Writers of Pro Football Prospectus 2008

Most Recent FO Features

FolesNic12.jpg

» 2014 KUBIAK vs. ADP: The Overrated

Where does KUBIAK differ most strongly from public opinion, and which players are most likely to disappoint their owners in the upcoming fantasy football season?

27 Aug 2010

Introducing YAC+, Part III: Quarterbacks

by Bill Barnwell

As previously promised, it's time to take a look at the impact of quarterbacks on yards after catch (henceforth YAC) using our new metric for analyzing a receiver's ability to run with the ball, YAC+. Although it's not realistic to separate entirely a quarterback's impact on YAC from a receiver's, we'll endeavor to find what sort of influence quarterbacks wield on YAC and how consistent that influence is from year to year.

Since the last YAC+ article was a laptop repair ago, I'll provide a quick refresher on how YAC+ works. For every completion, we calculate a baseline for expected YAC by looking at comparable passes from 2006-2009. We compare the actual YAC gained or lost on the play to the baseline. The resulting difference is YAC+. For more background on how YAC+ works, you can check out the first two articles on the topic here and here.

(For reference, the most YAC lost on one play in the four-year stretch is 10 yards. Brandon Marshall caught a pass thrown two yards behind the line of scrimmage from Jay Cutler on third-and-4 in 2007, promptly lost 10 additional yards, and fumbled in the process. That produced a YAC+ figure of -16.9, also the worst figure in the four-year period. )

First, let's check the leaderboard. Here are the top ten quarterback seasons from 2006 through 2009 by both raw YAC per completion (left side of the table) and YAC+ per completion (right side of the table), with a minimum of 100 completions in the given season to qualify.

Table 1: YAC vs. YAC+
Year Player Team YAC Year Player Team YAC+
2006 McNabb PHI 7.4 2006 McNabb PHI 2.2
2008 O'Sullivan SF 6.7 2009 Romo DAL 1.5
2009 Romo DAL 6.4 2008 O'Sullivan SF 1.2
2008 Cassel NE 6.3 2008 Cassel NE 1.2
2009 Campbell WAS 6.3 2006 Warner ARI 1.2
2006 Brunell WAS 6.1 2009 Rodgers GB 1.1
2009 McNabb PHI 6.1 2009 Campbell WAS 0.9
2009 Stafford DET 6.0 2009 Stafford DET 0.8
2009 Rivers SD 6.0 2006 Favre GB 0.8
2009 Rodgers GB 5.9 2009 McNabb PHI 0.8

What jumps out on those two top ten lists? For one, the groups are mostly the same, as only Mark Brunell and Philip Rivers fell out of the top 10 after switching from YAC to YAC+. The table includes two of the worst seasons by any quarterback over the timeframe, J.T. O'Sullivan's 2008 and Matthew Stafford's 2009. The only player who appears in the YAC+ top 10 twice is Donovan McNabb, and I've suggested that he struggled to hit receivers in stride during his time in Philadelphia.

The top season on that list is McNabb's from 2006. That was the year when the Eagles rallied to make the playoffs with a 5-0 finish under Jeff Garcia, who took over for an injured McNabb at the end of the year. The combination represents an example of one way to try and tease out the importance of a quarterback to YAC. Both McNabb and Garcia completed more than 100 passes on the year, one of 19 times in the past four seasons in which a team has seen two quarterbacks do just that. (Strangely, there were only two such teams in 2009: Buffalo and Tennessee.) While McNabb had the best YAC+ of the timeframe by more than a standard deviation, Garcia had a YAC+ of -0.9, the eighth-worst of any quarterback over the four-year stretch. Garcia was above-average in 2007 (0.5) and about average in 2008 (-0.1), which makes the huge disparity even stranger.

Of those 19 pairs, though, only seven of them had a difference in YAC+ that was greater than one standard deviation (.57). Some quarterbacks who look and play entirely differently -- like Kerry Collins and Vince Young in 2009 -- ended up with remarkably similar YAC+ figures.

If we look at the full four-year stretch (minimum: 400 completions), Tony Romo (.665 YAC+ per completion) just sneaks ahead of McNabb (.663). Also in the top five are Kyle Orton (0.5), Tom Brady (0.5), and Derek Anderson (0.4). On the flip side, Steve McNair has the worst YAC+ in the timeframe (-0.7), but remember that it's only considering 2006-07, not the earlier portion of his career. In the great statistical flukes category, Peyton Manning (-0.192) and Eli Manning (-0.195) are in a virtual dead heat during the past four years in this metric.

Guys like O'Sullivan, Stafford, and Anderson seem like they shouldn't belong in this conversation, but one thing they seem to have in common is that their work really came mostly in the course of one season. Stafford has only played one season so far, of course, while O'Sullivan had virtually all his attempts come during a stretch with the Niners in 2008, and Anderson had more than half of his completions (298, at a rate of 0.6 YAC+) come during the 2007 campaign. We need a much larger sample to say, but I suspect that these guys are just putting up an impressive YAC+ figure in a small sample and haven't been in the lineup enough for their statistics to regress to the mean.

The reason that comes to mind is because there's just not a whole lot of year-to-year consistency in YAC+ for quarterbacks. Among guys with 50 completions or more in consecutive seasons, the year-to-year correlation for YAC+ in 2006-07 was only 0.26. In 2007-08, it fell to 0.13, and in 2008-09, it was back up to 0.25. That's just not strong at all.

The sample size for those groups was either 26 or 27 quarterbacks in each season, which is still pretty limited considering we only have three sets of seasons to compare. In a trip to small sample size theatre, I went and looked at the 13 quarterbacks who had 100 completions or more in each season for the same team over the four-year stretch. Their year-to-year correlation was weak in 2006-07 (0.19) and 2007-08 (0.15), but in 2008-09, it jumped all the way up to 0.60! I don't think there's much to that figure, but it's worth mentioning considering how remarkably consistent league performance was as a whole between 2008 and 2009.

As much as I'd like to make the statement that quarterbacks simply don't have any effect on yards after catch, I don't think we have the data required to make that sort of statement as of yet. Furthermore, while it's certainly worth looking at going forward, I think YAC+ for receivers ends up being far more valuable than YAC+ for quarterbacks. The variance in receiver type is far greater than that of quarterbacks. Vincent Jackson and Wes Welker may have produced roughly similar DYAR figures last year, but they're obviously two dramatically different players stylistically. The difference between Philip Rivers and JaMarcus Russell might be great, but there's just not that much of a difference between Rivers and the fellow quarterbacks at his performance level when it comes to the types of routes he throws. Even the quirky throwing motion isn't as far off from other quarterbacks.

With all that on the books, it's time to put plus-minus and YAC+ on the shelf for the season. I'll re-visit the two statistics in Football Outsiders Almanac 2011. Let's just hope that it won't be a preview book for the UFL.

Posted by: Bill Barnwell on 27 Aug 2010

21 comments, Last at 30 Aug 2010, 12:40pm by Arkaein

Comments

1
by Doug Farrar :: Fri, 08/27/2010 - 2:54pm

Great series, Bill. I knew it would take something special for J.T. O'Smellingsalts to make the front page of FO.

2
by tuluse :: Fri, 08/27/2010 - 4:57pm

50 completions isn't a whole lot. That's 2-3 games worth.

What if you raise the baseline for calculating correlation?

3
by Joseph :: Fri, 08/27/2010 - 6:40pm

Probably means worse sample size. But I could see upping it to 100 or 200, just to see if it would be different. I'm guessing BB did it, and the numbers really didn't change at all.

4
by tuluse :: Fri, 08/27/2010 - 6:48pm

A smaller sample size of players, but each one would have a larger sample size of passes. I was thinking around 150.

11
by billsfan :: Sat, 08/28/2010 - 8:22am

Still, then you'd get to compare McNabb/Garcia 2006 vs. McNabb/Feeley 2007 vs. McNabb/Kolb 2009. Might at least tell us something about *one* QB or the scheme he plays in.

(I also like the Eagles)

5
by roguerouge :: Fri, 08/27/2010 - 7:26pm

Why would it be for the UFL?

6
by dbostedo :: Fri, 08/27/2010 - 7:59pm

Because the NFL will be in the middle of a lockout.

7
by Bobman :: Fri, 08/27/2010 - 8:37pm

I think a little sarcasm was in order there. Gonna hafta dock you some points.

9
by roguerouge :: Sat, 08/28/2010 - 1:18am

Sarcasm is now required for simple clarifying questions?

10
by Kibbles :: Sat, 08/28/2010 - 5:50am

I thought sarcasm was required to use the internet.

8
by fongs2 :: Fri, 08/27/2010 - 11:17pm

How about a defensive YAC+ and correlation? Wondering if certain defensive schemes are more vulnerable to YAC or are just bad tacklers.

12
by Theo :: Sat, 08/28/2010 - 10:29am

while you mention it... Ivory just makes a ridiculous play after a simple swing pass.

13
by Big Johnson :: Sat, 08/28/2010 - 1:56pm

Is there a coorelation between YAC+ and anything? Im failing to see it. It just looks like shit players mixed in with good players with some average players to top it off. What is the relevance of YAC+?

14
by billsfan :: Sat, 08/28/2010 - 3:40pm

I was hoping this might tell us something about "scheme quarterbacks" in the NFL. It didn't tell us that; I'm still not sure what exactly it told us, other than that, ceteris paribus, McNabb was better than Garcia, which, unless you listen to WIP, or even worse, call it, you probably already knew.

(I also like the Eagles)

15
by mm (not verified) :: Sat, 08/28/2010 - 9:50pm

A good QB should be able to hit a streaking receiver in stride for a large amount of YAC.

A good QB should be able to throw into tight coverage, getting the ball to a receiver who is tackled immediately for no YAC.

Not sure if YAC+ should be able to tell you something about the quality of the QB even in theory.

16
by Jerry :: Sun, 08/29/2010 - 12:55am

Maybe YAC+ will tell us something. Maybe it won't. The only way to know is to look at it. Even if Bill ends up with "I think YAC+ for receivers ends up being far more valuable than YAC+ for quarterbacks," that's information we didn't have.

18
by billsfan :: Sun, 08/29/2010 - 6:24pm

In the McNabb/Garcia case, it at least tells us something that DVOA (18.8% vs. 17.4%) doesn't.

(I also like the Eagles)

17
by bubqr :: Sun, 08/29/2010 - 9:55am

I'd like to be able to differenciate deep YAC (the like McNabb had while throwing bombs to Stallworth and R.Brown in 2006) to "real" YAC, where receivers catch a short/medium passes and navigate through traffic. One illustrates a receiver speed and a QB arm strength/deep accuracy, the other WR agility/strength/vision and QB timing/accuracy. I feel like one word is used for 2 different concepts.

19
by chemical burn :: Sun, 08/29/2010 - 7:18pm

Yeah, I was thinking the same thing: all of McNabb's YAC must have come from throws where the receiver had already beaten the coverage over the top - which is not really what I think of when I think of YAC (even if it technically meets the defintion.) The Garcia comparison is funny because it's telling us what we already knew: McNabb throws a much better deep ball and has far better arm strength than Garcia. Not valueless information, but not really anything about YAC as is it exists in the general imagination...

These stats seem crippled by small sample sizes and fuzzy sorts of definitions. I wish I could be productive and come up with a way they could be refined, but it's not really coming to me, even though I've been thinking about it since they started posting these...

20
by bubqr :: Mon, 08/30/2010 - 9:42am

An idea could be defining 2 differents stats such as : "Real" YAC as yards gained after a reception with at least a defender between a player and the endzone, while yards gained after a long reception with no one standing between the player and the endzone could be "Bomb YAC" or something.

21
by Arkaein :: Mon, 08/30/2010 - 12:40pm

Maybe a system like Adjusted Line Yards which only considers the first 10 or so yards of YAC on a single play. This would still give a QB and WR credit for throwing/catching balls in stride and making the initial defenders miss while filtering out footraces in the open field and any other anomalous plays (such as a 3rd RB gaining 90% of his YAC for the season on a single 50 yards screen pass).