Writers of Pro Football Prospectus 2008

Most Recent FO Features


» 2018 Free Agency Cost-Benefit Analysis

Is Kirk Cousins the best free-agent quarterback in recent memory? Should Trumaine Johnson or Malcolm Butler have gotten the larger contract? And what makes a free-agent contract good or bad, anyway?

26 Jul 2010

Wisdom Of Crowds 2010: QB Part I

by Bill Barnwell

Our Wisdom of Crowds feature returns for a second season, starting today. For those of you who are unfamiliar, we use my Twitter account to crowd-source fantasy projections for the upcoming season. You can read an explanation of the idea's origin from a year ago here.

This year, we won't be making any changes. Starting with quarterbacks, I'll nominate one player a day for predictions. The Twitter followers then provide three statistics for the player in question, assuming that the player makes it through a 16-game season without missing time. (I throw out those submissions that I consider to be disingenuous; it amounts to one or two per player.) For quarterbacks, those statistics include passing yards, touchdowns, and interceptions.

For each player, I'll provide an average projection from the users, a best- and worst-case projection in each category, and some thoughts on the likelihood of the player matching said projection.

I started with the most prominent quarterback to change teams this offseason.

Monday: Donovan McNabb
Average: 3,406 passing yards, 22 TD, 13 INT
Maximum: 4,180 passing yards, 31 TD, 7 INT
Minimum: 2,500 passing yards, 11 TD, 20 INT

Since we started with McNabb, he got about twice as many responses as any other player. It led to a wide range of responses. 11 touchdowns seems like a ridiculously low minimum projection, so you might want to consider the next-best projection -- 15 -- as the floor. There's only been nine seasons over the past 20 years in which a quarterback started 16 games and threw 11 touchdowns or less; Trent Dilfer threw an appalling four touchdowns in a full slate of games in 1995, which blended well with his 18 interceptions.

McNabb has averaged just under 3,925 passing yards per 16 games over the past three seasons, so it's pretty clear that the public expects his numbers to decline with inferior receivers and a questionable offensive line. Only three percent of the projections called for McNabb to throw for more than 3,925 yards. His touchdown total is right in line with his previous level of play, though, and his interception rate is only slightly higher. His maximum projection would be the best line of his career, which seems downright impossible at 34.

I think this is a wholly reasonable average projection. Of course, projecting McNabb to make it through 16 games is a dangerous bet, but we can pro-rate his numbers if he doesn't make it to 16 games at the end of the year.

Tuesday: Tom Brady
Average: 4,164 passing yards, 30 TD, 13 INT
Maximum: 4,800 passing yards, 42 TD, 7 INT
Minimum: 3,200 passing yards, 24 TD, 19 INT

This is basically an expected repeat of Brady's 2009 season, when he threw for 4,398 yards and 28 TD against 13 INT. It's a difference of 15 yards per game. We've written about how Brady faced an extremely difficult slate of pass defenses a year ago, one that he's not likely to face this year. Brady should also be a little healthier with another year removed from his knee injury. Then again, Brady also had Wes Welker at 100 percent for most of last season, and while Welker appears set to return at the beginning of the season, it's hard to imagine that he'll be the same player he was in 2007-2009, at least at first.

What does that all add up to? I think it's hard to say without knowing how healthy Welker's knee is. I would think that Brady beats that average yardage prediction and probably ends up around 4,400 yards again. That minimum prediction would sure make his contract negotiations interesting, though.

Wednesday: Matt Schaub
Average: 4,218 passing yards, 29 TD, 15 INT
Maximum: 4,904 passing yards, 43 TD, 8 INT
Minimum: 3,250 passing yards, 22 TD, 20 INT

Remember: Matt Schaub threw for 4,770 yards a year ago. Only 11 percent of respondents expect him to hit that total again or exceed it, a sign that the people following a Football Outsiders account on Twitter might be a little more comfortable with the concept of regression to the mean than a broader public. It could also be some sort of unconscious hedging for the idea that Schaub won't play 16 games again, something I've asked the respondents to ignore the possibility of. His average touchdown and interception figures are exactly the same as those of his actual figures from a year ago, which is something we'll continue to note going forward.

That maximum projection might seem like someone who placed a bet on Schaub winning MVP, but it's actually composed of three different projections (4,904/35/16, 4,200/43/13, and 4,100/27/8).

While Schaub's chances of playing 16 games are questionable, the bigger issue here in projecting his numbers might be the health of the players around him, something we don't ask projectors to consider. There's no reason to think Andre Johnson is going to get hurt, but an injury to Johnson would obviously dampen Schaub's numbers dramatically. Owen Daniels's healing process in recovering from a torn ACL could also slow Schaub at the beginning of the year.

Thursday: Eli Manning
Average: 3,705 passing yards, 25 TD, 15 INT
Maximum: 4,700 passing yards, 35 TD, 10 INT
Minimum: 3,100 passing yards, 15 TD, 22 INT

Again, you might think that the minimum projection came from someone with a bone to pick with the fairer Manning brother, but not so; the figures again come from three different projections (3,100/21/18, 3,825/15/16, and 3,200/22/22). The 4700-yard projection was part of a 4,700/24/15 line; only once in the past 20 years has a quarterback thrown for more than 4,500 yards and failed to hit 24 touchdowns. Warren Moon threw for 4,690 yards with 23 touchdowns and 21 picks in 1991.

Comparing Eli's line to last year's 4,021/27/14 figure is interesting. Eli threw for more than a yard per attempt more than he ever had as a pro, something buoyed by the emergence of Hakeem Nicks as an elite receiver after the catch. We don't ask people to project attempts, but I wonder whether people expect Eli to give some of his 509 attempts back to the running game.

Friday: Philip Rivers
Average: 4,146 passing yards, 30 TD, 13 INT
Maximum: 4,672 passing yards, 38 TD, 7 INT
Minimum: 3,200 passing yards, 24 TD, 22 INT

Again, there's a complicating factor here: The statuses of Vincent Jackson and Marcus McNeill are unknown. I didn't give any instructions on what to expect from them.

Rivers' projection expects a leap from nine interceptions to 13, but with an increase in touchdowns from 28 to 30, and he's the closest of the five quarterbacks to his actual 2009 yardage total (4,254). Considering that we've seen the userbase consistently project a larger dropoff in yardage, I'm thinking that people are expecting Rivers' attempt total to rise from 486, where it was a year ago. 486, though, is the most passes Rivers has thrown in a season; it seems remarkable that a Pro Bowl quarterback with four great years has never made it to 500 attempts, but Rivers is the exception.

I don't think Rivers's attempt total will rise, thanks to an improved running game, so I would probably regress his average prediction a bit across the board. Something more like 4,000/26/11. Of course, there's still a lot of time between now and the end of training camp, when we'll know a lot more about how much we should expect out of Jackson and McNeill.

We'll be starting up again today with a new quarterback, and continue through the week with four more. If you'd like to join in, make sure you're following @FO_BBarnwell on Twitter, and if you'd like to follow anyone else on the FO team, check out our list of Twitter names to the left.

Posted by: Bill Barnwell on 26 Jul 2010

35 comments, Last at 10 Aug 2010, 1:48pm by Anon


by Bobman :: Mon, 07/26/2010 - 12:10pm

The fairer Manning brother?

explain, please.

by dbostedo :: Mon, 07/26/2010 - 12:30pm

I think it's clear that Bill finds Eli to be better looking than Peyton.

by Anon (not verified) :: Tue, 08/10/2010 - 1:48pm

Eli is dark haired while Peyton is blonde. Fair also historically meant light of skin or hair.

by Marko :: Mon, 07/26/2010 - 1:12pm

Maybe it's on a scale from outstanding to poor, with fair somewhere in the middle. Eli is fairer, while Peyton is outstandinger.

by ohearn :: Mon, 07/26/2010 - 1:58pm

Similar to chocobo breeding in Final Fantasy VII.

by Bowl Game Anomaly :: Mon, 07/26/2010 - 3:23pm

I love this website.

by Shawn :: Mon, 07/26/2010 - 12:17pm

Can you explain the reasoning for combining projections to get the Maximum for certain players? You didn't provide an explanation in the intro, and since you only mentioned doing it with Schaub and Manning's Maximum Projections, I'm going to assume that is the only time you did it. But why?

by Anonymous11111 (not verified) :: Mon, 07/26/2010 - 12:27pm

He did provide a link to an explanation in the intro:

"We'll be providing the average projection for each player as well as the maximum and minimum figure provided. We'll draw these from multiple predictions, so if one prediction for Cassel suggests 3000 passing yards and 15 TD, and another suggests 2900 yards with 18 TD, we'll consider the maximum projection to be 3000 yards and 18 touchdowns."

by dmstorm22 :: Mon, 07/26/2010 - 12:17pm

Wouldn't regression to the mean be expected to occur around the league as well for QB's?

10 QBs? Threw for 4,000 yards last year, which is three more than any year previously. Now, most media people point to this being the feature of the league, but there were only 7 and 6 in 2007-2008. I would expect the number of 4,000 yd QBs to regress to about 6-7 as well.

by RickD :: Tue, 07/27/2010 - 1:16pm


(Number of QBs throwing for 4,000 yards or more) is not a statistic you would expect to "regress to the mean". (Let's call it QB4K)

If you are looking at a well-defined quantity such as "performance of player X", then you expect that statistic to regress to its mean. But that's not what QB4K is. QB4K measures the number of QBs performing at a given level. One expects this statistic to take a random walk over time, rather than expecting it to be uniformly distributed, independently over time. If there were 3 elite QBs 5 years ago there might be 5 elite QBs at that level. Or 7 or 2. One would have to be very careful how one modeled this statistic, but it suffices to say that a simple model presuming a time-independent binomial sampling hardly would appear to suffice.

We are in a period where the rules of the game have been changing, the officials' behavior has been changing, and a relatively high number of QBs are hitting their prime.

by Jeff Fogle :: Tue, 07/27/2010 - 2:00pm

Teams passing for more than 4,000 yards last year (2010 notes in parenthesis):

Houston: 4,654 (same coach, QB, good weather/dome)
Indy: 4,515 (same coach, QB, dome)
NE: 4,436 (same coach, QB)
NO: 4,355 (same coach, QB, dome)
SD: 4,338 (same coach, QB, good weather)
Dallas: 4,287 (same coach, QB, cathedral)
GB: 4,180 (same coach, QB)
Minnesota: 4,156 (same coach, QB?, dome)
Pittsburgh: 4,148 (same coach, suspended QB)
Philadelphia: 4,089 (same pass-happy coach, new but decent QB)
NYG: 4,019 (same coach, QB)
Arizona: 4,016 (same coach, new QB though, dome)

Three teams sneak over 4,000 by less than 100 yards. Next on the list is all the way down at 3,627...as this is a break between the have's and have-not's at QB last year. So, a dozen teams made it last year...at a time when holding has been liberalized in a way that seems to increase production and keep QB's more healthy than in the past. Most of a team's passing yards go to one QB if that guy stays healthy obviously.

Potential super-surgers:
Washington: 3,490, but now with Shanahan and McNabb
Chicago: 3,473, but now with Mike Martz at coordinator
Atlanta and Baltimore: in the 3,500 range with young QB's who are due to keep improving on the traditional QB maturity curve.

Sometimes outliers are outliers. Sometimes they're early warning signs of paradigm shifts. The NFL liberalizing holding may have created the latter.

Worth noting I think that:

*Some in baseball thought the power numbers would regress to the mean back in the early stages of the steroid era. The power numbers didn't regress until they cleaned out steroids several years later. That's actually a hint that we might be in the EARLY stages of yard inflation. Which QB's will be the McGwire/Sosa combo that captures the excitement of production? Manning/Brees tend to clamp down with leads. Maybe a Houston QB and somebody else in the middle of the standings rather than at the top.

*Some in pro hoops thought the trend toward lower scoring after Pat Riley ruined basketball (lol) would regress. But, scoring totals kept right on dropping until the league got more aggressive about calling hand checks and other obstruction fouls several years later. Scoring then kept trending up for a sustained period.

In other words, stuff doesn't always regress. Or, a forced regression is eventually imposed way down the road if things drift too far off course. Don't think the league is disappointed in big production and attractive fantasy stats.

by JMiller22 (not verified) :: Mon, 07/26/2010 - 1:05pm

Phillip Rivers - after reading FOA 2010 one would expect Rivers' attempts to climb for sure. Assuming they win less games means more playing from behind and therefore more pass attempts. Maybe all those 12-14 win seasons have held him under 500 attempts b/c they are pounding the rock with a lead almost every week?

by Joseph :: Mon, 07/26/2010 - 4:00pm

Or maybe because, except for last year, they had Hall of Famer LDT? (Last year, they had winding-down-his-HOF-career LDT).

by Tim Wilson :: Mon, 07/26/2010 - 4:51pm

Might not have mattered...LDT has at least one 100 reception season under his belt.

by Big Johnson :: Mon, 07/26/2010 - 4:19pm

that probably has some truth to it, but its still far from the exact truth. From what ive gathered, the san diego offense under norv turner is very rushing game friendly compared to other coaches when it comes to raw totals. The colts for example finished 31st in the league in rushing attempts per game. The chargers finished 19th (and 23rd in passing attempts per game). It speaks volumes that norv will rush the ball that many times despite how awful they were at it. Whenever people say that the san diego offense is passer friendly.... they are wrong. This is why Matthews will have a beast of a year. If cripple LT can get handed the ball that many times, imagine how many times a good runningback will get handed the ball by norv turner.

however, In 2008 the chargers were 7th in passing attempts and 20th in rushing attempts. the 8-8 record supports your theory that record will affect rivers passing totals this year, assuming they finish close to the almanac projection. To me its gonna be an interesting year for the charger offense because with a more effective rushing game but closer games it will be very difficult to predict where the attempts are gonna go in that offense.

by MJK :: Mon, 07/26/2010 - 4:01pm

Bill, just a suggestion for the format of this article. Telling us the maximum and minimum doesn't actually convey all that much information, since one crazy person or Raiderjoe can throw a projection out there that has absolutely no basis in reality (as you point out for McNabb).

What would be really neat would be to put little histograms in for each player.

If that's too much effort or too graphics intensive for the site, at least listing the median instead of the "average" (by which I assume you mean the mean), and listing the standard deviation (or, perhaps, the 10th and 90th percentile numbers) instead of the maximum and the minimum would be more informative.

by fyo :: Mon, 07/26/2010 - 5:20pm

I'd like to second this suggestion; I was wishing for the exact same thing while reading the article.

by Danish Denver-Fan :: Mon, 07/26/2010 - 5:33pm

I agree.

I think, however, that the average reader might find that too heavy on the math. I'm not calling this a mainstream site (yet?), but I think a lot of people have trouble with their statistics material from High School (or college?).

by Sophandros :: Mon, 07/26/2010 - 5:43pm

If they're reading this site, chances are they can handle the elementary stuff that we're talking about with this feature.

Sports talk radio and sports message boards are the killing fields of intellectual discourse.

by dbostedo :: Mon, 07/26/2010 - 6:11pm

Personally, I like it as is. I'd have no problem understanding the histograms or additional data, but it would also clutter up a nice, simple read.

I believe Bill is already doing some cherry picking to eliminate the outliers, so presumably, using the median wouldn't really tell you much more than using the mean. I, personally, don't care whether the mean or median will actually turn out to be more accurate at the end of the year, but I suppose I could see how some might.

Maybe an addendum to this article with all of the collected numbers would be good, so people that wanted to could just go to town on them?

by young curmudgeon :: Mon, 07/26/2010 - 6:23pm

re: comment 11 by MJK: "one crazy person or Raiderjoe" I'm glad you differentiated! My own position is that Raiderjoe is not a drunken dunderhead who types with half of his fingers taped together, but instead a talented (and at times even brilliant) performance artist. Or at least that's usually my position...sometimes even I wonder.

by AnonymousLurker100417 (not verified) :: Mon, 07/26/2010 - 10:40pm

I agree, and made the same point in response to RJ's comment on the fake Emmitt tSmith draft projections that were both a little RJ and a little ... fake Emmitt Smith. Glad to see others voicing the same suspicion explicitly.

Performance Artist. Yup, I hadn't thought of that term for it, but it's right on.

by tally :: Mon, 07/26/2010 - 11:46pm

Stop destroying my image of RJ!

by Independent George :: Mon, 07/26/2010 - 11:48pm

I honestly believe raiderjoe is one of the most knowledgeable football people here, and the consistency of the writing is truly astonishing. Performance art is a great term - I'm truly impressed by the performance.

by Big Johnson :: Tue, 07/27/2010 - 3:48am

couldnt agree more. Basically his value is that if he predicts a team will be good then they will win no more than 7 games(raiders three years running), and if he predicts a player will be good he will end up being the worst player of all time (jamarcus). NOT. raiderjoe couldnt hold peter kings jock!

by Mr Shush :: Tue, 07/27/2010 - 5:52am

He actually tends to be pretty accurate about anything not involving the AFC West. It's not that he doesn't know football, just that he's blinded by homerism and hatred for divisional rivals.

by Independent George :: Tue, 07/27/2010 - 10:25am

Which, again, is the performance art aspect of it. It's like the cast of 'Spinal Tap' remaining in character for the DVD audio commentary. In many ways, it's even more brilliant than the original movie.

I think you could make an interesting case study of postmodernism through raiderjoe. And I say this as someone who despises academic postmodernism.

by Big Johnson :: Tue, 07/27/2010 - 8:45pm

i dont think hes ever been remotely accurate on anything, AFC west obviously included. Id like to hear one of his predictions that wasnt one of the worst predictions of all time like his jamarcus russell one.

by dmstorm22 :: Mon, 07/26/2010 - 6:41pm

What is the sample size, because of its like 30, then the 10th and 90th percentile (3rd and 28th projections) wouldn't be all that informative. I would just throw out the max and min.

Also, I'm not sure I like the combining of projections. If someone projects a 3300/25/7 and another guy gives a 4300/28/14, then I wouldn't give the max projection as 7 ints. It doesn't make total sense, since they are totally independent samples.

by dbostedo :: Mon, 07/26/2010 - 8:06pm

Well, you aren't really "combining" anything because you aren't coming up with a single stat, you're really doing three different polls here : How many yards? How many TDs? How many INTs? When looked at that way, it makes sense to take the highest and lowest of each. Knowing the max and min and average of each of the three makes more sense than trying to figure out if, for instance 4500/35/20 is more of a max than 4200/38/10. That gets very messy.

by Dan :: Mon, 07/26/2010 - 6:50pm

Last year was the first time Rivers sat out most of week 17, otherwise he would've topped 500 attempts (Volek threw 30 passes that game). The Chargers have been passing more - they threw the ball on 47% of their plays in 2006, 49% in 2007, 53% in 2008, and 55% in 2009. But because of week 17 and the fact that they've also been running fewer plays (especially in 2008), the increase has barely shown up in Rivers' attempts.

by utvikefan (not verified) :: Tue, 07/27/2010 - 4:02am

I learned from this article that QB's over 34 can't have their best years (sorry Colts fans) and that apparently the only receiver that keeps Bill up at night with worry of injury is Andre Johnson, for some odd reason. =P.

I have decided I like these "wisdom" articles, and please do continue them. Great work as usual.

by Sunil (not verified) :: Tue, 07/27/2010 - 8:36am

Is there a review of fantasy projections from last year i.e. how did "FO crowd" do relative to NFL gurus? And relative to actual performance also?

by Snack Flag (not verified) :: Tue, 07/27/2010 - 3:25pm

How would one "regress an average prediction across the board"? Regression isn't synonymous with subtracting value and, for a stats based website, there should really be a more correct usage here. You don't regress something - the measurement itself goes towards the mean over time. Tommy Bennett at BP suggested that the stats community as a whole use this phrase sparingly because it's been so beaten into the ground that it's lost meaning. I really appreciate the work on this site, but this phrase is so overused here that it's taking away from the good analysis that surrounds it.

by peterplaysbass :: Tue, 07/27/2010 - 4:18pm

Just for fun.

McNabb 3340/22/11
Brady 4120/28/13
Schaub 4320/27/16
Manning 3750/25/14
Rivers 3860/27/12

Manning will get his 500 attempts this year. McNabb will struggle to hit 7 yds per attempt.