So the real reason behind the switch to DYAR reveals itself: Pats fan Schatz was looking for a way to make Brady's 2007 look better than Manning's 2004.

j/k

You still have "DPAR" in your finger memory: two table headers still say "DPAR" instead of "DYAR".

I'm glad there's no cluttering of the tables with "True Yards".

32 INTs for Kurt Warner in 2001? Methinks I smell a typooooooooooo.

We may not have the worst individual season, but we have 3 of the top 6, two alone from that godawful 2004 season.

*stares at 2008 Grossman/Orton battle to start, claws out eyes*

Question, do seasonal stats include playoffs?

I'm very amazed at the dichotomy between some of the seasons in the top ten dvoa. pennington, garrard were incredibly efficient, but neither put up gaudy stat lines like the others on the list. I think that supports your work.

It also leads to the question, was January 17, 1999 the greatest showcase of single season QB talent in league history?

It is led by Michael Vick in 2006 (275 DYAR)

I think Vick breaks the system here. I don't think that 700+ yards should be the definition of replacement level, even in a system designed around bootlegs.

It would be nice to see a table of QBs sorted by variance for DYAR per year or per game. Actually, graphs of DYAR per year might show interesting trends regarding rise and fall of productivity over a career.

What stands out to me is the total absence of Joey Harrington from the bottom QB's. I was at the University of Oregon when Harrington took over and played well, and I hoped he would do better in the NFL, but I was almost sure he'd be a lock for the Bottom 10 lists.

Re: #3

You are correct. According to profootball reference, Kurt threw 22 picks that year.

Hey, Dilfer may have put up bad numbers, but who cares?

He was a WINNER, damnit!!!

I think Vick breaks the system here. I don’t think that 700+ yards should be the definition of replacement level, even in a system designed around bootlegs.

DYAR isn't adjusted for offensive system, so being in a system designed around bootlegs has nothing to do with it. He "only" had 275 DYAR because he had 119 runs. A replacement level QB would average around 6 or 7 yards/carry in the situations Vick ran in, so he was only about 2 yards above replacement per carry. He still had the highest rushing DYAR of the DVOA era because he had so many carries.

It also leads to the question, was January 17, 1999 the greatest showcase of single season QB talent in league history?

In the playoffs, maybe, but November 4, 2007 probably wins out in terms of total QB talent. Interestingly, neither game was particularly high scoring, despite the fact that all four teams involved had top 5 offenses led by elite QBs.

"A replacement level QB would average around 6 or 7 yards/carry in the situations Vick ran in, so he was only about 2 yards above replacement per carry."

While this may be technically true, in reality, a replacement level QB would not be able to maintain that 6 or 7 ypc average over the number of attempts that Vick had, since defenses would adjust. Most QBs are able to maintain relatively high ypc numbers, in large part because of the element of surprise.

I remember when I was in high school, a friend of mine (who didn't understand basketball nearly as well as he thought he did) tried to tell me that since Jeff Hornacek had a higher shooting % than Michael Jordan, that if Hornacek took as many shots as Jordan did, then he would score more points than MJ. Of course, he actually would have scored far less, since his shooting percentage would have plummetted if he took that many shots.

14. Or more accurately, Hornacek would not be left open nearly as often as he was if opponents noticed him taking a huge number of shots. Just like when a QB runs 119 times a season, defenses are far less likely to neglect keeping a spy or two on him.

Wait, the future NFC North had the top three quarterbacks in 1995? Kelly...Young...Marino...Elway - what the hell happened to you guys?

Re: 13

The Jets-Broncos AFC Championship game was played in severe winds (up to 45 mph), which helped to keep scoring down. Testaverde had a pretty good game under the conditions, however - 31 of 52, 356 yards, 2 interceptions (both in the last 5 minutes).

The Jets were held to 10 points thanks to Curtis Martin (13 rushes, 14 yards), 3 lost fumbles in Denver territory, and a missed field goal.

On Vick's QB rushing DYAR:

It doesn't mean that a replacement level QB would rack up huge rushing totals. It essentially means that on running plays with Vick, he did the equivalent of 275 yards better than what average running plays would accomplish in the same situation. Keep in mind that Vick's raw totals are probably much better than his FO stats - QB scrambles are pretty boom-and-bust for a guy like Vick..

I'da thought Kurt Warner would have had two seasons up there (1999 also IIRC with GSOT v1.0), not just one. Hmmmm.

Aaron -

I've got a question about how some of the numbers work out. Looking at the top two seasons Brady threw 1 more TD and tallied 150 more yards than Manning. He did this in just under 100 more attempts than Manning had. However, he also threw 2 less picks, and assuming FUM is fumbles lost, he turned it over a total of 3 less times with 93 more attempts.

Is there any added value that shows up from having more possesions and less turnovers? Obviously he's not losing as many success points, but does it offset at all the lesser production per attempt?

(The fact that this example is Manning v. Brady is purely coincidental, I'm just looking at numbers.)

Re: 16.

I can't answer about the other guys, but since I charted San Francisco '95, I can tell you Young fell down because he was banged up--lots of time for Grbac that year.

Re #20
The AdjYPA formula is that a TD is worth 10 yards and an INT is worth -45 yards. I'm pretty sure these are Hidden Game of Football numbers, but could be wrong. In terms of AdjYPA with the table numbers, Peyton comes out at 8.52 and Brady is 7.79. Throw in SOS and year-basis variations and the numbers make sense to me.

DVOA: rate stat, DYAR: counting stat. So, Brady gets more DYAR because he has more attempts. I have a hard time saying, though, that we should think of Brady as the better QB because he posted a slightly lower DVOA in 606 plays than Peyton did in 519. They're both NFL starting QBs who played most every game.

re: 19

I was kind of surprised by that too. If you check out the stats from that year, Warner was actually 2nd in DYAR to Manning after the Warner gets a huge negative adjustment for playing a cake schedule.

Re #19
Check out the opponent adjustments from the 1999 season. Warner loses 9.0% of VOA and almost 300 YAR. Plus, he only had 533 plays that year-8 of the top 12 DYAR seasons had 590+.

Re: #22

I wasn't trying to say the numbers were incorrect, I was just wondering how they work. I'm not familiar with all of the ins and outs, so I was curious if there was anything that factored in the lower turn over rate.

Also, not trying to say that Brady's year was better, it just happened that their numbers were an example of what I was asking about.

Re #25
Another way to put it: every play has a value. Turnovers (I presume) have negative value. DVOA is a measure of the player's value per play over the course of the year, measured against a baseline and adjusted for opposing defense.

Ceteris paribus, Manning's 2004 DVOA would be higher if he turned the ball over at the rate Brady did in 2007 (less frequently). Ceteris paribus, Brady's 2007 DVOA would be lower if he turned the ball over at the rate Peyton did in 2004 (more frequently).

Keep in mind that Vick’s raw totals are probably much better than his FO stats - QB scrambles are pretty boom-and-bust for a guy like Vick.

That's what I thought at first, too, but Vick's "true" yards were actually higher than his conventional yards. He really was worth 1000+ rushing yards that year. Problem is, a replacement level QB averages about 6 or 7 yards, or the equivalent, on runs (QB scrambles are very effective). I guess the problem is that a replacement level QB would be unlikely to sustain that average if he ran 119 times per season, but I'm not sure how to quantify that effect, or even test whether it exists. Most QBs wouldn't benefit from scrambling that much because of injury concerns, and because most QBs are probably more effective passing than they are scrambling, even with the element of surprise that they generally have with the occasional run.

I took a quick look at Donovan McNabb's 2000 season, when he had 629 rushing yards, and tried to see whether his average yards/carry decreased as he ran more, both in terms of carries/game, and in terms of first half of season vs. second half. I graphed his YPC in each game against the number of carries in that game, and then against the week of the season, and I didn't see any clear pattern of decline in rushing average. He seemed to be able to sustain whatever rushing ability he had throughout the year, even though he was running a lot. Now, obviously, McNabb isn't a replacement level scrambler, but if we have to find a replacement level scrambler with 119 carries, I'm afraid we're out of luck.

In a quick check of McNabb and other QBs with seasons that had relatively high rushing totals, I didn't see any clear pattern in the YPC vs. number of carries. Vinny Testaverde, for instance, ran 38 times in 1990 (his career high), and averaged 7.4 YPC, one of the highest averages of his career. McNabb, in 2000, ran 86 times (his career high), and averaged 7.3 YPC (also a career high).

Now, there's obviously the problem of selection bias - QBs who are having a lot of success running will probably run more often, and then if they aren't as effective on the ground the next year, they'll run less. I don't see any easy way out of that problem, though. We can't just take a replacement level scrambler and tell his coach to have him run 120 times a year. Short of that, I don't see how we could answer the question of what a replacement level scrambler would do with that many carries. Honestly, I'm not entirely sure that they wouldn't put up 700+ yards.

27. The only thing I can think of, off the top of my head, is to treat Vick like a RB. Surely defenses did not play Vick like a normal QB and instead defended against his running abilities to an extent, much as they would a RB. So his replacement value is lower than a normal QB, who does not receive this same treatment. However, surely his replacement value is not the same as a RB-- even the best of pass defenses must at least cover the receivers to guard against the pass threat, which they would not against a RB.

In the end, it is perhaps a moot point, since you could say the same for many other situations (say, a QB's pass DVOA suffers because he faces different defenses than most QBs due to his team's weak running game... much as Vick's rush DVOA suffers somewhat because he's running against different defensive looks than most QBs). That's just part of the team-oriented and multi-variate nature of the NFL and I don't know that anyone has come up with a way to fully isolate any player's performance.

"This list looks pretty similar to the first list, but of course players with more opportunity will do better in a total stat than a rate stat. That gives Tom Brady the most valuable quarterback season of the DVOA Era."

Most *productive* season. Most valuable would be Manning's 2004.

Or am I misunderstanding DVOA/DYAR?

29,

Brady contributed more *total* value. Manning was more valuable per play. I'm not sure what distinction there would be between being valuable and being productive.

I still can not believe that in 1995 the top two QBs in the league played for the Lions and Bears and the universe did not implode in on itself.

Re 31: You probably shouldn't believe that, because the top QB in 95 played for GB.

Amazingly, neither Mitchell or Kramer made the pro bowl, with Aikman and Young both getting in on reputation. Also amazingly, throw in Warren Moon's numbers for Minny that year: 4228 yards, 33 TDs, 14 INTs. 4 out of 5 NFCN QBs had essentially MVP-caliber seasons, along with Emmitt Smith who had over 1700 yards and a then record 25 TDs.

http://www.pro-football-reference.com/years/1995/probowl.htm

Re: 22, especially the comment that "They’re both NFL starting QBs who played most every game."

I don't know about Brady specifically, but I've charted Manning's 2004 season. He sat out 96:06 of gametime, effectively breaking the TD record and passer rating record in less than 14.5 games. The former record which required all 16 games, almost one hundred more passing attempts, and some 4th quarter stat-hounding by Brady to equal and then break.

However I'd actually rank Manning's 2006 season at the top of the list. I thought he was better in 2006 than in 2004, watching him carry the team to such efficiency, despite being saddled with one of the worst if not the worst run defenses in the history of the league. The 2004 season changed how teams played the Colts, forcing them to become much more patient.

Great Article. I'd like to see previous years added in for a better comparison of Young, but that's about it.

The Vick thing, I just take it as fairly impossible to adequately take him into account. He broke the mold of what it is to be a running QB, and I can't think of any way to really put that into effect via statistics.

It's kind of like Barry Sanders to me. both Vick and Sanders caused the defenses to play in a different way based on their skill set. That radically changes a number of other factors that we don't have enough of comparative players to make any real concrete explaination of it.

We know defenses changed for both of them. I saw it with my own eyes. What that meant in relation to everything else going on was probably significant, but stats can't tell us how.

At least not yet.

Then again, other players can make that same claim. Offenses scheme for defensive players. Run packages to protect against good defenses.

But Sanders and Vick made the defense actually play differently. I'm sure there are other examples too, that's just what comes up in my head.

Then again, I guess you could say Bledsoe made defenses play differently, I.e. Rush like crazy at him because he can't move. :)

33: You're just asking for a trip to the irrational thread, Richard.

Couple of thoughts --

1) Scrambling QBs -- doesn't the NFL consider a tackle on a scramble to be a sack if behind the line of scrimmage? So the QB really gets a huge break from the NFL for scrambling. If he threw the ball away, it lowers his passing stats. But if he pulls it down and runs, it's a win, win, win (for him, if not the team). If tackled behind the line, it's not his fault (he was sacked). If he gets past the line, he's likely to have gotten past the pass rushers and will get some big yardage until the LBs can tackle him. Good for his running stats.

2) Does FO adjust for this bias?

3) re: replacement value -- in 2003, McNair had big numbers behind incredible pass pro. Volek and O'Donnell (coming out of retirement with a couple days practice) both played better QB for the Titans that year. If replacements were better than the starter when playing with the same teammates, does it impact the starter's DPAR? ;)

I tried to turn it back to the topic by the end ...

33,

So give him a bonus for his running. And a big minus for his passing. Because the changes defenses made to stop his runs made it even easier for him to pass.

OOPS!

make that 34 regarding Vick and defensive changes.

Ahh, the good ol' Bobby Hoying era. *shudders*

Re #36
1) I know we mark in game charting if a tackle on a pass play is a scramble, but I don't think it's incorporated into QB passing stats. Thus, the combined Pass + Rush DYAR totals.

3) If you're willing to accept that Craig Nall in 2004 was the best QB of the DVOA Era (92.4% DVOA!), I'll go along with that w/r/t Volek. O'Donnell, though, gets the "It's Week 17 and Nothing Matters" award.

So you know what else this means? If a rookie QB is just awful, chances are he'll never get it. Look at the worst seasons of all time--it's littered with recent high-pick QB busts--players that always got the "he'll have great game experience and he'll get better" thing. Look at that group--Carr, Akili, Alex :( , Ryan Leaf, Andrew Walter, Orton--all guys that played right away and sucked. As Aaron pointed out, really, the only QB who was dreadful as a rookie (on this list) and did anything was McNabb.

I'd be curious to see such a list among rookie QBs. It looks like there's a point at which, if he performs poorer than that point as a rookie, then he just might not ever get it. Think of all the QBs that HAD to play as rookies that have panned out...Peyton, Cutler, Big Ben, Brees...they weren't great, but they were at least alright. Even guys like Eli, and Vick, and Vince--they were OK.

Re: 43

Brees played one game his rookie season. Walter didn't play at all.

Troy Aikman's rookie season was terrible. I'd say he turned out ok.