to 2010.

That's pretty slim pickings, but I decided to make the best of it and see what I could mine out. I took every rate stat-qualified QB this year, plus any other QBs that had 1000+ attempts in the 2006-2010 span, and attempted to merge in the YAC data with the rest of their stats. The regular stats won't line up perfectly in this case, because I didn't include any numbers for any QB from a year where they didn't finish in the top 20 in their conference in YAC (Tom Brady in 2008, for instance), but it's pretty close. Anyway, here are the raw passing totals and YAC totals for the 39 QBs that qualified for the study:









PlayerYards, 2006-2010YAC, 2006-2010
Drew Brees2272210636
Brett Favre182239233
Philip Rivers192008903
Peyton Manning213758577
Donovan McNabb168178094
Tom Brady164347636
Tony Romo166507306
Eli Manning175987240
Ben Roethlisberger172167143
Jason Campbell130926677
Jay Cutler147956491
Carson Palmer156566438
Matt Schaub141716365
Matt Hasselbeck136545949
David Garrard141955929
Kurt Warner131305779
Aaron Rodgers121655479
Kyle Orton104274902
Matt Cassel96184807
Jon Kitna106414794
Joe Flacco100814727
Vince Young105914615
Marc Bulger94134170
Jake Delhomme89803930
Matt Ryan98253777
Chad Pennington87703752
Alex Smith73343730
Derek Anderson83553294
Jeff Garcia64613009
Ryan Fitzpatrick63272799
Chad Henne61082586
Mark Sanchez57352381
Shaun Hill54172371
Kerry Collins54242330
Michael Vick54922240
Josh Freeman50512116
Sam Bradford33571812
Matthew Stafford22671236
Jimmy Clausen1376757



Most fans could probably have guessed the top nine correctly (at least if they remembered who had a job in 2006), but it's interesting to see which QBs are more reliant on YAC than others. Some QBs get more than half their totals from YAC, while on the other extreme you have QBs like the Manning brothers who rely on the vertical passing game.

Pro-football-reference.com puts together all sorts of tweaked "yards per attempt" stats. I decided to use the data above to cook up my own, cribbing off the ideas Brian used in his old revised passer rating article.

My formula for "Normalized, Optimized, Air Yards Per Dropback" is:

NOAY/DB = (Passing Yards - Sack Yards - YAC - 32*INT)/(Attempts + Sacks) + 3.9

The 32 is in there because that preserves the ratio of INT and yardage Brian used in his passer formula. The 3.9 is in there because it gives the average 2010 QB a NOAY/DB of roughly 6.2, which is the average yards per attempt this year.

Here is the table of those same 39 players again, this time sorted by NOAY/DB, with their traditional Y/A numbers also listed for comparison:









PlayerNOAY/DBY/A
Peyton Manning7.397.54
Tom Brady7.077.7
Matt Ryan7.016.9
Tony Romo6.988.04
Aaron Rodgers6.957.98
Philip Rivers6.958.04
Jeff Garcia6.917.25
Matt Schaub6.897.86
Drew Brees6.837.64
Kurt Warner6.747.59
Chad Pennington6.617.18
Ben Roethlisberger6.567.77
David Garrard6.557.16
Carson Palmer6.557.02
Michael Vick6.447.23
Donovan McNabb6.417.39
Eli Manning6.46.9
Vince Young6.366.92
Jay Cutler6.357.28
Kyle Orton6.356.93
Kerry Collins6.336.23
Josh Freeman6.326.84
Joe Flacco6.257.22
Derek Anderson6.156.33
Chad Henne6.126.6
Jake Delhomme6.056.83
Mark
Sanchez
5.996.58
Marc Bulger5.996.69
Brett Favre5.987.12
Matt Cassel5.986.74
Jon Kitna5.967.21
Matt Hasselbeck5.926.58
Jason Campbell5.846.74
Shaun Hill5.826.61
Ryan Fitzpatrick5.666.08
Alex Smith5.56.51
Sam Bradford5.396.06
Jimmy Clausen4.475.17
Matthew Stafford4.456.01



The top of that list pretty much speaks for itself, but this list does contain some surprises for most fans, such as Favre's location near the bottom.

Hopefully someday play-by-play data will include yards in the air and yards after the catch as a matter of course, and we will be able to use this data to more precisely separate out QB and receiver contributions on individual plays. Until then, we're left making wide-angle analyses such as this one.

39 comments:

James said...

I'd like to see the percentages of YAC/Total Yards and Y/A - NOAY/DB without having to do it manually or in my head for easy comparisons. For instance, despite having similar NOAY/NB, Alex Smith is clearly more reliant on YAC than Ryan Fitzpatrick. This is at least partly because Alex Smith had Frank Gore and Vernon Davis to run after the catch for him, but it's still good to note.

Tarr said...

I sort of cringe at the idea of doing percentage breakdowns between NOAY/DB and Y/A. NOAY/DB is sort of a fake stat, and looking at the list there makes me think I screwed up the normalization anyway (almost everyone is lower in NOAY/DB than Y/A).

Percentage Air Yards is a nice unitless stat, though, and easy for me to get at. Here it is, in ugly unformatted form:

55.0% Jimmy Clausen
54.5% Matthew Stafford
54.0% Sam Bradford
51.0% Jason Campbell
50.9% Alex Smith
50.7% Brett Favre
50.0% Matt Cassel
48.1% Donovan McNabb
47.0% Kyle Orton
46.9% Joe Flacco
46.8% Drew Brees
46.6% Jeff Garcia
46.5% Tom Brady
46.4% Philip Rivers
45.1% Jon Kitna
45.0% Aaron Rodgers
44.9% Matt Schaub
44.3% Marc Bulger
44.2% Ryan Fitzpatrick
44.0% Kurt Warner
43.9% Tony Romo
43.9% Jay Cutler
43.8% Shaun Hill
43.8% Jake Delhomme
43.6% Vince Young
43.6% Matt Hasselbeck
43.0% Kerry Collins
42.8% Chad Pennington
42.3% Chad Henne
41.9% Josh Freeman
41.8% David Garrard
41.5% Mark Sanchez
41.5% Ben Roethlisberger
41.1% Eli Manning
41.1% Carson Palmer
40.8% Michael Vick
40.1% Peyton Manning
39.4% Derek Anderson
38.4% Matt Ryan

Tarr said...

It's worth noting, perhaps, that having tons of YAC as a QB, or even a very high percentage of YAC as a QB, doesn't mean you're doing anything wrong. It can be an effective use of the talent on your roster. You can hardly blame Tom Brady for giving the ball to Welker and Woodhead in space and letting them create yards. And QBs like Brees and Brady are near the top of the rankings even after we take out the YAC that makes up a large share of their totals.

The point, really, is that once the ball is in the possession of the other skill position player, the QB really has very little impact on what happens. We don't give the QB any credit for what happens after he pitches the ball to a running back on the edge, so why should he get credit for what happens after he throws it to him in the flat?

Ian Simcox said...

Tarr - easy counter example. Say a QB does a massive pump fake, gets the CB to bite, throws it up for his now-open WR who takes it in stride and runs in untouched for the TD.

The QB may have no direct effect on the YAC after he's thrown it, but through the pump fake he bought the receiver the opportunity to gain all that lovely YAC.

I would say that a QB's job is to find the receiver who offers the greatest opportunity to gain yards. If that involves a little dink followed by tonnes of open field YAC, then that's hardly a bad decision by the QB.

Anonymous said...

I agree with Ian. I find it hard to believe that no QBs have the skill of being able to find the receiver with the most open field ahead of him.

Favre, for example, has had many different receivers among the YAC leaders over two decades. And that was all just luck? Or did Favre help them gain YAC? When Rodgers (another excellent QB) took over, the same receivers that played with Favre suddenly had fewer YAC.

Tarr said...

Did they have fewer YAC, Anonymous? I'd like to see the study. A quick look sees that Driver has beaten his career average in his years with Rodgers, even if you count this year where he's clearly lost a step. Jennings is about the same as in the Favre years. Those are the only two guys that we have a decent amount of data for on both sides of that transition. Who's the guy who "suddenly had fewer YAC"?

Or look at Wes Welker's career. He had his best-ever YAC year when Matt Cassel was the QB for the Pats. Is that because Cassel is better at fakes and "hitting the receiver in stride" than Brady? Please. Heck, his second-best YAC came in the year when GUS FREROTTE was throwing to him. Try to explain that one by giving credit to the QB - I dare you.

I realize that this is one of those classic cases where our perception flies in the face of what the numbers tell us. People continue to believe that there's such a thing as a clutch hitter in baseball or a hot shooter in basketball, despite study after study that doesn't find any link.

Again, just read the old article from Brian that I liked to in the first paragraph of this article. The evidence strongly suggests that YAC is a function of the receiver (or, at most, a function of the usage patterns of the receiver) and not a function of the QB throwing to him.

The burden of proof here is really on those who want to demonstrate the link, not on those who say the link doesn't exist.

Tarr said...

Ian, one last note: I fully agree that there's nothing inherently wrong with a dink-and-dunk offense if it's successful. There's also nothing wrong with a heavy run-based offense if it's successful. But in both cases, it's not the QB who deserves the lion's share of the credit for the success of that attack.

Anonymous said...

Tarr, consider the 2007 vs. 2008 Packers. Same coaches, same system, same receivers. Identical offenses except for the QB.

The top four receivers all dropped significantly in YAC (stats via Yahoo): Driver 5.3 to 4.9; Jennings 7.4 to 4.8; Jones 4.7 to 3.1; Lee 6.0 to 3.3. That works out to nearly 400 yards fewer YAC with Rodgers than with Favre. Is receiver YAC usually so volatile from year-to-year?

Meanwhile, the 2008 Jets and 2009 Vikings saw an increase in YAC with the addition of Favre.

Rodgers has clearly improved since 2008 by pretty much any measure. Anecdotally, he has become less reliant on the long ball and has mastered the slant, which has been a staple of the GB offense for years and often results in large amounts of YAC. He's become close to what Favre was in his prime, in my view.

I totally agree on clutch hitting and streak shooting (and I'll throw in DIPS for good measure). But I don't think the evidence is quite so convincing on QB YAC. Certainly the receivers deserve much of the credit for YAC, perhaps even most of the credit. But I think more research needs to be done before we can conclude that they deserve *all* of the credit.

James said...

Anonymous, from what I remember there actually is a large variation in YAC from year to year.

Tarr said...

Anonymous, you're cherry-picking there, and if you've actually looked at the stats as much as your post would lead me to believe, you must be aware of that. For instance, yes, Driver's numbers dropped, but they were still the third-highest of his (at that point) 10 year career. In other words, Rodgers' performance in his first year as a starter was "better" than most of Favre's years from the perspective of "producing" YAC.

Meanwhile, I'm not sure where your assertion that "the 2008 Jets and 2009 Vikings saw an increase in YAC with the addition of Favre" comes from. Just pulling up the leading Jets receiver that year, Cotchery's YAC dropped significantly when Favre arrived. It's possible Favre put up more YAC than previous QBs simply by throwing more, but I don't see any evidence that his receivers were suddenly getting more YAC/reception.

As I said, the onus is really on those who want to demonstrate the link. Statistically speaking, the "null hypothesis" here is that YAC is a random stat. It's trivially easy to disprove this when it comes to receivers - some receivers (typically RBs and other underneath route runners, but also some big play threats like Desean Jackson) consistently put up higher YAC totals than others. There's a lot of year-to-year variation, but some receivers definitely have higher averages.

On the other hand, once you've accounted for the receivers, there's very little evidence that the QB is also making a significant contribution on a per-catch basis. Some QBs will get the ball to a receiver more often, allowing them to put up higher totals, but on a per-catch basis, the evidence is definitely lacking.

Once again, I'm not suggesting that the check-down for 2 yards to a receiver who runs for another 8 yards is a bad play. The question is whether the QB deserves the same share of the credit there as he would for completing a 10 yard pass downfield, or whether it's closer to the credit he gets for pitching the ball to a running back that runs for 10 yards.

Anonymous said...

The numbers posted on the Green Bay receivers weren't intended to be a complete study proving the existence of QB YAC as a skill. They simply were a counter-example that seems to contradict the assertion that no QB has any control over YAC. If you have a better explanation why those receivers had such a dramatic drop in YAC in identical offenses (except the QB) please share it. If team receiver YAC really does commonly fluctuate that much year-to-year even with the same QB, let's see the evidence.

See, I think when you make a contention that is so at odds with conventional wisdom ("QBs deserve no credit for YAC"), the onus is on you to provide convincing evidence. And then let others reproduce your findings independently.

Unfortunately, until reasonably complete data on air yards are publicly and readily accessible, either as cumulative totals or play-by-play form, I don't know if that is even possible. It's certainly an interesting theory, and I'm not dismissing it completely, but I don't think the burden of proof has been met to make the theory widely accepted, even among hardcore analytic geeks.

Anonymous said...

"Once again, I'm not suggesting that the check-down for 2 yards to a receiver who runs for another 8 yards is a bad play."

But you're implying that it's a worse decision than completing it 10 yards down the field to a double-covered receiver who is tackled instantly. That would seem to assume that the QB has no idea whether the check-down receiver has any more room to run than the receiver downfield. Are QBs really incapable of making such judgments?

Tarr said...

"If you have a better explanation why those receivers had such a dramatic drop in YAC in identical offenses (except the QB) please share it."

There is no such dramatic drop observed, so I don't have any need to explain it. You gave a few isolated stats that pointed in that direction, but you did not show a consistent pattern. It was trivial to find counter-examples.

"I think when you make a contention that is so at odds with conventional wisdom ("QBs deserve no credit for YAC"), the onus is on you to provide convincing evidence."

Again, for the fourth time, just read the original article from Brian. Whether you find it convincing is up to you, but the fact that accuracy and YAC are completely uncorrelated is pretty astounding.

At any rate, when I say the burden of proof is on those who want to provide a link, I mean that in a scientific/analytical sense. The baseline assumption is that a stat is not meaningful until one can demonstrate its meaning. YACs status as a meaningful stat for a receiver is very easy to demonstrate.

I recognize that, if I want to actually convince people to abandon conventional wisdom, I need to present a convincing case. Some people will see what Brian wrote and will suspend their belief in the conventional wisdom. Other people will see Welker getting more YAC/reception with Cassel and Frerotte than with Brady and will be convinced by that. But other people are simply going to believe the conventional wisdom even if the evidence points the other way. I don't deny that.

We have what appears to be a stat that follows the career arc of receivers fairly consistently, albeit with a lot of random variation from year to year (one huge play can have a pretty significant impact on a player's average).

It's very very hard to delve much deeper here, but there are some avenues for people with a lot of patience. The year-by-year YAC/reception numbers for receivers are available online. You could mine out all of those stats for every active receiver, removing years where they played on multiple teams or with multiple QBs. This gives you a very large data set, and the question of whether a QB transition or team transition has a significant effect on a receiver's YAC could be addressed with that data.

Tarr said...

" "Once again, I'm not suggesting that the check-down for 2 yards to a receiver who runs for another 8 yards is a bad play."

But you're implying that it's a worse decision than completing it 10 yards down the field to a double-covered receiver who is tackled instantly."

No, I'm not implying that at all. I'm implying that the QB deserves less credit for the success of the play. That's very, very different.

A pass completed to a receiver 10 yards downfield who is tackled instantly was usually pretty well-covered. That's often a difficult pass to make that requires a lot of accuracy. By contrast, the two-yard pass is often to a wide-open back in the flat, who is then expected to make a defender miss in space in order to pick up the yards.

There's nothing wrong with a QB taking the easier pass and relying on the skill of a teammate. They just deserve less credit for the outcome, is all.

"That would seem to assume that the QB has no idea whether the check-down receiver has any more room to run than the receiver downfield. Are QBs really incapable of making such judgments?"

Of course they are capable. But we only measure QB decision making indirectly, not directly. There's nothing wrong with a QB audibling to a run play when they see a soft front - that's often an excellent decision by the QB. But the QB receives zero credit there.

Anonymous said...

"the fact that accuracy and YAC are completely uncorrelated is pretty astounding"

I never found this particularly surprising. Comp % and QB YAC don't really measure the same thing, and both metrics already have a lot of noise with respect to the skills they are trying to measure.

In general, comp % is more about physical accuracy but isn't adjusted for distance or receiver catching skill or defense. QB YAC is more about mental decision making but isn't adjusted for receiver running skills. There probably is an accuracy component to QB YAC, but I'd argue it's more about knowing where to throw the ball (reading defenses) than it is about pure physical accuracy. Sure, some QBs have good physical and mental skills, but others are weaker in one area than the other.

I certainly think this is an area of research worth exploring further. I'd love to see year-to-year correlations of QB YAC and team YAC as a percentage of total passing yards.

BTW, I know for a fact that the NFL compiles complete YAC data on a play-by-play level because I've seen the sheets that list it. That the league chooses not to publish this data is a great disservice to its fans, in my view.

Tarr said...

100% agreed on that last point, Anon.

Comp % and QB YAC absolutely don't measure the same thing. That's not really up for debate, nor is it the point. But there's also no debate that completion percentage is:

a) A meaningful stat that measures an intrinsic skill of the QB (as opposed to just something extrinsic, in the QB's environment), and

b) Is strongly related to the ability to put the ball into a precise spot.

If "hitting the receiver in stride" is a driving factor in producing YAC, then we would expect to see some correlation here. The fact that we don't see that correlation means that either this is not a driving factor, or other factors are actually working counter to this. The first of the two seems much more intuitive, though.

Additionally, we know that:

a) Short passes consistently produce more YAC/reception, on average, than long passes.

b) Individual receivers tend to have very strong year-to-year correlation in their YAC, and variations in the overall production of their QB don't consistently influence their YAC/reception numbers one way or the other.

When you put all of these points together, it presents an extremely strong case that YAC is a function of the type of passes a QB throws and the receivers they are throwing them to, as opposed to the skills/talents of that passer. Does the data above completely exclude the possibility that QBs can consistently influence YAC by decision-making or somesuch? No, but the evidence certainly hasn't been seen.

If year-to-year QB YAC correlations were very poor, that would certainly go a long way toward disproving the notion that QB YAC is a skill, but I would be very surprised to find that. I would expect to find high year-to-year QB YAC correlations, but that would not really prove anything. I would expect them to be high. This just means the receivers and passing system stay consistent on most teams.

(By the way, we could could calculate those correlations already, using the stats.com data. We get a limited data set but it's better than nothing.)

The best way to really get at this, in my opinion, is to do a regression model for YAC that includes BOTH the QBs AND the receivers as independent variables, and looks at a very large data set (e.g. all QBs and receivers over 5+ seasons). This could show us how much of the variance in YAC is accounted for by the QBs after we have already accounted for which receivers they are throwing to. If the QB variables fail to meet significance, that would pretty much drive the nail home.

(Additionally, I'd probably want to throw Yards before catch in the model as well. So it would be a regression model to predict YAC based on QB, receiver, and yards the ball travels in the air.)

Anonymous said...

I think you are overstating how well comp % measures pure QB accuracy. There's a lot of noise in there due to other factors. And recall that comp % is fairly strongly connected to the team, as p-f-r showed when looking at mid-season QB changes.

I also would disagree that the type of passes a QB throws is independent of his skills. My guess is that they are very closely related.

I would probably agree that YAC is more strongly a function of the receiver than the QB, and that there might not be much variation among NFL QBs in their ability to generate YAC. But that's not the same as saying *every* QB has *no* influence on YAC.

Tarr said...

"I think you are overstating how well comp % measures pure QB accuracy. There's a lot of noise in there due to other factors. And recall that comp % is fairly strongly connected to the team, as p-f-r showed when looking at mid-season QB changes."

I said it's strongly related, not that it's the only factor. I don't think you disagree.

Moreover, since the short passes that tend to produce more YAC, such as check-downs to RBs, are also completed at a higher percentage than other passes, there is a _built_in_bias_ that would tend to produce a positive correlation between completion percentage and YAC. This, to me, is what makes the complete absence of that correlation so surprising.

"I also would disagree that the type of passes a QB throws is independent of his skills. My guess is that they are very closely related."

I don't disagree with that... what made you think I did? I never said that. There's a reason that David Carr checked down all the time - it's because he was bad at the deep ball. QBs select their receiving options, and offenses design their receiving routes, based on the skills of the QB. That's inherent to this analysis - otherwise we wouldn't expect to see any variation in QB YAC/reception at all!

"I would probably agree that YAC is more strongly a function of the receiver than the QB, and that there might not be much variation among NFL QBs in their ability to generate YAC. But that's not the same as saying *every* QB has *no* influence on YAC."

You seem to be pretty much agreeing with what I've been saying the entire time while setting up a straw man to disagree with. I've said that receivers deserve the lion's share of the credit, and that we have no evidence that the QB's impact extends beyond simply choosing to throw to athletic receivers on YAC-friendly routes. That doesn't mean that it's impossible that QBs can influence YAC in other ways. It just means that we have no compelling reason to believe they can.

Obviously, it's easy (as Ian did early in this comment thread) to come up with examples where a perfect pass creates a large YAC opportunity that a slightly less accurate pass would not have. We know such plays happen. The question then becomes whether the incidence of those plays as a percentage of the QB's completions happen at a statistically significantly different rate for different QBs. At the moment, the evidence seems to point to no.

Anonymous said...

"we have no evidence that the QB's impact extends beyond simply choosing to throw to athletic receivers on YAC-friendly routes."

I don't think the decision to throw to receivers with separation is simple at all. You seem to be dismissive that it's an important skill. Unless I'm misunderstanding, your argument seems to hinge on the assumption that QB YAC would be all about physical throwing accuracy.

Perhaps people tend to overlook decision-making skills as a crucial component of being a good QB. They suppose if a guy can hit a dime eighty yards away he must be a good QB. If that were true, Jeff George would be a Hall of Famer.

JMM said...

The other variable is the type of offense, which is a function of the talent. A ten yard come-back or out route has less opportunity for YAC than a ten yard slant.

Some offenses or QB's only seem to throw passes to receivers when both numbers are visible. Others, like the Greatest Show Rams, are based on timing.

Tarr said...

JMM, that's absolutely true, however, the way a receiver is used (and, consequently, the way the offense is designed) is a function of his talents. Nobody is about to send Welker out running fade routes.

If we flip that around to the QB, what the data appears to be saying is that there isn't really a discernible talent to throw the sorts of routes that are YAC-friendly. Throwing out into the flat is a relatively easy pass, after all. And completing the deep post is probably a similar skill set to completing the deep out, despite vastly different YAC opportunities that result.

Tarr said...

"I don't think the decision to throw to receivers with separation is simple at all. You seem to be dismissive that it's an important skill."

Close. I think that the ability to pick the receiver that has the best chance to create a positive play is obviously a crucial skill. What's missing is a logical step from there, to "this QB can make a receiver running a given route, open by a given amount, get more YAC per completion than the average QB can". Note that that's YAC per COMPLETION, not attempt. The more accurate thrower is already going to get the ball into the receiver's hands more often on a given route.

Let me try to boil it down to an example. Say we look at a single type of pass - say, a slant completed 5 yards downfield. Say we remove every incompletion - so now we are just looking at every completed slant 5 yards downfield. Now say we average out receivers' numbers over the course of their career, to get a sense of their intrinsic ability to create yards in that spot (say for argument's sake that we get rid of plays involving receivers who have only played with one QB).

The question then becomes - if we are only looking at completions on this one sort of pass, and we've removed the baseline of what that receiver usually does, does the QB, himself, have any further influence over the outcome? If some QB's completions tend to be better at "hitting the receiver in stride", the answer should be yes. If some QBs better select their targets for YAC, the answer should be yes. But as it stands, we don't have any evidence that the answer is yes. In fact, the evidence we do have seems to suggest that you can pretty much predict the YAC based on who is being thrown to and where they make the catch.

Again, the consistently biggest YAC opportunities are on check-downs into the flat, which are probably the easiest pass a QB throws. So there's a significant bias in YAC towards QBs that simply throw the easiest throws more often. Fades, outs, ins, and comebacks don't produce as much YAC.

"Unless I'm misunderstanding, your argument seems to hinge on the assumption that QB YAC would be all about physical throwing accuracy."

Not at all. My argument hinges on the lack of evidence that QBs contribute to YAC in ways that extend beyond the receiver and route they are throwing to. The fact that throwers who complete a higher percentage of passes (which suggests that they are throwing MORE YAC-friendly passes, since those are easier to complete) don't tend to produce more YAC is just a prominent example of this lack of evidence.

"Perhaps people tend to overlook decision-making skills as a crucial component of being a good QB. They suppose if a guy can hit a dime eighty yards away he must be a good QB. If that were true, Jeff George would be a Hall of Famer."

I don't see this as at all apropos to what I'm saying. After all, completion percentage is a measure of in-game successes and failures, not some mechanical test of throwing accuracy. Are you arguing that you would expect QBs that are better at getting their receivers to produce YAC (if such QBs exist) would NOT be better at getting their receivers to catch the ball? That seems truly bizarre to me. It seems like the former skill would almost necessarily be a subset of the latter (again, if it exists).

However, just for fun, if you want to see the mechanical accuracy of the YAC leader, check this out:

http://www.youtube.com/watch?v=tVoqA-LKGb4

Buzz said...

Tarr,

Check my article a few posts back. http://community.advancednflstats.com/2010/12/are-short-passing-qbs-secret-to-nfl.html The last table basically shows which QB's have gained the most YAC per attempt above average after taking into account the depth of the throw (we can filter out some of the easy dumpoffs).

If you look at this list you can pretty clearly see that WR YAC hardly correlates to QB skill at all. QB's like Troy Smith, Alex Smith, Gradkowski, Cambell (coincidently who are from the same teams another reason why WR's should get most of the credit) and Mcnabb are near the top. Where as QB's like Peyton and Eli Manning, Flacco, Ryan and Brees are near the bottom.

Anonymous said...

"The fact that throwers who complete a higher percentage of passes (which suggests that they are throwing MORE YAC-friendly passes, since those are easier to complete) don't tend to produce more YAC is just a prominent example of this lack of evidence."

Not necessarily. Consider the following example: On 1-and-10, a QB has two receivers running the same slant route 5 yards downfield, one to the left and one to the right. Suppose the QB is more accurate throwing to his left (75% completions) than his right (50% completions). Or alternatively, the receiver on the right is more closely covered or has worse hands (same percentages). Now say the QB sees safety help on the receiver on the left such that even if it's complete we would only expect an average of 5 yards after the catch. There's no safety help on the receiver on the right (or alternatively the receiver on the right has more YAC ability) and a completed pass would expect 15 yards after the catch. So a better QB would throw to the receiver on the right because he would expect a bigger gain per attempt (.50 x 20 = 10 y/a) rather than to the receiver on the left (.75 x 10 = 7.5 y/a) despite the fact that his completion pct. would suffer.

And of course this could get even more complex when considering routes of different depths, not to mention down/distance. The point is that I want a QB who can influence YAC per *attempt* not per completion if it increases my team's yards per attempt. Correlating QB YAC to comp % doesn't tell me anything about whether such a QB exists.

Tarr said...

Buzz, that is a great article, thanks a ton for linking me to it. It's very interesting to see how the YAC above expectation (for pass depth) seems to be completely uncorrelated with the other positive aspects of QB play, or with our intuitive sense of which QBs are good.

Anon, an example where QB could influence YAC independent of pass depth is not really news. Ian gave an example early on in this thread. What's at issue is whether any QB can consistently make distinctions like that more often than an average QB, on a per-completion basis.

Buzz said...

Tarr - I just did two quick comparisons using my QB yards/att and QB YAC/att.

1st I took every QB year in Brian's database and compared current season to the next season for all QB's. QB Yards correlated at a rate of .52 to the next season. On the flip side QB YAC only correlated at a rate of .22. (Plain YPA correlated at .50)

2nd I took all teams that had 2 QB's starting within the same season. For this formula QB yards correlated between the two QB's at only .11 where as QB YAC correlated at .26 (plain YPA correlated at .38). Based on this (although a small sample size of 43 pairs) the QB air yards would seem to be a lot higher correlated to QB skills than YAC. (you would want a lower correlation if it was a QB skill and not a team skill).

Other notes - Success rate correlates at .59 looking like that has a LOT to do with teammates and scheme. Pass comp% correlates at .32 where as Pass comp% added (see my article) only correlates at .18 which again is better.

Tarr said...

Buzz, good stuff.

Have you found a more extensive database of QB YAC than what can be gleaned from stats.com's top 20 from each league list?

JAFAman said...

I think this article reflects why there are those who coach and those who blog. Its the discretion of the QB and his personality-- efficiency vs efficacy. some QBs just want to give it to the guy with the best chance of making something happen so our more willing to dump off. other QBs want to make it happen by their arm/will so they care less about YAC as they do their throw. in other words, Marino might get bored with short passes/dump offs, but Montana might be more willing take what he gets.

after that, then u can look at talent pool of receivers, but u would have to avg the YAC of the entire pool after having graded pool just based on talent. then do an apple to apple comparison. who writes this stuff, anyway?

Tarr said...

JAFAman, I think your response reflects why there are those who blog and those who comment on blogs.

For the who-knows-how-many-th time, there is nothing wrong with dumping it off to a RB in the flat, just like there is nothing wrong with handing it off to the RB. It's just that both of those are easy plays that doesn't tell us much about the abilities of the quarterback. For some reason, though, we give the QB 100% credit on the first of those plays, and 0% credit on the second of those plays. It's illogical.

Buzz said...

Tarr - Football outsiders has QBYAC in their books. I'm not sure if they have that type of information in their "premium" section on their website since I'm not a subscriber. If they had information such as that, their injury database, etc it would certainly be worth getting. I was using my YAC approximation process from my other post for correlations. There are a lot of sites that have WR YAC. Hopefully it will just be a matter of time before they have the QB YAC too.

Tarr said...

By George, you're right! They do have QB YAC! That gets us all the way back to 2006. Better than nothing, for sure.

Do you have it all spreadsheeted out? If so, I'd love to get it from you rather than re-invent the wheel. Or you could just post it on a google spreadsheet.

Of course, I can't do the sorts of analysis I really want to do unless I get YAC data for each individual passing play. But it does allow a bit more depth than what I've done so far.

Buzz said...

I don't have it in a spreadsheet. It does look like the play by play available on their website has it. But I haven't ever gotten any of those.

JAFAman said...

And u my friend, Tarr, prove that those who know how to play football do, those that dont, blog. those of us who dont create blogs arent nerdy wannabees. Now, if u think u can go one-on-one about the difference between a quarterback and a passerback, then lets do it. if u want to try and humiliate me, u're gonna get ur flaccid noggin rocked. FYI, I just handed out lunches to a group of college debate wannabees on this same issue at Tampa Stadium. U're next.

Anonymous said...

Some of the data you're looking for is available. For example, "percentage deep". On this website, I think.

mike said...

CORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITYCORRELATION DOES NOT IMPLY CAUSALITY

mike said...

Here's the problem.... a WR will be used on shorter passes if the QB has limitations and on 3rd and 10 a good WR is going to be able to get 7 YAC on a -1 yard screen every time. A better QB may throw a 3 yard slant in stride and the player may get 7 yards for the first down. Both situations you have 7 yards after the catch. The WR stats are very highly correlated as if the WR is "7 yards" good per play regardless of who is at QB. The difference is the team schemed for it's QB and the QB that lead his WR on the run in that situation was in a much more difficult position to have the WR convert the pass into YACs and the pass had to be much more accurate to get that same 7 yards of YAC. The team's best option with a weaker QB still might be to throw the screen on 3rd and 10....
Does that mean that if the worse QB had thrown the 3 yard slant route that the WR still would have gotten a first down because of the WR's stats?
NO! The better QB made the throw that was more difficult and left little room for error. The WR has a high correlation because the team is wise to find a way to give the WR room to run. If the QB can't throw the slant route as well and make the right read, the WR isn't going to have the room to run and potentially get the first down. The good QB will be capable of hitting the hole in the zone in the right timing for the offense to convert or have the best shot at it and the bad QB will throw an erant pass and the WR will be hung out to dry and get hitas he catches it. Any offensive coordinator in his right mind would say that if the WR is going to be hit as he throws, you are better off going to a posession physical jump ball type WR and throwing AT the 1st down sticks or past it, and if you are going to utilize the WR who can get good run after catch you are going to do it in a way that the QB and offense can utilize his strengths by getting him the ball in a situation where he has room to run.....

In this hypothetical the stats may not show a difference because a screen pass is thrown but in reality it does.


Problem!... Unless you are looking at the same exact formation against the same exact routes in the same exact offense to the same exact WR against the same exact coverage with the same exact presnap read or at least very very similar situations, you aren't going to be able to draw strong correlations and any strong conclusions about a QB's ability to create YAC. Despite teams throwing on say 550 attempts in a given year, the sample size of Matt Ryan playing a trips right formation against the Baltimore Ravens with the same exact starters to a backside slant to Julio Jones is very small, and difficult to compare it to say a QB throwing in the same situation against the same team with a very similar WR with a similar career YAC and Average depth of target similar speed and attributes and catching ability.
You cannot do it. The sample size to draw any sort of real conclusions is FAR FAR too low. You can attempt to generalize but all this is telling you tis that it may be possible to conclude that a WR will have a more consistent career YAC regardless of QB, but a QB will have inconclusive data on whether or not he can influence YAC... Yet over a career, there seem to be plenty evidence that Brett Favre has created a much higher percentage of his yards via YACs than Derek Anderson.

You cannot conclude that because WRs numbers correlate well and two different QB's throwing to the same WR in a similar offense do or don't who CAUSED the YAC to occur..


To restate.... Unless you are looking at the same exact formation against the same exact routes in the same exact offense to the same exact WR against the same exact coverage with the same exact presnap read or at least very very similar situations against very similar defenses, you aren't going to be able to draw strong correlations and any strong conclusions about a QB's ability to create YAC. There is not enough information!

mike said...

There's a big difference between statistics not being able to conclude that QBs CAN create YACs over the long run and being able to conclude that QBs are not able to create YACs. Not being able to prove a positive is not the same as being able to prove a negative. It is very difficult if not impossible to use the current database of statistics to really conclude that a QB can boost a player's career YPC. But in reality a screen heavy scheme will produce a greater amount of YACs. Unfortunately that isn't the issue. However a QB that audibles to screen on 3rd and long every time certainly will boost his YAC. That doesn't mean it's smart to do.

also, the problem with Ian's comments about the pumpfake to get the CB to bite leaving the WR wide open is that a QB with a good enough arm will get it to the endzone and the WRs will be zero. Even though technically they could be almost unlimited since no one has a chance to catch him, the rules of the game dictate the play ends after reaching the endzone. So that may unfortunately take away the ability for those created YAC to show up on the QB's stat sheet. On the contrary, how hard is it if a defender screws up and misses his assigned man coverage with no safety help for a QB to throw to the wide open WR. A QB in that situation will be able to pad his YAC if he hits the WR early in the route rather than in the endzone. The good QBs often make the decision of throwing it far enough in front of the WR that he catches it very close or in the endzone so it takes away the advantage of any "QB created YAC" that otherwise could show up IF that's what the player wanted to do. This is the problem with trying to analyze the stats sometimes.A QB can score a TD by a 80 yard bomb in the air to the WR, or a 20 yard bullet to the wide open WR or a 40 yard quite pass. The real thing we need to look at is the yards a player gains after being open on a particular route. If a WR like Randy Moss in his rookie year in single coverage breaks open after 2 steps, does the QB get the ball to the WR in a spot that will get him 6 points? The proper way to throw that pass is as far and as high in the air as possible to give the WR as much field to utilize his speed advantage over the DB as possible. That will result in the fewest YACs but it doesn't matter if the QB made the defender slip by a pump fake or not....

mike said...

Oh and here's the other part you forget... The CB in MAN coverage doesn't read the QB so the WR is often the one that makes the defender miss.However the zone coverage the QB can make him bite and usually the QB will be able to cause the safety to bite on the right side by his eyes and pump fake, and cause the left side WR to be open for step.... But how many more YACs are created on a screen pass or short pass after doing this? Generally the QBs that can get defenders to bite don't get the maximum YACs where as the same talented WRs will be thrown more screen passs and short yards if the backup lacks the NFL arm to get him the ball... And of course in a Denny Green system the #1 requirement was that the QB to make the team had to be able to throw the ball deep down field to hit Chirs Carter, Jake Reid or Randy Moss.

QBs may do great things to create YAC but for some QBs they might show up, for others they might not.. it depends on the route and where the best spot to hit the WR is and the scheme. An elite QB will actually underthrow the route if the defense is guilty of overpursuit and the WR will come back for the ball and run the opposite direction even if it's a slant route. A good scren pass, the QB will have the correct timing so the WR gets the abll early enough to move, but the throw is late enough to fool the defenders, but not early enough that they either figure it out or sack the QB. there is a lot of skill in the screen pass in getting YACs but it STILL is very dependent upon the OL and the RB... That doesn't mean the QB doesn't contribute to it, but he might more on one route than another!

If a QB wants to boost YAC it usually results in suboptimal decisions such as throwing a dart 10 yards down the field, rather than putting a 60 jump ball so the WR that beat the CB can gain the yards BEFORE the catch is made...
So that is the problem and why it won't show up but always remember...

CORRELATION DOES NOT IMPLY CAUSALITY

mike said...

The problem with the one anonymous is he is showing a small sample size, just because something happened doesn't mean something else caused it, if there are other variables involved, you can't say. He is guilty from the same problem as the initial conclusion of the study that you could use stats in general situations to determine who should get credit for YAC. Correlation does not imply causality. If you site that the stock market goes up every time you jump up and down does that prove that jumping up and down causes the stocks to go up or that stocks going up and down cause you to jump up and down, or is it just a coincidence that you happened to be doing jump and jack exercises. We don't know just based on the fact that more often than not you might have observed the stock going up when someone was jumping up and down. If you think that because you observe something that you automatically can deduce the cause whether using stats or your own limited "examples" you are fooled by randomness or fooled by other variables that make things appear to have either a correlation, randomness or luck, not necessarily luck and randomness or correlation itself.

Post a Comment