Reflections on a racing form: January 2012

O.k., this is and is supposed to be a horse racing blog, but I'm sure you'll allow me this one departure. I'm sure, actually, it won't be the last, but I am allowed other hobbies as long as I get my work done here, right? Anyway, to the subject at hand. I can remember arguing with people about Michael Vick back in his Atlanta Falcon days. It was hard to evaluate the effectiveness of his mixture of running and passing. But what I said was, look, do the Falcons have one of the best offenses in the league? Do they look like the Indianapolis Colts, the Kansas City Chiefs, or the Denver Broncos? Or do we see a fair amount of punting and stagnation? It was clear that the Falcons' mixture was not stupendously effective.

Yards-per-drive can be harshly criticized, but I know that if someone presented this information to me as an outsider, I would want it. I think it is about the best way to look at quarterbacks.

Here are some of the assumptions that go into it. First, the advantages of looking at effectiveness per drive, rather than over per play, or over a game of indeterminate possessions, should be obvious to all who have thought about this sort of thing to any extent.

Second, the idea is that running the ball doesn't matter a whole lot, so the passing game carries the lion's share for yards per drive. (What kind of offense did the 1984 Rams have with the record-setting Eric Dickerson, right?) People talk about a quarterback's win-loss record -- well, quarterbacks don't play defense or special teams.

So, while we might not be isolating the quarterback's performance looking at yards per drive, we're looking at him a lot more closely than if we looked at team record, and we're still staying attuned to the bottom line.

Another assumption is that clutch play and ability to score touchdowns as opposed to field goals is basically a myth. Give me the guy who can move the ball, and eventually he'll score touchdowns, too. Points and points per drive are extremely dependent on field position, which is why I prefer yards per drive.

This is a historical project. I was interested in the quarterbacks of yesteryear as much as the quarterbacks of today. So what I did was to take all seasons when a quarterback was the main starter for his team, and figure his yards per drive versus the league average. For instance, if his team averaged 28 in a league that averaged 26.5, his score for that year is 1.5. If his team averaged less than the league average in yards per drive, I recorded the number, but didn't deduct from the player's total. So here is an example of a quarterback's season-by-season numbers, and his total.

Norm Snead

1961 -6.9
1962 -1.7
1963 -0.2
1964 0.8
1965 4.9
1966 -1.7
1967 1.1
1968 -1.8
1969 -0.9
1970 0.8
1972 5.1
1973 1.6

Total 14.3

It is in vogue in baseball, and I imagine in all sports, to compare performance to replacement-level players. Maybe that would be a -3 season for quarterbacks, instead of 0.

My instinct is that better rankings come from using 0 as the base, though. Otherwise, the ratings are too dependent on length of career. Yes, in some theoretical sense, if you take ratings with deathly seriousness, you are interested in how many wins the player added over the course of his career. Some people believe that a meaningful line can be established between players below and above a certain point. This approach may capture value very well, but I'm not sure it captures quality. Players do not have a deathly responsibility to add as many wins as they can, and we do not have a responsibility to assess players in exactly that light. We are not paying players with taxpayer money. From a practical point of view, general conclusions are not going to differ much using 0 as the base or -3 as the base, however.

I will have much more to say about method towards the end of this post. My goal in all of my decisions was to have the right general approach, to be in the ballpark. I did not quadruple the time of the project to dot all the is and cross the ts. I have a good intuitive sense of what affects the results markedly and what does not. When I faced multiple ways of defining some aspect, and they could all be carried out without great inconvenience, I chose carefully, however. I don't think anyone can say that I was haphazard or not thoughtful in my decisions.

What else can I tell you? My main source was the 2007 ESPN Pro Football Encyclopedia. Yards per drive are listed for all teams, starting in 1941. Football Outsiders' website is the source for yards per drive starting in 2007 through 2011.

Starting in the '60s and '70s, quarterbacks got credit for a team's yards per drive if they started more than half the games. If they started exactly half the games, they got credit for a team's yards per drive if they threw the majority of the passes. Going back in history, games started by quarterbacks the way that the Pro Football Encyclopedia defines it can be deceiving and fail to indicate the truly most active player at quarterback. So I used throwing the majority of the passes as the standard there.

I compiled the career numbers for 116 quarterbacks in all, with a hope to get all significant quarterbacks in NFL history and a good many more. I started out just freelancing, letting my curiosity determine whom I examined, and free associating. This produced a great many names. To make sure I had all significant quarterbacks, I then found the quarterback for all teams that were in the top two in the league or conference before maybe 1960, and the top three in the league or conference afterwards (top being defined by yards per drive). It appears that just about every quarterback who has a significant career at some point rates among the two or three best in his conference. This is normative for a player who spends, say, ten years as a starter, with more negative numbers than positive numbers. So, while Bernie Kosar isn't in the rankings, or James Harris or Neil O'Donnell, the method erred on the side of overkill.

There were 12 quarterbacks who scored 40 or higher, 18 who scored 30 or greater, 37 who scored 20 or greater, and 79 who scored 10 or greater. If I missed running more than a couple who scored 20 or higher, I'd be surprised. I'm sure I'm missing quite a few who scored 10 or higher, but that's a modest enough score that I don't care. Anyone who scored 10 or higher still had a good career. 3.0 is a nice score for a season, often good enough to rate in the top three in the conference, so 10 is more than three seasons of that. 10 for one season is a monster season. Twenty or more for a career is very good, and we can at least start talking Hall of Fame for that guy. I will list the quarterbacks who scored 20 or more in order, and deal with the rest who didn't in alphabetical order.

One more note of method: starting in 1960, and continuing to the AFC/NFC years, the Pro Football Encyclopedia shows conference yards per drive, but not NFL yards per drive. So, while the NFL YPD would have been preferable, conference yards per drive was my base. Players are compared to the average yards per drive in their conference, not to the average yards per drive in the NFL as a whole.

Even I do not think yards per drive is sufficient on its own to begin to rank quarterbacks. It lumps in a lot that you want, including quarterback running, third-down-efficiency, etc. But it leaves out turnovers.

After I had completed yards per drive for all quarterbacks, I also figured how many interceptions they saved or cost their teams over the course of their careers. Interceptions are difficult to look at historically, so this yielded fascinating clarity on who was and wasn't an interception thrower in the various eras.

The method for calculating interception differential was by no means perfect. It's conceptually dirty but should have given an assessment close to the ideal method in just about every case. I first took the average INT rate in all seasons that a quarterback was a starter. What the Football Encyclopedia has for this is the NFL average, not the conference average, except for 1960-1969, where interception rates are broken down by NFL and AFL. I then took this INT rate and multiplied by the number of career passes the player threw, and compared to the number he did throw. Yes, in an ideal world, I would have weighted the league INT rate by number of passes thrown in each season, but the departure is a small thing. For instance, if you look at players who began their careers within five years of each other, their expected interception rate rarely differed much. And it's not like I'm missing important seasons entirely that the player had.

Unlike for the YPD compilations, INT differentials were allowed to be negative. The sum is indirectly the sum of all seasons, negative and positive. The idea was just, "Was this guy an INT thrower over his career?" The idea was not to directly integrate the YPD and INT data and achieve mathematical nirvana.

Keep in mind when assessing a player's INT differential that length of career was an essential component, even more so than with YPD. If you throw three times more passes than another guy, you have the opportunity to save or cost three times as manhy interceptions. So use your mental idea of how many passes a quarterback threw when assessing the INT number.

I've rated the players by YPD but included two other pieces of information next to the number: the career-best season, and then the INT differential. Positive here actually means that the player saved interceptions, and negative that he yielded excess interceptions. This should be obvious when you see who is strongly positive and who is negative. When I add my commentary to each player on the list, I repeat the three statistics, but call saving 20 interceptions, "20 good" and costing 20 interceptions, "20 bad," just to be clear.

Without further ado then.

(This post was written after the 2011 season. Numbers and rankings have not been updated, with the exception of immediately below, where the current scores are appended to the ones that go through 2011.)

Manning, Peyton 80.8, 9.5, 33 // 87.6, 9.5, 33
Marino, Dan 64.3, 10.8, 57
Luckman, Sid 58.4, 11.9, 32
Young, Steve 52.8, 11.0, 34
Brees, Drew 50.0, 11.7, 24 // 54.1, 11.7, 19
Montana, Joe 49.3, 8.4, 71
Brady, Tom 47.3, 12.2, 50 // 56.8, 12.2, 62
Fouts, Dan 46.0, 12.0, 16
Van Brocklin, Norm 44.3, 11.0, 22
Tarkenton, Fran 43.6, 6.1, 83
Unitas, Johnny 41.5, 6.5, 53
Favre, Brett 41.1, 6.6, -11
Hadl, John 38.4, 6.5, -1
Moon, Warren 37.0, 9.1, 13
Culpepper, Daunte 35.7, 10.1, 0
Staubach, Roger 35.5, 5.2, 48
Baugh, Sammy 35.5, 13.7, 58
Warner, Kurt 34.2, 10.5, 2
Elway, John 29.4, 7.1, 42
Esiason, Boomer 29.4, 6.7, 9
Griese, Bob 28.7, 5.7, 13
Aikman, Troy 28.5, 7.8, 19
Anderson, Ken 28.4, 7.4, 59
Green, Trent 28.4, 8.9, 6
Rivers, Philip 28.3, 8.7, 13 // 28.3, 8.7, 14
Hart, Jim 27.5, 6.3, 17
Tittle, Y.A. 26.4, 6.8, 27
Layne, Bobby 26.1, 5.0, 12
Brodie, John 24.2, 5.1, 27
Kelly, Jim 23.8, 6.2, -3
Starr, Bart 23.5, 5.9, 41
Garcia, Jeff 22.7, 6.8, 35
LaMonica, Daryle 22.5, 6.5, 8
Stabler, Ken 22.3, 6.2, -36
Dawson, Len 21.2, 5.1, 34
Ryan, Frank 20.3, 6.1, 4
Gannon, Rich 20.1, 7.5, 39
Bartkowski, Steve 8.1, 3.9, 15
Blanda, George 19.0, 10.2, -25
Bledsoe, Drew 8.0, 2.6, 9
Bradshaw, Terry 11.0, 3.7, -11
Brown, Ed 10.6, 3.7, -15
Brunell, Mark 18.4, 4.4, 40
Bulger, Marc 9.9, 5.3, 5
Christman, Paul 7.8, 4.1, 25
Collins, Kerry 6.8, 4.4, 4
Conerly, Charlie 6.3, 3.2, 34
Cunningham, Randall 11.0, 9.2, 25
Cutler, Jay 10.9, 8.6, -10 // 10.9, 8.6, -14
Davidson, Cotton 6.1, 3.0, 4
DeBerg, Steve 13.6, 4.8, 7
Detmer, Ty 2.7, 2.7, -3
Dickey, Lynn 3.8, 3.5, -35
Everett, Jim 8.9, 3.2, 2
Ferguson, Joe 13.2, 5.7, 8
Ferragamo, Vince 3.4, 3.0, -20
Flutie, Doug 7.8, 4.5, 7
Gabriel, Roman 10.9, 4.4, 94
George, Jeff 9.7, 4.5, 22
Graham, Otto 15.9, 6.4, 20
Griese, Brian 12.1, 6.4, -7
Grogan, Steve 11.7, 5.4, -32
Haden, Pat 3.9, 3.5, 8
Hasselbeck, Matt 16.0, 5.4, 7 // 16.0, 5.4, 9
Hipple, Eric 4.6, 4.1, -5
Jaworski, Ron 6.3, 2.5, 25
Johnson, Brad 10.1, 4.6, 16
Jones, Bert 6.2, 5.8, 27
Jurgensen, Sonny 16.5, 5.2, 45
Kemp, Jack 10.7, 3.2, -2
Kilmer, Billy 6.7, 3.7, 12
Kramer, Tommy 7.7, 3.7, -1
Krieg, Dave 2.7, 1.7, 3
Landry, Greg 8.8, 5.3, 17
Livingston, Mike 2.5, 1.6, 10
Lomax, Neil 9.7, 4.5, 39
Lujack, Johnny 12.1, 5.6, 7
Manning, Archie 6.4, 4.3, 13
Manning, Eli 17.2, 5.1, -11 // 20.2, 5.1, -10
McMahon, Jim 6.8, 2.9, 13
McNabb, Donovan 16.1, 4.8, 50
McNair, Steve 10.7, 4.8, 26
Meredith, Don 13.0, 5.3, 14
Mitchell, Scott 10.9, 4.5, -6
Morrall, Earl 12.0, 6.3, 5
Morton, Craig 10.2, 4.4, 2
Namath, Joe 17.6, 4.8, -17
Nelsen, Bill 8.4, 3.9, 0
O'Brien, Ken 6.7, 3.2, 42
Palmer, Carson 12.4, 4.6, -6 // 12.4, 4.6, -7
Parilli, Babe 3.9, 2.7, -14
Pennington, Chad 10.6, 3.6, 13
Plum, Milt 16.3, 6.2, 16
Plummer, Jake 13.2, 5.0, -22
Rodgers, Aaron 19.3, 8.7, 25 // 20.6, 8.7, 31
Roethlisberger, Ben 13.1, 6.1, 3 // 13.1, 6.1, 5
Romo, Tony 17.8, 6.4, 6 // 20.1, 6.4, 6
Rote, Tobin 8.1, 6.8, 12
Ryan, Matt 17.0, 5.8, 15 // 19.5, 5.8, 16
Rypien, Mark 14.0, 4.6, 6
Schaub, Matt 19.5, 5.9, 10 // 20.7, 5.9, 12
Schroeder, Jay 7.3, 2.7, -1
Simms, Phil 8.6, 3.7, 29
Sipe, Brian 17.4, 5.4, 16
Snead, Norm 14.3, 5.1, 2
Stewart, Kordell 10.5, 6.9, -6
Testaverde, Vinny 11.9, 4.6, -39
Theismann, Joe 9.0, 3.1, 24
Thompson, Tommy 11.7, 5.2, 18
Todd, Richard 4.5, 2.9, -25
Vick, Michael 11.1, 4.3, 7 // 11.1, 4.3, 5
Wade, Billy 8.2, 4.6, 20
Waterfield, Bob 14.8, 5.9, 11
White, Danny 6.3, 2.6, -5
Williams, Doug 3.4, 2.0, 20
Zorn, Jim 7.3, 3.6, 10

1. Peyton Manning

80.8 9.5 (33 good)

There's never been anyone like Peyton Manning before, and I don't know if there will be again. Marino is freakish in his own right at 64.3; Luckman benefited from being a quarterback pioneer, and his number of 58.4, if not skewed, is an apple and oranges number. A score of 50 should really be vying for the all-time best, and yet Peyton zoomed right by that and hit 80.

It would probably surprise many football fans, too, to know that throughout the game's history, careers of 15+ years are not unusual for quarterbacks. Of course, one can argue whether Peyton's incredible record of starting every game meant that he was more likely to meet an earlier end or a later end than another vastly successful but less durable quarterback. But Peyton seemed on his way to going right on to 100, and may still. The neck injury is likely to either end his career or become a non-issue.

Peyton (I call him Peyton because of Eli's and Archie's right to Manning as well) was precocious, scoring 5.7 in his 2nd season, and 7.4 in his 3rd. It is in particular his run from 2004 to 2009, or 2003-2010 if you take a more tolerant view, that is just unprecedented. He produced six seasons in a row that can comfortably be called great (seasons of 7 or more). His career-best season of 9.4, the 49-touchdown '04, has actually been surpassed by 12 players. Describing Peyton as consistent seems not laudatory enough for his numbers, but his greatness is stationed there in a sense.

Because of interception differential, I do not think it is absolutely right that Peyton is "the best ever," although he would be my pick, both based on analysis and intuition. Marino saved 24 more interceptions over the course of his career. Montana saved 38 more, and there are arguments for Sid Luckman. Still, as quarterbacks churn touchdown after touchdown in this era, it is easy for them all to blend together, and I am here to say that there is no one like Peyton Manning. He should not be thrown over for the newest flavor of the month. Saying that Eli could be as great as Peyton if the Giants win the Super Bowl is just a case of losing all perspective.

2. Dan Marino

64.3 10.8 (57 good)

I was a huge fan of Marino's growing up watching him as a kid, but like I was with most of my heroes, I fretted over his performance and transferred my inferiority complex over to him. I was always afraid that he would be found out not to be as great as some people said he was. It was great pressure to be Dan Marino, particularly after 1984. I was always nagged by the idea that Marino simply put up big passing numbers without passing terribly well, and this was why the Dolphins never made another Super Bowl.

Compiling this information has enabled me to see my inferiority complex for exactly what it was. Marino was unambiguously great. He did everything well. He threw the ball well, on many attempts, stayed away from interceptions, and was brilliant at avoiding sacks. It all made for outstanding yards per drive.

Marino was a star who shone early, and his numbers generally show gradual decline through his career. But he did have a renaissance in the last Shula years, scoring 6.1, 7.7, and 6.2 in his final three years with the steak-house man. Marino's yards per drive were positive in the first 14 years of his career (excluding 1993, when he didn't qualify as a starter).

3. Sid Luckman

58.4 11.9 (32 good)

That Luckman was the most dominant of early NFL passers (at least after 1941) was a surprise to me. I thought Otto Graham and Sammy Baugh would rate as the all-time greats. But the rating shouldn't be jarring when you consider that Luckman was the Bears starter in nine seasons, and they won the NFL Championship four times. They also had an 11-0 season in '42 and lost to the Redskins in the championship game (with Luckman apparently not playing). He also led the league in yards passing three times, touchdown passes three times, yards per attempt five times, and completion percentage once.

He can be discounted as a pioneer, or as the product of a system that was ahead of everyone else. To take the idea to the extreme, if only one team was allowed to pass, and Luckman was on that team, identifying him as the best ever would seem simply a formal argument. On the flip side, Babe Ruth usually is given a pass as a pioneer and actually given credit for outhomering all of the other teams in the league and so forth (I can't say that I ever thought he deserved special credit for that; it's not like it was harder to hit home runs in those days. Not like the fences were 1000 feet away and that's why other players didn't reach them.)

Luckman also excelled in 1942-1944, which were war years. But he scored a sensational 11.9 in 1941, 8.6 in '45, and 6.8 in '47.

The bottom line is that, if you are going to ask who was dominant by the record book, Sid Luckman has to be near the top of the list. His 58.4 score came over just 8 seasons. His average seasonal score of 7.3 exceeds Peyton Manning's 6.2, and Steve Young's 5.9.

His total does not include 1940, excluded only because the Pro Football Encyclopedia started yards per drive the next year. 1940 was another Bear's championship year, in fact the year they beat the Redskins 73-0 in the championship (talk about having an edge on an opponent, and being years ahead).

4. Steve Young

52.8 11.0 (34 good)

I actually began this project first wanting to confirm that the 49ers offense truly was better as defined by YPD under Young than under Montana. In terms of yards per pass and the other categories that comprise passer rating, I knew this to be true. Yards per drive gives the same verdict. Under Young, the 49ers outdid the conference average in yards per drive by 7 or more five times; they did this just twice under Montana. Montana never came closing to reaching Young's scores of 11.0 in '92 and 10.4 in '93. Young and Kurt Warner are the only quarterbacks to post double-digit YPD differentials in consecutive seasons.

We give the nod to Montana over Young because of the four Super Bowls to one, and because Montana came first. How can Young be downgraded as a clutch player, however, when he threw a number of touchdown passes in the Super Bowl that was historic for any playoff game, let alone a Super Bowl?

Young was good at avoiding interceptions, although not as good at Montana. So this would be the argument for Montana as the better player, and lend some credence (although I don't think very much) to the idea that Young put up some stats that were better, but did not play better all-around, team football.

5. Drew Brees

50.0 11.7 (24 good)

The best of the juggernaut offenses of 2011 was not the Packers but the Saints, and I believe their yards per drive of 42.4 is an all-time NFL record. The differential vs. the league average of 11.7 "only" ranks 5th all-time, however. Adding 2011 to previous seasons of 7.5, 7.2, and 6.7 means that Brees has compiled a very Peyton-like run. But he'll need to add a handful of additional great seasons in succession to match Peyton. Brees was very good as well from '04 to '07, totaling 16.4 during that span.

I don't know that Brees is a legitimate contender for "greatest ever." Number one, he has to take a backseat to a player like Peyton who's played more seasons. Number two, his interception advantage is a little low for an all-time great.

6. Joe Montana

49.3 8.4 (71 good)

For Montana fanatics (and as someone who knows his fortunes every season in great detail, who knows what "Montana 1986" and "Montana 1989" mean quite exactly, I am one), the career high came in 1984.

7. Tom Brady

47.3 12.2 (50 good)

8. Dan Fouts

46.0 12.0 (16 good)

A Hall of Famer without question but probably not the 8th-greatest quarterback of all time. Interception rate unremarkable and missed 13 starts while getting credit for team's 14.8 score from 1983-1985. 12.0 season came in the strike-shortened 1982 season. The 37.8 YPD for that year was still remarkable; it was better than the '84 Dolphins score of 36.5, and would have led the NFL in 2009 and 2010, been 2nd in 2007 and 2008, and 5th this year. The '82 Chargers could go up and down the field with the best teams of today.

9. Norm Van Brocklin

44.3 11.0 (22 good)

It was heck trying to separate Van Brocklin from platoon-mate Bob Waterfield, as you can imagine; Van Brocklin showed as the more active passer in any year when there was a question, so the 44.3 is the ceiling for his score. It puts him tops for his era; none of the guys ahead of him were active when he was, although you can technically count Sid Luckman, who was really winding down in 1949 (Van Brocklin's first year). Van Brocklin with the Eagles was positive all three years, but barely so.

10. Fran Tarkenton

43.6 6.1 (83 good)

It was impossible to come away from charting Fran Tarkenton without adoring him. I think my first idea of him was a guy who just passed for 47,000 yards because he played forever, and my second was of a short guy who scrambled magnificently and was more entertaining than effective. These understandings couldn't have been more wrong. Who would have thought an improvisational player would have saved the second most interceptions in history?

There's been talk about the best quarterback in Giants history, with Tarkenton taking a backseat to Eli Manning, Phil Simms, and Chas Connerly, if he's mentioned at all. Well, Tarkenton spearheaded excellent offensive teams in New York, scoring 6.1, 2.5, 3.3, and 4.9, when those kinds of numbers were rarer than what we see today. Certainly, Eli could be picked first because of length of his tenure, but Tarkenton's data per year is more impressive. And that's before you get to Tarkenton saving a whopping 30 interceptions with the Giants, while Eli has been -10 over his career.

Over a period of 15 years, Tarkenton missed five starts. His score was positive in 14 straight years, with the only chink in the armor being a top score of only 6.1. I certainly hope whatever stains the cartoon-like images of Tarkenton have left for me are not permanent.

11. Johnny Unitas

41.5 6.5 (53 good)

Neither Unitas's career total or interception differential leave any doubt that he was an outstanding player. I've heard such different things about him from analysts. Allen Barra and George Ignatin said over a couple of decades ago that Unitas was good, but Starr even better in the '60s; Brian Kenny finds vindication in the numbers for the awe that most ex-players and writers feel for his play.

Kenny's approach is era centered, but depends on the player's rank in a number of different categories year by year. I guess it's possible I don't duplicate his ranking because Unitas was leading the league or ranking highly but by small margins. I don't just give a '1' for YPA, but a distance from the center.

Regarding Unitas's year by year statistics, the Pro Football Encyclopedia shows him particularly dominant in the '57-'60 period, perhaps largely because he led the league in pass attempts three times. YPD is generally buoyed by lots of pass attempts, but Unitas's numbers in those years were 3.6, 2.9, 3.2, and 2.2. Consistent and very good, but hardly dominant (he was actually 5.4 the next year, despite 16 TD and a NFL-high 24 interceptions). Unitas led the league in yards per attempt just twice, in 1964 and 1965: his yards per drive differentials in those years were 4.0 and 2.1.

His high-water mark of 6.5 came in '67, when the Colts were 11-1-2 in the year of the Ice Bowl Packers. They only rushed for 3.7 yards per attempt, so Unitas does deserve a lot of the credit.

Granted no one can "era adjust" on the fly, but the Unitas legend has never quite added up in the statistics for me. For one thing, after 9-3 championship seasons in '58 and '59, the Colts were mediocre until '64, Shula's second year with them. Under Shula, I have to acknowledge that their won-loss records were outstanding.

There are several quarterbacks who appeared immortal watching them, but don't quite have numbers consistent with their apparent playmaking skill. To Unitas, we can add Favre and Elway. Not only did Unitas look more in command and more expert than just about anyone else, he had the whole clutch aura around him too, and the combination made his supporters impervious to the data.

Unitas was in positive territory on my measure for all 14 seasons he started.

12. Brett Favre

41.1 6.6 (11 bad)

Didn't I just mention him? The career total and seasonal best are virtually identical to Unitas's profile, although the interception differential trails Unitas by 61.

I bet a lot of people would have thought it worse; Favre was actually throwing interceptions at around the league average for his career. The problem is that near the league average leaves him with the worst differential of all of the quarterbacks in the top 30 in my yards per drive statistic. You'd sort of think that interceptions are just a matter of discipline and only weakly connected with being a good passer. Nothing could be further from the truth. Interceptions evidently point to a deeper issue of control. Accuracy is certainly a part of this; if you don't know where the ball is going, you're not going to complete passes, and you're going to throw some to the other team.

But I think the inverse correlation between yards per drive and interceptions goes deeper than this. I statistically modeled a few years ago, and interceptions are not directly lowering yards per drive by ending drives. I can see why people would think that, but it's not the case. The guys who keep the ball away from the other team also have higher yards per attempt, and that's why they're moving the ball better.

Another reason why the top quarterbacks in yards per drive are outstanding as a group with avoiding interceptions is that length of career has a lot to do with avoiding interceptions. Throw interceptions, and you won't be around to complete the seasons where yards per drive diffential accumulate. But again, I really don't think that goes very far towards explaining why the best guys in yards per drive are some of the best in interception rate as well. Simple longevity and seasons as a starter will NOT put you on the yards per drive list; I could cite many examples for that. The guys who lasted a long time with moderate reputations generally had interception differentials of around 10 good; the killers are liable to be much higher in interception differential.

Now, we'll come across examples of this, but the big question is what to think when a quarterback excels in interceptions, but not yards per drive. Which do you believe more, in judging his quality? Did he compile the great interception rate through exceeding caution? Or was he a very good player, who simply didn't have the players around him to move the ball up and down the field like some of the other quarterbacks? Probably both factors are always at work to one degree or another, but I think more times than not, the low-interception quarterback, mediocre yards-per-drive quarterback is a very good player in a bad system.

Whether it's fair or not, we have a sense which explanation is right for various players. Ken O'Brien was cautious more than untapped; Roman Gabriel, Donovan McNabb, and Phil Simms quality players who didn't have the surrounding parts or the green light to show what they could do. The fact that they didn't lose games often was a testament to their ability. Interception rate may be more of of a reflection on an individual than yards per drive. (Now this doesn't mean that yards per drive has a lot to do with the running game. The evidence is not consistent with that. But I do think yards per drive has a tremendous amount to do with coaching, and probably a little bit to do with receiving and passs blocking as well.)

Back to Brett Favre -- YPD shows 1995-1999 as his likely peak, as you would expect, although he had a better season than any of those in 2004 (huh?), 2007, and almost 2009. The first MVP season was a 4.5, but the Super Bowl season of 1996 a very underwhelming 2.2.

This was a season of offensive parity in the NFC so extreme that it was comical; the Philadelphia Eagles, with Ty Detmer and Rodney Peete at the helm, nipped Green Bay for the league lead by 0.5. Green Bay won 13 regular-season games because they were the top defensive team in the conference in yards per drive, and picked off a conference-leading 26 balls. To be fair, Favre also saved 5 interceptions, with an INT rate a full percentage point under the league average.

On the field, Brett Favre rarely made things look easy, and I had that sense too compiling his data. It was a long journey to get to 41.1.

13. John Hadl

38.4 6.5 (1 bad)

Early AFL ('60-'64) 1.9 of total

A dead ringer for Favre on the three numbers, Hadl has never even been a Hall of Fame finalist. Hadl was clearly respected in his day, making seven Pro Bowls, and making first or second team All-Pro three times. The strikes against him are presumably

1. Benefited from Sid Gilman, in his day a Bill Walsh type.

2. Benefited from Lance Alworth, whose dominance in San Diego did include almost all of the Hadl years.

3. Didn't win much.

4. Threw interceptions (he was to the league average, as you can see).

I won't take them one by one. As far as where the Chargers fit into the league's power players during the Hadl era, they lost the AFL title game in his first two years as starter, losing to Buffalo both times. Subsequently, they finished 3rd in the Western Division three straight years, finishing behind Oakland and Kansas City each season. They were mediocre and no match for those two teams.

In 1973, when Hadl was 33 he joined the Rams. He scored 5.1 on my statistic, and the team went from 6-7-1 to a division best 12-2. They were beaten by the Cowboys in their first playoff game. Hadl only had a couple of more years as a starter, neither making team or individual impact.

Hadl's 1973 with the Rams is some evidence what he could do away from Gilman and Alworth. Also relevant: in 1970, he was without Gilman as coach; he had a small positive differential, of 1.7. In 1971, with Gilman back but Alworth gone, he had a very good season, at 4.0.

The interceptions are what they are, but this guy is 13th all-time in career YPD. He wasn't an early AFL phenomenon, either, like George Blanda, as he lit it up after the common draft, including a 6.5 in the regular season preceding the Jets' Super Bowl triumph.

14. Warren Moon

37.0 9.1 (13 to the good)

Certainly fits in beautifully to make a class with Favre and Hadl, although Favre and Hadl never had huge seasons like Moon did in 1990 (9.1) and 1992 (Houston was 8.8, although Moon only started 10 games). Moon always leaked some interceptions, like Hadl and Favre.

My sense is that he was on plenty of winning teams, but he only was the quarterback for 10 or more wins twice, and he certainly never made the Super Bowl.

There is much more to celebrate than to fault here for a great player, however. Moon turned 34 during the 1990 season, and his career score before that was 3.1. Hadl was wrapping up his career at 34, remember. Moon's total score late in his career is probably the all-time best, however you define late in the career.

Now, he wasn't in Canada until age 34. A 0.4 differential in his NFL rookie season is all that prevented him from starting his first five seasons in negative territory. So the run-and-shoot deserves lots of credit for Moon's success. I have a feeling, however, that poor surrounding talent and management had a lot to do with his earlier failure.

Moon was good with Minnesota and Seattle as an old player, too (2.3, 2.9, 3.1, and -3.0), although perhaps not as good 4264 and 4228 yards passing in seasons with Minnesota would indicate.

15. Daunte Culpepper

35.7 10.1 (0 good)

How did Daunte Culpepper lead one of the most prolific stretches of offense of any team in NFL history with little credit or even acknowledgement of the fact? The 2000-2004 Vikings were treated as so within the ordinary that I myself remembered few of the basics about them; I remembered them as at least a winning team, for instance. They weren't, and that is the first reason we probably don't give Culpepper even 5% his due. In 2000, Culpepper's first year as a starter, they went 11-5 and lost the NFC Championship game to the Giants. Their best record in the four other seasons with Culpepper as starter was just 9-7. While the Vikings were over the league average in yards per drive by 1.6, 6.6, 8.4, and 10.1, their defense was allowing 5.7, 1.9, 3.5, and 8.5 more yards per drive than the league average. It was an awful defense. It was the same story in 2000, with the offense carrying the defense, scoring 10.1, while the defense was -5.9. It is a fair point to note that Culpepper's record as a starter in Minnesota was 38-42, despite the offense being more brilliant in yards per drive than the defense was inept. However, the problem was not so much Culpepper's interceptions, as he is at the league average in that for the seasons in which he played.

Another question is if Culpepper's individual numbers jive with the Vikings' sensational yards-per-drive statistics. He led the league in yards per attempt twice, in touchdown passes twice, and had a completion percentage of 64.4 with the Vikings, including 69.2 in 2004, coming close to the all-time NFL record. If he was a guy who just threw the ball up to Randy Moss, is a 69.2 completion percentage consistent with that?

Culpepper's rushing statistics (yards, average, and touchdowns) are also all very good and probably contributed positively to the team. In 2002, Culpepper had an off year throwing the ball, yet his statistic was what most players dream of, 6.6. Eighteen touchdowns throwing didn't even have him in the conference's top five in that statistic, but if you added in Culpepper's 10 rushing touchdowns, he leapfrogged everyone in front of him except Aaron Brooks. This is more than an exercise; it is certainly realistic to think that Culpepper could have thrown for touchdowns on some of the ones he took into the endzone himself.

I'm of two minds in contemplating the quality of Culpepper's supporting cast. He had the exquisite Robert Smith behind him in 2000, and he had an aging Cris Carter through 2001. Other than those guys for a brief period, I see forgettable players starting at running back (Moe Williams and Michael Bennett), tight end (e.g., Jim Kleinsasser), and second receiver (e.g., Nate Burleson). He had one very big help, though, in #84, Randy Moss.

It's hard to resist the idea that Moss was one big powerful force who literally drove offenses to an extent thought beyond a receiver's control. Randall Cunningham's six previous seasons as a starter produced five negative scores, but in 1998, he was at 9.2, a Culpepper-like number. That was a great year for Randy Moss, and his rookie year.

Moss left the Vikings after 2004; the next year, with Culpepper starting seven times, then hurting his knee, and Brad Johnson starting nine games, the Vikings offense plummeted to 2.6. It might be of interest that Culpepper remained vastly more dangerous than Johnson, beating him in yards per pass attempt, 7.2 to 6.3.

The final example of the "Moss reality" came when he came to the Patriots, and they set a new record for points scored in a season.

Culpepper, however, is hardly the only player high on this list who had help from a great player, not to speak of a few of his friends.

Could we compare Culpepper to Kurt Warner? You'll see that Warner's close behind Culpepper on this list. Yards per drive shows him as at least slightly better than Culpepper during his brilliant peak (the three Rams years), but those years made up 84% of his career score. Warner was clearly a guy who had help from players with lightning speed and Hall-of-Fame savvy, and played in a system that churned out "in the zone" quarterbacks. Warner was no better at avoiding interceptions than Culpepper (2 to the good for his career).

I'm not sure actually that being as good as Kurt Warner is a huge compliment, but I think the comparison is apt. Warner is probably overrated. Culpepper is unquestionably underrated.

Culpepper suffers just as much historically from what happened after he left Minnesota as for the .500-or-so record during his tenure. Barring a Peyton Manning kind of injury or a car crash, it just doesn't seem like anybody great and not of an advanced age could fall off as dramatically as Culpepper did.

I hear this argument, but the man did hurt his knee, let's not forget. He has one of the most interesting encyclopedia entries of any player, and it will get increasingly interesting with time. I think there are some very definite but very incorrect reasons he has been underrated, and I do not see his sensational score here as a mere curiosity.

16. Roger Staubach

35.5 5.5 (48 good)

Like Tarkenton, another player whose record jumps off the page for me in a way that Unitas's doesn't. He started eight years for the Cowboys, with a positive score every year (indeed his worst score was 3.0), and a top score of 5.5 but an average improbably not far behind (4.4). His interception savings were 10th among all of the players that I calculated, but considering this was achieved in a relatively low number of passes, it could be considered the second best mark ever, and certainly in the top five, with Roman Gabriel and Joe Montana. (Aaron Rodgers is an early candidate to join the group).

It's hard to know how much to credit Cowboy support for Staubach's numbers; I think that was certainly a factor. Don Meredith scored 12.0 total as the Cowboys starter from '66 to '68; Craig Morton totaled 9.8 taking his efforts in '69, '70, and '72. Morton wasn't worth anything, yards-per-drive wise, after he left the Cowboys (well, technically he was worth 0.4). On the other hand, neither of those guys was quite as strong in YPD as Staubach, and there was a definite fall-off in the Danny White years, even as the Cowboys continued as one of the best teams in the NFL; White was typically around 1.0, and he threw interceptions at a problematic rate.

17. Sammy Baugh

35.5 13.7 (58 good)

First, Baugh was the main passer for three Washington teams that predate the available data, so his true score is higher. I'm probably equally ill-equipped to understand even a small part of his place in NFL history, but hunting and pecking, I have found a few things that are at least interesting to me.

First, although I don't believe in rating any players, including quarterbacks, by team championships and won-loss records, I think it's necessary context. It's sort of where you start; you have to be able to explain where and when there's a disconnect with the statistics and that.

Baugh played from '37 to '52, exclusively with the Redskins. The team success came in the first half of his career. Although he generally was one of the most active passers in the league (i.e., pass attempts), we know that the offense was not T-formation until probably 1944. Sid Luckman and the Bears were the first to do the T in the NFL. Oddly, for one of the only guys who could really pass it, the Redskins were a big winner before they went to the T more so than after it. Baugh's attempts are consistent with going over to the T; he threw 354 times in 1947, and 315 times in 1948 (both higher numbers than Ben Roethlisberger had in 2004 and 2005), but the Redskins' last championship appearance with Baugh occurred in 1945. Baugh figured very prominently in the leading passer stats after '45 (in fact, you could even call him dominant), but the team wasn't good. His yards per drive numbers after '45 are excellent but not great: 2.0, 4.9, 5.5, 2.9 from '46 through '49.

1945 is the big point of interest. This is the greatest single-season score of the study -- 13.7. He also completed a crazy 70.3% of his passes (Sid Luckman ran 2nd at 53.9) and was intercepted just four times, with an interception rate less than half as high as the next guy's. The Redskins went 8-2, losing the championship to Bob Waterfield's Cleveland Rams. They were only fourth in the league in scoring in a 10-team league, so in a sense this is a weird team to single out as tops offensively. But with so many completions, and so few interceptions, it's no wonder they didn't turn the ball over often, by downs or otherwise. Indeed, the Redskins had 12 fewer drives than anyone else in the league (the Chicago Bears, who were 8.6 in yards per drive, and garnered even more first downs than the Redskins, had the second fewest). The Redskins' running was not sensational, with 4.3 yards a rush, tied for 2nd in the league in a very spread out group (half the league was under 3.0).

Certain things definitely stand out about this offense, but I have to say that I still regard the huge yards per drive as a little bit fishy (by which I mean not just misleading as to their overall effectiveness, but subject to the vagaries of 1945 data collection, and perhaps in error). It seems relevant that the number doesn't fit at all with the Redskins' surrounding seasons -- of Baugh's 10 marks, his second best is 5.5.

Finally, the degree to which Baugh was ahead of the league in general when it came to interceptions should also be stressed. (The last thing I want to do is offend any relatives or friends of relatives of Sammy Baugh.)

18. Kurt Warner

34.2 10.5 (2 good)

His Rams career is thoroughly known (although I am surprised looking at his incomplete 2002 and 2003 statistics that he was really apparently not playing well; I thought virtually anyone could do well in that offense -- even an out-of-sorts Kurt Warner). His score with the Giants in his one season there was -2.1, but I'd bet it was closer to even before Eli took over and proved unready. His Arizona scores were 1.5, -0.1, 4.0, and -1.1. A negative in his final season is surprising, but a couple of these scores probably would have been better again without the meddling of Josh McCown and Matt Leinart.

Warner's interception statistics with the Giants and Cardinals seem to come from a different player than his interception statistics with the Rams. But you knew that.

He was so good right away with the Rams that we blame Mike Martz for his interceptions, but he may not have been seasoned in every respect with the Cardinals. On the other hand, an interception rate getting worse not better after his first season in St. Louis suggests experience was not the problem from 2000-2002.

19. John Elway

29.4 7.1 (42 good)

I think there are separate parallels between Elway and Brady, Elway and Unitas, and Elway and Bradshaw. Well, maybe the parallels between Elway and Unitas and Elway and Bradshaw aren't separate; all were thought of as prodigiously talented, clutch players, but the numbers (to my way of thinking) show them as something less. Unitas's numbers are certainly better than Bradshaw's, though, and even with Elway's strong finish to his career, probably better than Elway's. Elway and Bradshaw go together better than Elway and Unitas because Elway and Bradshaw were both the first player drafted, both were good runners, and both had exceptional arms. Unitas wasn't drafted high and was probably more of a smooth passer than a "tools" guy. Elway and Unitas get special points because they have the same first name -- maybe that plays on me in seeing them together.

I thought of Brady for Elway because I look at Elway's seasons, the total of which summed to 8.4 after 1994 (12 years in), and I think, "Gee, I'm glad he got to show he was good. I'm glad he got to prove it." He'd been to three Super Bowls by that time, but you couldn't see it in his numbers (although his interception body-of-work was good). Brady also had that team success early, and had a burgeoning reputation but didn't fully turn around his numbers until 2007 (and he did so with aplomb). Elway languished statistically for longer, and didn't come up as far as Brady did, but it's a similar pattern.

Elway blamed Dan Reeves for his first career's statistics. This seemed spot-on when Wade Phillips became the head coach and Elway led the NFL in passing yards in 1993. But his YPD differential was only 1.5 that season. Two seasons later, with Mike Shanahan as coach, it was 7.1, and the '96-'98 seasons weren't far behind that. Elway averaged 5.3 with Shanahan as head coach.

In Shanahan's remaining 10 years with Denver, eight of the seasons had positive scores, with an average of 4.5 in those years. You'd still have to take Elway's performance as the best of the bunch of Shanahan quarterbacks. But Jay Cutler scored as high as 8.6, and Chicago's offense is dependably in the bottom quarter of the league since his arrival. Statistically, Jake Plummer was a bad player apart from Mike Shanahan. Brian Griese is the third guy who was a season's starter for Denver. His numbers with Shanahan were good, but he fell out of favor, and took on the role of journeyman. Despite Shanahan's failure to ignite the Washington offense, I'd love to see him with a really good quarterback there. I guess we saw his son with Matt Schaub, however, and that made for some pretty serious offense.

19. Boomer Esiason

29.4 6.7 (9 good)

The claim for best offense in the AFC in the second half of the '80s came down to Cincinnati (25.3 total under Esiason from '85-'89) and Miami (26.3 total during that time). Although, for his career, Marino saved his team 57 interceptions and Esiason saved just 9, there is not much to choose from in their interception rates during the mid and late '80s.

21. Bob Griese

28.7 5.7 (13 good)

I'm certainly reminded a little of Staubach in Griese's season-by-season scores. They are close contemporaries, too -- facing each other in the '71-'72 Super Bowl, and retiring after 1979 and 1980, respectively. 1971 was also the best season in terms of YPD differential for both: 5.5 for Staubach, 5.7 for Griese.

Once Don Shula came to town, Griese was, like Staubach, marvelously consistent, scoring at least 2.9 in his first seven seasons, but never above 4.0, except in '71. (The famous '72 Dolphins, led principally by Earl Morrall while Griese was on the sidelines, scored 6.3.) Griese's pattern is pure Staubach, but his 19% lower career YPD total in 50% more seasons tells you he wasn't as good.

Staubach certainly came into the lineup ready to play, however, and also knew when to retire. The per-season difference between the two probably overstates the difference, but the difference in career totals underestimates it; Stauback overcame being in the Navy. In terms of interceptions, Griese and Stauback were certainly different breeds, too. (Not that "13 good" in 3429 throws classifies a quarterback as an interception problem, even in this elite company.)

What Griese's YPD seasons show more than anything else is that a run-heavy offense can score well if the quarterback passes well. (I'm assuming here that what you hear about Miami running so predominately is true, although I haven't checked it out. It was obviously true in some playoff games during the Super Bowl years at least.) The dichotomy between YPD and pass attempts might not be so clean as the dichotomy between yards per pass attempt and pass attempts, where the two are independent, or even inversely related. But Griese's numbers show that run-heavy quarterbacks are not at an insurmountable disadvantage in YPD.

22. Troy Aikman

28.5 7.8 (19 good)

23. Ken Anderson

28.4 7.4 (59 good)

Brian Kenny champions him for the Hall of Fame. In fact, I think Kenny would have him outrageously high (in some people's mind)s among the Hall of Famers themselves. My YPD methodology doesn't show Anderson as a Hall-of-Fame slam dunk, but his interception savings stand fourth all-time. Super Bowl XVI was actually a contest between #3 all-time in interceptions savings (Montana) and #4 (Anderson). I'm not sure whether the game went to true to form in that respect or not: there were two interceptions in it, both thrown by Anderson. His other strength as a player was his high completion percentage, so Anderson was the classic high-completion-percentage, low interception quarterback. Anderson set a record in completion percentage in the strike-shortened season of 1982, and that is also the year that he recorded his 7.4 YPD differential.

23. Trent Green

28.4 8.9 (6 good)

An 8th-round draft pick, Green started one year for Washington at age 28. His score was -2, but with 23 TDs and 11 interceptions, St. Louis tapped him to be their starter the next year. He blew out his knee in pre-season and missed out on Marshall Faulk, Torry Holt, Isaac Brice,, and Mike Martz. Back as the starter with Kansas City in 2001, Green compiled a stellar five-year run, including scores of 8.9 in 2004, and 7.3 in 2005.

Certainly, he had a coaching staff who gave him every chance to be successful, but in my observation he fit in with the group of current quarterbacks whose precision leaves the defense helpless. He was certainly a guy who scared you if he had the ball and you needed a stop.

25. Philip Rivers

28.3 8.7 (13 good)

Leads his 2004 draftmates Eli Manning and Roethlisberger with room to spare and can be among the all-time greats if he keeps his form for three or four more years. He would then be where Brees and Brady are now, I guess. He reminds me a bit of Marino, not just for his throwing motion, but because his team's success has been real but somehow unsatisfying, and because he puts up numbers with just about anybody, while sometimes not giving you that unstoppable feeling I mentioned with Trent Green.

It's not quite the mystery of Washington in 1945, but the perception of San Diego's 2011 is at odds with their 8.7 YPD differential. The top three teams in raw YPD (not comparing to a separate average for conference, in other words) were New Orleans, New England, and Green Bay, with San Diego closer to Green Bay than New England was to New Orleans. Football Outsiders has Carolina the fourth team on their method for rating offenses, with the expected healthy gap from 3rd to 4th. They do have San Diego 5th, though, and they penalize for turnovers, so perhaps the Football Outsiders ranking is more consistent than inconsistent with the picture presented by yards per drive.

26. Jim Hart

27.5 6.3 (17 good)

Hart's rating surprised me, but I'm not sure if this was my own myopia, or if he is truly underrated in general. I would gather, though, that with just two 3,000-yard passing seasons and a 51.1 career completion percentage, not to mention no Super Bowl appearances, he gets lost in the shuffle. You do hear about him sometimes as one of the prolific Don Coryell passers, I think, and he made the Pro Bowl four times in five years under Coryell. Those Pro Bowl seasons were also his four best in yards-per-drive differential, ranging from 3.3 to 6.3. He scored positively in all of his first 10 years as a starter, however, and there's not a catch to the yard-churning such as that he was an interception machine, which he wasn't.

27. Y.A. Tittle

26.4 6.8 (27 good)

Two years as a starter in the AAFC not analyzed. Tittle's best years in YPD were with San Francisco in 1953 and 1954, and the yields from those years account for half his career YPD total.

28. Bobby Layne

26.1 5.0 (12 good)

I think some people regard Bobby Layne as a "winner" who didn't have good statistics, but he was very effective for the 1950s era, racking up eight consecutive positive seasons. His prime was spent in Detroit, but his performance at the end of his career with Pittsburgh (1.9 in 1958, 4.1 in 1960, with 1959 and 1962 negative but not disasters), suggests he was capable of carrying his own weight.

29. John Brodie

24.2 5.1 (27 good)

Brodie shows well in both YPD Differential and interceptions, but only made two Pro Bowls, so perhaps he was underrated. The San Francisco teams under his stewardship were the absolute epitome of mediocrity until his last two years as a starter in 1970 and 1971. Curiously, 1971 shows up as the better YPD year, 4.9 to 1.5, despite statistics that look better in 1970, and an All-Pro selection in 1970. Brodie scored positively in his last seven years as a starter. He succeeded Y.A. Tittle as the starter in San Francisco, and produced the exact same interceptions savings as Tittle in his career, in just 96 more passes. He's also remarkably close to Tittle in career YPD, since only one player ranks between them on my career list.

30. Jim Kelly

23.8 6.2 (3 bad)

Kelly was very good from 1988-1992, scoring 20.6 in that five-year stretch. That's not quite Boomer Esiason good from 1985-1989, though, to compare him to the no-huddle operator who preceded him. The 1988-1992 stretch includes three of the Super-Bowl-losing seasons, as well as a 12-4 AFC Championship-appearing team in '88. Outside of '88-'92, obviously, Buffalo's offense was unremarkable moving the ball under Kelly. He's also only one of four quarterbacks with a career YPD over 20 who has a net negative interception record. He's a guy who you have to think benefitted more than suffered from the players around him, and from the offensive system. I feel that the compilation of my data gives me perspective on Kelly, while before he was lost in a haze of great players for me.

Aside from the fact that he didn't play for an extremely long time, I wouldn't have had any suspicion that his career score was under 40. I also had no idea about all of the interceptions, because I don't remember that tendency being a part of his reputation.

31. Bart Starr

23.5 5.9 (41 good)

There's probably such a thing as looking at numbers too closely, but I'm going to risk that with Starr. Some points

*The strength of his record is with Lombardi, yes (he was negative as a starter in four of five seasons without Lombardi), but particularly early in his partnership with Lombardi. His career high is his first NFL Championship year, 1961, when he scored 5.9. Each of his four subsequent seasons declined from there, going all the way down to -3.3 in 1965, when the Packers still somehow won the championship.

*While I don't embrace Unitas as a brilliant quarterback on the record book, in some ways Starr held his own against him, scoring higher than he did in five of the 12 years they were starters in the same year. In the five seasons Starr worked under Lombardi, and Unitas under Shula, the pattern was the same: Unitas was better, YPD wise, in three of the five years. An obvious difference in their records, and probably a large part of why Unitas scored 41.5 and Starr 23.5, is that Unitas was 14 for 14 in positive seasons, and Starr an unimpressive eight for 13. Starr's interception value approached Unitas, though, (41 good vs. 53 good), and with less than 61% as many passes; Starr contributed greatly to the Packers' dynasty with his smart handling of the football.

32. Jeff Garcia

22.7 6.8 (35 good)

We were certainly numb to quarterback success in San Francisco by the time Garcia took over in the City by the Bay, but Garcia's performance as the starter from 1999-2003 should be given serious attention. Perhaps I am not lowering average completion percentage and other passing statistics from their current zenith adequately in my mind, but it does seem to me that Garcia is an interesting case, because the 49ers under him seem to have been much better in YPD than they were in most of the passing statistics. Part of their secret may have been the low number of sacks that Garcia accepted -- just 22 per 16 games over his time there. Garcia was also something of a factor running the ball. When you add one of the better recent interception rates to the equation, Garcia was an impressive player.

The stars were never perfectly aligned for a big playoff run that might have elucidated this. The 49ers had losing records in Garcia's first two years as a starter because they had bad defenses. (I'm loathe to use points allowed as a barometer, but they gave up 453 and 422 points those two years.). In Garcia's third year as a starter, the 49ers went 12-4, but they had a tough road in the playoffs, consigned to the wild card as the Rams registered 14-2, and they lost the wild card game to the Packers.

Everyone who assesses Garcia of course must note that he played several years in Canada, and didn't become a NFL starter until age 29. His 22.7 score is even more impressive, then, and we can look at it almost the way we look at similar current players like Matt Schaub and Tony Romo, who have scores around 20 in unfinished careers.

Trent Green is also an interesting comparison, as he begun to play at a late age. He was even better at moving the ball than Garcia, actually, but he yielded more interceptions.

33. Daryle Lamonica

22.5 6.5 (8 good)

The Mad Bomber was a very good player, achieving his career total with just six starting seasons. He scored positively in every year. Based on my data, he was a better player than Stabler, who succeeded him as Oakland's quarterback, although they eerily end up side-by-side in the ratings. Stabler had 11 seasons as a starter, and his cumulative score with Oakland was 18.6 in seven seasons (by the way, moored to horse racing as I am, I at first wrote Oakland as Oaklawn. I was tempted to leave it.) Lamonica's career high of 6.5, achieved in the Heidi season of 1968, surprisingly beats Stabler's 9.3-yards-a-pass, Super Bowl-winning 1976, when he came in at 6.2. But the biggest difference between Lamonica and Stabler is that Lamonica, ironically enough given his moniker, was more careful with the ball than Stabler, who ranks 115th in interception differential among the 116 quarterbacks investigated, ahead only of Vinny Testaverde, who was a notorious leaker of interceptions, and three spots behind the likewise notorious Richard Todd.

Since Stabler's Raiders went an outrageous 51-12-1 with him at the helm before a couple of 9-7 seasons at the end, it's hard to argue too vociferously with the decision to turn over the team to him in 1973. He also was over four years younger than LaMonica, and probably in much better physical shape (bet that's the first time anyone ever praised Stabler's shape, in any context). But LaMonica's last season as a starter was his usual fine work, at 4.0, and he had an interception rate under the league average that year, too. He made just three more starts the next season before getting the boot. LaMonica's career record as a starter with the Raiders was 62-16-6 (move over, Tom Brady).

34. Ken Stabler

22.3 6.2 (36 bad)

Despite the fluid offense the Raiders showcased during the Stabler years, their excellence in the midst of all of Stabler's interception is certainly something that deserves further consideration. It wasn't just penalties with the Raiders, but turnovers, too, or at least interceptions. Maybe the interceptions came on passes deep down field, and the damage was contained. However, that would only be true to a dergree because the Raiders must have thrown plenty of short passes, too. They didn't play an altogether different game from the other teams; football just doesn't work like that.

35. Len Dawson

21.2 5.1 (34 good)

21.2 includes 6.2 from 1960-1964 AFL

Dawson led the AFL in completion percentage five straight years, and he led in touchdown passes four years out of five (the two five-year periods for these runs overlap but are not identical), so he is nowhere near the standout in YPD some might have expected. He was in-the-black for his first seven years as a starter, however, before tailing off after the Chiefs' Super Bowl win. He also was a smart quarterback, interception wise. The snapshot of him here makes his selection to the Hall of Fame reasonable, if not clearcut. But it's probably more of a "not-Hall-of Fame" profile than a Hall-of-Fame profile, even making some allowance for lower top YPD differential scores in Dawson's day than today.

36. Frank Ryan

20.3 6.1 (4 good)

The only Browns quarterback to reach 20 on the statistic (or at least of the logical candidates Graham, Plum, and Ryan; I'm not counting a guy like Jeff Garcia, whose Browns run was forgettable). The seasonal totals for Milt Plum ('58-'61) and Ryan ('62-'67) are really indistinguishable: Ryan hits 20 because he kept the job longer than Plum did. Ryan was a Blanton Collier quarterback, not a Paul Brown one, and had his best season in '66, without the other famous Brown on the team, either. That's pretty interesting, and speaks well for Ryan, although the Browns were still a great running team, with Leroy Kelly going for over 1000 yards in '66, while averaging a league-best 5.5 a carry. The Browns continued winning games after Ryan exited the starting stage, coming one game away from the Super Bowl in '68 and '69 with Bill Nelsen as quarterback. Their YPD Differential was down to just 1.4 in '69, though. Ryan was not as good at avoiding interceptions as Graham and Plum, owning just a "4 good" for his career.

37. Rich Gannon

20.1 7.5 (39 good)

Gannon's late-career emergence with Oakland included a very low interception rate.

THE REST ALPHABETICALLY

Steve Bartkowski

8.1 3.9 (15 good)

George Blanda

19.0 10.2 (25 bad)

18.7 of 19.0 early AFL

Blanda's 10.2 was his Player-of-the-Year, championship-winning 1961. That season used to be remembered, because Blanda threw 36 TD passes, which was a new all-time record, and not surpassed by anyone until Marino in 1984.

The amazing thing is that Blanda only started 11 games in '61, and the Oilers as a whole threw 48 TD passes, not to mention racking up 4,568 passing yards. They only played a 14-game schedule, so the composite numbers were on a par with teams like the 2004 Colts and the 2007 Patriots.

Blanda followed up his 36-TD season with 42 interceptions the next year, far and away the record for interceptions in a season (Testaverde is 2nd with 35 in 1988). Blanda would lead the AFL again in interceptions in '63, '64, and '65.

Drew Bledsoe

8.0 2.6 (9 good)

With the exception of the 2003-2004 Bills, who distinguished themselves negatively, Bledsoe tended to lead average offenses. His tools exceeded his production.

Terry Bradshaw

11.0 3.7 (11 bad)

Just two seasons over 2.0 (both were Super Bowl-winning years -- 1975 and 1979), and one of 2.0 (1980). I don't want to suggest that, even on numbers, Bradshaw was an awful player; he wasn't, being positive every year from '75 - '80. But the reverence with which he is held actually comforts me somewhat, because it suggests to me that I would have found the mindless search for winners and deliverers of Super Bowl rings just as prevalent in the '70s as I find it today. I don't need to pine not being born in another decade, nor to think we are actually going backward in collective insight.

How exactly do you defend Bradshaw as one of the greats? He didn't stand out moving the ball despite having a strong supporting cast, and he cost the Steelers more interceptions than he saved them over his career. If at least he had the interception rate going for him, he would be a candidate to be a hidden great player (see the Favre comment). I don't know if his post-season statistics would place him in different light than his regular-season ones, but regular seasons proved that the Steelers were good enough on defense to win prolifically despite what they got from their 'O': the Super Bowl champion Steelers all had regular-season records of .750 or better.

Ed Brown

10.6 3.7 (15 bad)

I won't pretend to really know anything more about a guy like this than you do, but I'm guessing you won't mind being saved a trip to basic reference sites, so I'll share what I've found. Brown was a Pro Bowler in 1955 and 1956 as a new starter with the Bears. He took over for George Blanda, but Blanda continued to see some action, including when the Bears were shellacked 47-7 by the Giants in the 1956 NFL Championship (Blanda threw 27 passes and Brown threw 20; somehow the Bears had more first downs than the Giants in the game, despite the final score). After 1956, the Bears were below-average in offense the next three seasons with Brown starting. Brown's second and last period as a team's starting quarterback also came directly after a Hall of Famer had manned the position; he succeeded Bobby Layne with the Steelers, and scored positively in both '63 and '64. Also punted 493 times over his career, averaging 40.6 yards a kick.

Mark Brunell

18.4 4.4 (40 good)

Quite impressive when you combine YPD with interceptions....15.8 points in his first six seasons as starter, just 2.6 in his last five. One thing I've noticed in the study is that a positive year is likely to be followed by a positive year, particularly with a young player. There seem to be very few quarterbacks who were good every other year.

Marc Bulger

9.9 5.3 (5 good)

I suppose this is as good a place as any to get one of the methodological issues out of the way, which is the method of using the season's statistics for YPD and then assigning them all to the starting quarterback, as long as he saw a majority of the action. Ideally, I would have data on what exactly a QB's yards-per-drive were when he was in there, but I don't have that. Then this number could be weighted by games played, or percentage of the season played, so that a 11-game starter in 2011 wouldn't get as much credit for positive seasons as a 14-game starter in 1965.

The net effect of giving quarterbacks credit for games they did not play varies from player to player, and not just because their durabilities differed. If the backup played as well as the starter when he was in there, simple inflation of career-score results. But if the backup didn't play as well, yes the starter's 69%-active season is counting as much as a 100%-active season, but the YPD differential will be lower, diluted by the backup's poorer performance. Since most backups are not as good as most starters, there is generally (and imprecisely) a penalty for not playing.

Bulger's career is very unusual because he was essentially a half-time quarterback in three seasons, but just missed out on playing a majority of the time in each of those seasons, and so has no opportunity to score. 2002 was Bulger for seven starts, Warner for six, Jamie Martin for two, and Scott Covington for one. 2005 was Bulger for eight starts, Martin for five, and Ryan Fitzpatrick for three. The 2009 Rams were not only 1-15, increasing their three-year-win-total to just 6, but -5.1 in YPD differential; Bulger started eight games that year. 2002 is when he missed out to some degree. Bulger had a league-leading 8.6 yards an attempt, and Warner just 6.5, for a team with a 2.7 YPD.

Paul Christman

7.8 4.1 (25 good)

Christman was the quarterback of the 1947 NFL-Champion Chicago Cardinals. He scored at 4.1 that year, and at 3.7 in 1946. In the championship game itself, Chicago prevailed 28-21 despite Philadelphia's quarterback completing 27 passes to Christman's 3, for 297 yards to Christman's 54. The Cardinals rushed for 282 yards, however; Christman didn't contribute there (and he virtually never did).

The Wikipedia entry says that "Christman was a notoriously poor ball-handler," referring to a supposed propensity to fumble. Ironically, when it came to interceptions, he wasn't Joe Montana, but he was ahead of his time; you can see he was 25-to-the-good, and that in a career that only numbered three seasons as a starter.

With 14.5 yards per completion, I don't think he was just throwing screen passes, but I'm not sure to entirely trust the meaning of interception rates so soon after the introduction of the T-formation. There might be some obvious explanation that had little to do with Christman's ability for his impressive interception rate. As important as systems are today, I suspect they were much more important still in Christman's era.

Kerry Collins

6.8 4.4 (4 good)

All of his points came with the Giants, where he was positive three times in four years as the starter. For a guy who was given the reins enough times to throw for over 40,000 yards while not being a good yards-per-attempt player, his score obviously isn't impressive, but yards-per-drive doesn't show him as as bad a player as I thought it would. My expectations weren't high, however.

Charlie Conerly

6.3 3.2 (34 good)

How can I say this? As a ball mover, Conerly was consistently and relentlessly bad, scoring negatively his first eight seasons as a starter. These included single-season disasters of -6.7 in his 4th year (1951) and -5.5 in his 6th (1953). If you reverse my procedure, and only add up his negative numbers, they sum to -27. It's a curiosity that anyone could stay in the lineup that long having so little success.

Perhaps coach Steve Owen must share in the blame for the offensive futility -- he'd been with the Giants since 1931, during which time the game obviously had changed tremendously. With Jim Lee Howell and Offensive Coordinator Vince Lombardi in, and Owen out, Conerly finally began to improve. He moved up to -1.7 in 1954, then was better still in each of his next three seasons. Finally getting respectable offense to assist their defense, the Giants won the championship in '56, while losing it in '58 and '59.

Conerly does show up as a game manager of sorts of the 1950s, with a very good interceptions savings of 34. Still, the seven Hall-of-Fame-finalist nominations he procured from 1971-1980 does not provide any confidence that the selectors knew what they were doing. I would bet that the 173-167 TD/INT ratio, made possible by the low interception total, played a role in the misreading.

Randall Cunningham

11.0 9.2 (25 good)

See the Daunte Culpepper comment.

Jay Cutler

10.9 8.6 (10 bad)

See the John Elway comment.

Cotton Davidson

6.1 3.0 (4 good)

All points from early AFL.

Appears to have been a pretty anonymous quarterback from the early AFL; a top-two conference rating in yards-per-drive with the Dallas Texans in '61 probably doesn't indicate anything deep about his significance to the game.

Steve DeBerg

13.6 4.8 (7 good)

Quite a solid profile. In The Blind Side, Michael Lewis tries to show how much Bill Walsh improved the San Francisco offense, including in '79 with Steve DeBerg. With 6.3 yards-a-pass-attempt and a 2-14 record, I wasn't buying it. It turns out, however, that the Niners were 3.7 yards-per-drive over the conference average that year. Montana became the principal starter in 1981 (of course winning the Super Bowl his first year), but he wouldn't better DeBerg's mark until '83. Pro Football Reference shows DeBerg as having only gotten sacked 17 times in '79; perhaps that goes some way towards explaining why the Niners were as effective as they were. They also put the ball in the air down after down; DeBerg led the NFL with 578 pass attempts. So, many passes, and short passes in lieu of runs, were the approach. (Sounds a lot like today's offenses, doesn't it?)

DeBerg qualified as the starter for 11 seasons, but only started more than 13 games three times: in '79, '90, and '91. I don't know how many games he missed because of injury, and how many because the team had another quarterback who seemed more exciting, but he missed games regardless.

1991 also rates as a very good year for him, at 4.8, and again, his summary statistics don't stand out, outside of just being sacked 19 times. If you think the Chief's running attack was the explanation for the driving average, their yards-per-carry in '91 was 4.3, not providing probable cause for the 4.8 YPD differential.

Ty Detmer

2.7 2.7 (3 bad)

Has the makings of a trivia question -- which quarterback led his conference in yards-per-drive in his only year as a majority NFL starter? This came in 1996, when he started 11 games and Rodney Peete 5, in the midst of forgettable years for the Eagles -- years neither good enough nor bad enough for those of us outside of Philadelphia to remember them. Between Cunningham and McNabb, a quarterback is going to get lost unless he was inherently interesting. (The trivia question comes with ready and appropriate hints: his brother was also a NFL quarteback; he won the Heisman Trophy.)

Looking at his career as a whole, I do not think Detmer was an especial diamond in the rough. Many guys who weren't that talented have registered better than a 2.7, including players I have no use for, like Kerry Collins. 2.7 just happened to lead the NFC in 1996.

I'm not a fan of what Jon Gruden does with offenses; Rich Gannon went particularly wild the year after Gruden left Oakland. And with Tampa, Gruden couldn't settle on a quarterback and seemed to make his offense cute and complicated rather than effective. But he was the offensive coordinator for the '96 Eagles, so the success with that offense makes his case stronger. It seems fitting, though, that Detmer would have had such limited staying power with Gruden as offensive coordinator.

The team regressed the next year, with Peete and Bobby Hoying getting their snaps in addition to Detmer. Sacks shot up from 39 to 64, with Detmer doing a much better job of getting rid of the ball than the other two guys: he was sacked once every 13.8 pass attempts, while they were sacked once every 8.6 pass attempts.

See also the Favre comment.

Lynn Dickey

3.8 3.5 (35 bad)

I can probably trace my uninformed decision to run Lynn Dickey on having had a 1984 NFL Record & Fact Book. I lived in that thing. In 1983, Dickey led the NFL in yards, passing touchdowns, and yards per attempt. On the other side of the ledger were an NFL-leading 29 interceptions. 1983 was his best year. Overall, Dickey was a bad starting quarterback, if not worse than that. He got eight years as a starter (although four with at least three starts missed), which says something positive for him, as long as the people making the call on him had any expertise at all.

Jim Everett

8.9 3.2 (2 good)

In general, Jim Everett's highs probably do not register as well as I thought they would, but his lows also don't reach the depths I had anticipated. The conventional understanding of Everett is that he was one of the bright, young stars in the league, then lost his heart and wasn't good for much for the majority of his career. My numbers don't quite bear this out. The only season where he was really bad was when he had a -4.3 in 1996, his last year as a starter. If you think interceptions were a problem for him, he never threw more than 20, and his yearly interception totals are not impressive but very consistent. The Rams clearly ceased being a competitive team after 1989, losing at least 10 games their last five years in Los Angeles, but I would suggest their offensive performance was not a primary reason for this.

Everett was considered by some (or at least by Jim Rome) as a chicken. He started 16 games an impressive seven times, and was a majority starter for 10 consecutive years. In other words, by the numbers, he was durable. Other quarterbacks with the reputation of having been faint-hearted also stayed in the lineup more than typical quarterbacks. One can argue this occurs because the reputations are undeserved, or because the approach of self-preservation that is there protects the player.

Joe Ferguson

13.2 5.7 (8 good)

Ferguson owed a huge assist to O.J. Simpson and the run blocking of his line in his rookie year of 1973. The Bills totaled 3,088 yards on the ground, just shy of 2 1/2 times what they threw for; adding in their sacks, they also averaged 0.89 fewer yards a pass attempt than a running attempt. Simpson was the rare running back who could carry an offense, however, leading the Bills to a 2.5 YPD differential. Two years later, Ferguson had his legs under him, and he and Simpson were a lethal combination: 25 touchdown passes for Joe, tied with Tarkenton for tops in the NFL; 1,817 rushing yards for Simpson (remember also that the season was just 14 games at this point). The Bills led the NFL in scoring. Their YPD differential was 5.7. But they went 8-6 and missed the playoffs. After 1975, Ferguson held his own as a starter through 1984, although the Bills were certainly cumulatively a bottom-half offense. I spoke of durability in the Everett comment; Ferguson was an iron man, starting every game in 10 of his first 11 seasons. He also had five seasons as a backup and part-time player after 1984, drawing a paycheck past age 40.

Vince Ferragamo

3.4 3.0 (20 bad)

Ferragamo was the starter (at least during the playoffs) for three Rams playoff teams in five years, including for the 1979 NFL runner-ups. The evidence suggests he could throw the ball. If you just look at his yearly interception totals, they don't really stand out, but then you do the math, and you see he was ceding a lot of interceptions, so this was his problem. Oddly, he follows Joe Ferguson in the alphabet, and he broke Ferguson's streak of seasons as the starter in Buffalo. That 1985 season in Buffalo was also the last hurrah for Ferragamo as a starter.

Doug Flutie

7.8 4.5 (7 good)

I always rooted for Flutie (unless he was playing against the Giants or something); I felt like he needed me. I felt like he was the underdog, that he was on the precipice of embarrassing himself. Who knows whether he could have kept it up, but it turns out he was a really good player with the Bills.

He was about as bad as I thought with the '88 Patriots (-4.9). He compiled a Tim Tebow-like record as a starter that year, though, going 6-3.

I'm not sure how we fill in those years between '88 and '98, and speculate what he would have done with a body of NFL work. But his ascension to superstar in the CFL, and his later very good work with the Bills, suggests that he should be taken very seriously by historians.

I thought that was a pretty strong finish above, but I see I forgot to make one point I wanted to: having immersed myself in these quarterback records, I'm convinced that the stigma against short quarterbacks is undeserved. I don't see how anybody can look at quarterback records, and quarterback successes and failures, and take the position that short quarterbacks are unlikely to succeed. If you hold that position, you are besieged by counterexamples.

Roman Gabriel

10.9 4.4 (94 good)

From 5'10" Doug Flutie, to a man whose measurables would probably impress scouts in 2012, 6'5" Roman Gabriel. Gabriel was just a guy in yards-per-drive, neither good nor bad. Despite being the 1962 2nd overall pick behind Ernie Davis, his career got off to a slow start, and he wasn't entrenched as the Rams starter until '66. But from '66 through '75, he was a starter every year, and between -2 and 2 every year except 1973 (when he was 4.4).

However, he really stood out when I shifted the analysis over to interceptions. They used to be so much more prevalent than they are today that his judiciousness can be missed. But Gabriel scored as the #1 all-time interception-saver with my methodology. Interestingly, according to pro-football-reference.com, he only led the NFL in interception percentage twice. But again according to pro-football-reference.com, he was in the top 7 in interception percentage all 11 years that I count him as a starter, and in eight of those years, he was in the top five in interception percentage. So big, powerful Roman Gabriel earned his stripes with his discipline. Good prospects often grow into good players, but look different than we thought they would.

Jeff George

9.7 4.5 (22 good)

I don't really remember him as interception prone, but I was still surprised he came in at "22 good." Character assessments creep into the discussion of interception rate, rightly or wrongly, and George was rightly regarded as an underachiever. I say rightly, because he certainly didn't play to the level of his talent. Character assessments also wade into surveying of sack rate (QBs who are sacked make mental mistakes and should unload the ball) and, of course, clutch play. George was the most sacked quarterback in the NFL in both 1991 and 1997, with two different teams, Indianapolis and Oakland. To be fair to him, those teams were not exactly the San Francisco 49ers. George was sacked over 43 times per 16 games in his career, which certainly sounds like a lot, and draws a "Sack%+" index of 91 from Pro Football Reference (it's clear that under 100 indicates an above-average-rate of sacks allowed, but whether 91 is a notable score, I can't tell you).

Otto Graham

15.9 6.4 (20 good)

(Additional 27.4 if AAFC included)

Trying to put Otto Graham's ratings into context with other quarterbacks from the same era, the following statements seem to be true.

*The Sid Luckman Bears were really in a class by themselves. Whether you consider the NFL to have been competitive enough in those years to take the ratings at face value is another question, and not one I can opine knowledgeably on.

*Graham's AAFC Browns ('46-'49) were about as potent as Van Brocklin and Waterfield's Rams,'50 -'54.

*Once moving to the NFL, the Browns performance slipped so significantly that their claim as one of the all-time great offensive teams must be constricted to the AAFC years.

*Greatness is fleeting, and Otto Graham was aging, so we cannot simply take the NFL results as the real Mccoy, and dismiss the AAFC ones as reflecting a lower level of competition. The Browns also did bounce back with seasons of 5.8 in '52 and 6.4 in '53.

*I suspect the Browns were still improving offensively in the early NFL years, and suspect their '52 and '53 units would have racked up even greater numbers in the AAFC than did the earlier Browns.

*The NFL Browns in Graham's time resemble the 49ers under Tittle, YPD-wise. The Browns only had the two seasons with memorable scores, and the same was true for the 49ers under Tittle. In fact, the two teams' short peaks overlapped, with the 49ers getting good just a year after the Browns did, and edging them 6.8 to 6.4 in '53, before posting another outstanding season in '54 (6.4). The Rams, however, were in the stratosphere in '54, at 9.4.

*I don't know if the gap in scores for Graham's NFL Browns between '52 and '53 and the other years makes sense, but it makes sense from the basic passing statistics that '52 and '53 were the best years. Graham threw 364 passes in '52, 99 more than he did in any other NFL season; if you look at offense from the point of view that runs are pretty much giving away yards, and passes adding them, throwing that often is an advantage. 1953 obviously didn't feature as many passes as 1952, but was a freakish yards-per-attempt year, at 10.6; that was over 2-yards-a pass-better than he'd done in either the the '50, '51, or '52 seasons.

*The pattern from season to season looks different for Graham than for his Cleveland successors, Milt Plum and Frank Ryan. So it might not at first be apparent, but Plum's and Ryan's overall numbers with Cleveland actually eclipse Graham's NFL numbers by a little bit.

*Cleveland's winning run in the NFL under Graham was fabulous, with a perfect six for six record of making the NFL Championship, and victories in three of the championship games. Without assessing defense directly, we can still be confident that that was the main variable separating the Browns from the Rams, who were a better offensive team, and the 49ers, who were as good. The Lions were also about an equal match for the Browns offensively in the Graham years, and beat them two championships out of three from '52-'54, but did not post impressive records in '50, '51, and '55.

Brian Griese

12.1 6.4 (7 bad)

See the Elway comment. My YPD method creates a positive picture of Griese in Denver, even though everyone gave him a thumbs down as a franchise quarterback when he left there. I said before that most quarterbacks who have a mixture of good and bad seasons rarely alternate such seasons. What you usually see is maybe two negative seasons to start off, then five positive seasons, then three negative seasons. Or some guys bloom late, and start off negative for a longer period of time. But Griese was 6.4 in 10 starts in 2000, -2.1 in 15 starts in 2001, and 4.7 in 13 starts in 2002. I wonder what the best preceding season is for a guy who was negative, and whether Griese's 2000 is in contention for that?

Inconsistency could certainly be identified as having been a problem for Griese. He is a guy who was intercepted just four times in 336 attempts in 2000, yet finished his career "7 bad" in interceptions.

Steve Grogan

11.7 5.4 (32 bad)

The fourth mostly costly quarterback on the list when it came to interceptions....Had the best YPD-differential in the AFC in 1978, when the Patriots went 11-5....If you normalized yards per completion for era, I bet Grogan would rank highly. With a strong 7.5 yards per attempt for his career, but a completion percentage of just 52.3, you know his yards per completion were good.

To me, Grogan's whole profile goes together, particularly in the years before 1980, when he was a running quarterback. You'd think an improvisatory quarterback would complete some long passes, a la Ben Roethlisberger, Tim Tebow, and Michael Vick, and you'd think a quarterback who threw long would have a lot of interceptions.

Pat Haden

3.9 3.5 (8 good)

Matt Hasselbeck

16.0 5.4 (7 good)

Few surprises in his data: has not been the same player since the 2005 Super Bowl season. That goes for his INT rate as well as his teams' YPD numbers.

Eric Hipple

4.6 4.1 (5 bad)

I should probably say nothing and not draw any attention to the fact that I ran him. Many people perhaps remember him for having a funny name and being a backup. He was run because of Detroit's 4.1 season in 1981, just Hipple's second year in the league. He started 10 games that year and led the NFL in yards-per-attempt and yards-per-completion. His yards-per-completion number was off the charts, at 16.8. The completion percentage was just 50.2.

A quick glance at the names of the receiving corps does not shed light on the sky-high yards per completion (which admittedly was quite a bit lower with the throws of Gary Danielson and Jeff Komlo added in). The one average that stands out is Billy Sims having averaged 16.1 yards a catch, a tremendous average for a running back. That was on just 28 catches, however. Yards per reception used to be higher for running backs, so if we pretend Sims averaged 10 yards a catch, we'd be pretending a normal figure. With 10 yards a catch for him, Detroit's receptions yards would have diminished 171 yards, which would have lowered their overall yards-per-reception by about 0.7. So Sims played a role in the high yards-per-reception figure, and could possibly have played even a bigger role in Hipple's individual number, but it was a modest role.

In his three other seasons as the majority starter for Detroit, Hipple's yards per completion were 12.6, 13.2, and 10.0. The first two would be highish numbers for today's NFL, but were probably average or below in the '80s, when Hipple played.

Ron Jaworski

6.3 2.5 (25 good)

A productive player from '79 to '81, shaky otherwise. On the positive side, he was at or below the league-rate in interceptions every year from '78 to '86.

Brad Johnson

10.1 4.6 (16 good)

Bert Jones

6.2 5.8 (27 good)

Since his Pro Football Reference page can be sponsored for just $55, it might be stretching it to say that Jones has become something of a legend, a Pete Reiser of football, but it is also true that people who saw him play feel it's an opportunity missed not to respond without some sort of guttural exclamation when he comes up. He's compared to John Elway.

Like the early Elway, his reputation is not completely borne out by the record book. He had a 5.8 score in 1976, but in his other surrounding years of starting before his injury problems, was right around average.

If he wasn't clearly on his way to a 30-point+ career, at least it can be said that his record does have some special elements. He was a 60% passer in 1976, and in the top in the top five in the NFL in completion percentage in 1975 and 1977, too. Interception-wise, he was one of the better quarterbacks we have seen, ranking #1 in interception percentage in '75, and #2 in 1976 and 1977. Per pass, he saved almost as many interceptions over his career as Aaron Rodgers, and quite a few more than Dan Marino. He would have been a perfect guy to have cited in the Favre comment as a player whose excellent interception rate gave a clue of his superior talent.

Only one Pro Bowl during his glory years does give some pause.

Sonny Jurgensen

16.5 5.2 (45 good)

It's always hard to know what expectations are for individual players, to know what the consensus rankings are, but Jurgensen has to qualify as a YPD disappointment. Forgetting the laudatory things people say about him as a pure passer, just in terms of statistics alone, he was a starter for just ten years (nine if you exclude a nine-game season), and he led the NFL in passing yards five times. The system does have those as five of his six positive seasons, although they score as low as 0.3 and 1.0. Certainly, since yards is the measuring stick for both yards passing and yards per drive, it's particularly surprising Jurgensen doesn't do better in the latter. But any number of factors could explain this, such as a high number of passing attempts, weaker performance getting first downs than big plays, a high number of drives, or a bad running game. You might think the latter is the biggest missing factor with YPD in general, but I don't think you are right. Certainly, the more running explains the discrepancy in a given case, the less culpable the quarterback is for the failure in yards per drive.

Jurgensen certainly had an interesting career, sitting behind Norm Van Brocklin for multiple years at the beginning of it, and behind Billy Kilmer for multiple years at the end of it. His two best YPD scores came in 1961, his first year as a starter, and 1970, his last (in case you were wondering if that was the Lombardi/Jurgensen season, it wasn't -- that was the one before). 1961 was the only year his team had any real success, as the Eagles missed out on the East championship by a half game behind the Giants. He threw a league-leading 32 touchdowns that year.

Jurgensen ranks 11th of the quarterbacks I sampled in interceptions saved, with 45. Even though other methods and subjective opinion would put Jurgensen higher than my YPD/interception two-pronged approach, 16.5/(45 good) could conceivably fit into the Hall of Fame, particularly when those numbers apply to a quarterback of a number of years ago.

Jack Kemp

10.7 3.2 (2 bad)

8.6 of total early AFL

Kemp was a mainstay of the AFL Championship game, quarterbacking two of the first seven winning teams, and three more of the first seven losers. He started for eight years, making seven Pro Bowls. He is not in the Hall of Fame, however, and based on my data, I think the voters get that right. It is somewhat reassuring to see that, using a league standard, he did at least direct a good offense in most of those championship-contesting years. Good does not equal great, however, and Kemp is evidence that smart guys can throw interceptions.

Billy Kilmer

6.7 3.7 (12 good)

Perhaps best known for being the Redskins' Super Bowl quarterback vs. the 17-0 Dolphins, the Redskins were 44-19-1 in his starts from 1971 thru 1976, despite YPD numbers overall that were average or a little below. Kilmer started at least 10 games every year from 1968 to 1975, but only the season's full slate of 14 in 1969. That came with the Saints, and was also his best year, YPD-wise.

Tommy Kramer

7.7 3.7 (1 bad)

Dave Krieg

2.7 1.7 (3 good)

I'm not sure I'm trying to convince the "stats are for losers" (Raheem Morris and Ross Tucker, you are acknowledged here) crowd of the relevance my work, but their likely intuitive sense of the balance of quarterback play will be reinforced by my low rating of Krieg. Surprisingly reinforced, too, since Krieg was often seen as someone who was missing intangibles but put up numbers: he threw for 7.2 yards per attempt over his career, on enough throws to total 38,000+ yards, and is 12th on the all-time list in touchdown passes. Add to that the fact that the Seahawks, in particular, had their fair share of success under his leadership, and it seemed unlikely that even a nuanced rating system would yield a negative view.

That is exactly what happened here, however. How a player can start for 12 years and not have positive seasons even summing to 3 is beyond me.

Perhaps the list of aspects that can separate conventional passing stats from yards-per-drive listed in the Jurgensen comment are worth contemplating here. Whatever the case, the Krieg teams' yards-per-drive seasons do seem to show that they were missing something, and were not a good offensive team.

To be fair, that Krieg might possibly come in last in YPD points-per-season-started is a real fluke: he does not out-and-out have the worst yards-per-drive data, even among guys who played for a while. He does not rate with Charlie Conerly, for instance, in my mind. Krieg's worst seasons were only -3 to -4, and two of those came when he was past his prime. If I used a replacement-level number as the base, instead of 0, Krieg would do better. The super-low score reflects that he never had a good season, although it's interesting that he's not devoid of positive seasons (which would yield a 0.0 score), nor is anyone else.

And it should be added that there may be quarterbacks who started several seasons and were always in negative territory. I was aiming to find good quarterbacks, so I would not have tested them.

Also, to be fair, John Elway's data for the first 10 years of his career (also beginning in 1983) is Krieg's, just somewhat better. Elway was almost on the ground floor at 5.2 at the end of his 10-year partnership with Reeves. So maybe those old Ignatin and Barra articles about whether Krieg was better than Elway weren't that far off.

Greg Landry

8.8 5.3 (17 good)

I started following football as a little kid in 1982. The last year Landry saw any significant play was in 1981. Even among players recently retired, conversation persisted about most of the ones of any importance persisted; knowledge of them informed the current ones. But no one was talking about Greg Landry (Greg Landry the player, anyway) in the '80s, that I can recall.

Perhaps that's because he was pretty much washed up by the mid-70s. Just looking at his modern bubble-gum card (to label the current Encyclopedia entries), he seemed to have a lot going for him in his early NFL years. He was 6'4"; a first-round draft pick, despite being from Victor Cruz's school; and the owner of simply gaudy yards-per-attempt rushing statistics. From most indications, he could pass, too: his completion percentage numbers aren't eyesores; he passed for 8.6 yards an attempt in '71; and his career interception-differential can safely be called good.

Anyway, Landry was drafted by the Lions in '68, started seven games for them in '69, then scored 5.3 for them in '71, and 3.2 in '72. But his '73, '74, and '75 seasons were partial each year, with between three and seven starts. So presumably, he was injured, and never again was the player of '71 and '72. His last two seasons as a starter, in '77 with the Lions, and '79 with the Colts, are almost the exact inverse of his '71 and '72 seasons in yards per drive.

Mike Livingston

2.5 1.6 (10 good)

Neil Lomax

9.7 4.5 (39 good)

Digging into his record has given me a new-found respect for him. As Dan Marino's record for passing yards was first endangered and then broken this season, we heard a lot about how he achieved his totals under less favorable conditions for the passing game than exist today. The irony of that is that 1984 was considered a "year of the quarterback" in its own time. Marino's achievements were part of a league-wide trend (or at least that's how I remember things.) Perhaps 1984 was to offenses in football what 1987 was to offenses in baseball. Montana was the second-biggest quarterback star after Marino. Lomax might have been the third guy: a player who put fear into opposing fans and seemed to have limitless potential. Marino's and Montana's brilliance was obvious, but Lomax was the next guy who was going to join their ranks. It never happened, and so we think of Lomax as a disapointment.

The fall-off in his play was magnified by the Cardinals going backwards in the standings. In Lomax's four remaining seasons after his big '84, they had a losing record each season (although they were .500 with him as a starter in '87 and '88).

Lomax actually righted the ship considerably in those last two years. He led the NFL in passing yards in '87, and was 5th in '88. His YPD scores were 2.9 and 2.3. But anything less than getting back to where he had been in '84 was less than completely satisfying. Because Lomax fell off at least some, it seemed he was responsible for the team's decline. The question that wasn't being asked was whether anyone else would have been good enough to win with the team he had.

Lomax took a lot of sacks: more than 1 per 10 passing plays over his career, including 61 in 1985. Conversely, he threw very few interceptions (he saved about 5 a year for 39 for his career). It's important to realize that sacks are a part of YPD. There aren't additional costs in a sack that don't come out in YPD. So, if you have a player with good YPDs and excellent interception rates (Lomax was no worse than 6th in the NFL in INT percentage in his seven seasons as a starter), that is very definitely a good player.

With his hip injury, Lomax never made it into a NFL game past age 29. Still, it is doubtful he could have reached a career total of 20 if this hadn't have happened. Even his '84-'88 peak contained its share of struggles. Heading into the '89 season, he could have been looked at as either the 2-3 YPD player of '87/'88, or essentially a net 0 player who had ups and downs. His interception value could not have been questioned. But perhaps the truth of Lomax was somewhere between 1C to Marino and Montana's 1A and 1B, and the cowardly mutt exasperated fans wanted out of town.

Johnny Lujack

12.1 5.6 (7 good)

A Heisman Trophy winner from Notre Dame, appears to have been one of those athletes everyone should know, and apparently did know, in his day. He also succeeded Sid Luckman as the Chicago quarterback and was immediately one of the best players in the league at the position (although I suppose one could credit the foundation that had been set in Chicago, as much as Lujack). The fact that he was an All-Pro quarterback in his second year as starter, and a Pro Bowler in his third and final year, however, suggests his play generated the respect his statistics say that it should have.

Apparently, the TD/INT ratio was not the bottom line it is today for most pundits, because Lujack was neither a Pro Bowler or an All-Pro in 1949 when he had 23 TDs passing and 22 interceptions, but was both in 1950 when his ratio looks abysmal at 4 to 21. My first reaction seeing the 1950 numbers was that they were fishy -- that the Bears were probably playing in a way that did not lend itself to TD passes. This idea received some support when I noted that Lujack led the NFL in rushing TDs in 1950 with 11 -- in 1949, he only had 2.

Don't think that the Bears curtailed their passing in general to an extreme extent in 1950: Lujack threw 312 passes in 1949, and 254 in 1950. In other words, it appears the Bears didn't stop throwing, they just stopped throwing on the goal line.

Their overall performance was down some in 1950, too, from 42 touchdowns running and passing and 332 points in 1949, to 30 touchdowns running and passing and 279 points in 1950. I'm happy to note that YPD has the Bears with a raw YPD of 27.9 in 1949, and 23.9 in 1950. Lujack's touchdowns passing are probably of very little significance in gauging the quality of his play year to year, however. And touchdown passes are really the issue with the ratio, since his interceptions actually declined from 1949 to 1950 (if they did increase a bit on a per-pass basis).

My grasp of early NFL history is limited, but Lujack seemed to fit with players whose contribution to football is best understood by taking stock of both their college and pro careers. I'm sure football was a tempting way to earn a living for most of the great athletes, but there wasn't any kind of a sense about legacy at all, particularly since pro football didn't count for much. There wasn't the feeling that if you left the pro game after a few years, you had in some ways been incomplete.

Maybe Lujack would have fought harder to come back after 1951 if there had been a bigger picture, but the fact appears to be that he retired because of a knee injury (source: polishsportshof.com). We should not lose sight of the fact that the expectation for long careers was also absent because the game was more brutal than it is today, and so premature and unavoidable ends occurred more in the natural scheme of things.

Archie Manning

6.4 4.3 (13 good)

While quarterback statistics can be deceiving if we judge them by standards of a different era, anyone who has even eyeballed Archie Manning's has probably had his faith in the famous patriarch shaken. Commentators are usually effusive in their praise of his efforts, but we do see the stray graphic from time to time showing that Archie was the worst ever in this or that, which the commentator sells to us a shocker. The graphic is always presented with the theme of how horrible Archie's teams were. No one ever thinks (or dares) to say that maybe Archie wasn't very good himself, no matter his raw talent.

Teasing apart Archie from his teammates would be an ambitious project, and not something I have any desire to do here, as it doesn't fit with my mission. I will say that I found it interesting that Archie was effective for a short time, in 1978 and 1979, which also happened to be his two Pro Bowl seasons.

Through all of his career futility, most of which came with New Orleans (as most know), he did also maintain a positive interception differential.

Eli Manning

17.2 5.1 (11 bad)

As I write this, in the wake of the Giants' fresh Super Bowl championship, we are in Eli Manning silly season. Having run his numbers weeks ago, I thought I was going to be primarily pointing out their virtues when I went to the blog, and coming to terms with Eli as a good statistical quarterback, which was rather a fresh idea for me.

But now the debate is whether Eli is better than Peyton. And not just whether he is elite, but how elite is he? Well, four Super Bowl championships didn't stop me from leveling about Terry Bradshaw's record (or perhaps better put, being the iconoclast), and two Super Bowl championships won't stop me from doing the same about Eli's. Peyton sits at 80.8 in YPD differential, yes in 13 years, but 80.8; Eli sits at 17.2. Eli's not going to catch him. Given where he sits, and the way the modern game is going, if Eli is in the top 15 all-time when he retires, I'll be surprised.

He's ninth currently among active quarterbacks, although only P. Manning, Brees, Brady, and Rivers are well ahead of him; the rest in front of Eli are very close to him and very bunched. We can throw Matt Ryan, at 17.0, in that group as well. More important than Eli's total rank is that, of the big 10, only Mark Brunell has a lower average per season started; in fact, Eli's 2.5 per season is well underneath the other eight guys, none of whom is lower than 3.6 a season (Romo).

I don't really consider all of this a knock. More a reality check. Eli's record in YPD is good -- very good.

In his first few years, his passing stats were mediocre, but his YPD was better than that. It's been positive all seven years he's been a starter. The Giants running game helped, no doubt, but that started to even out this year. And even if it didn't, Eli is supposed to be a smart quarterback who sometimes adjusts the plays to get them to the right hole. Eli's total lags a bit because, by and large, he doesn't have big seasons: his high is 5.1 (2008), and his second best year, 3.3 (2009).

This year, he was 2.3, with the Giants coming in 9th in the NFL in yards-per-drive, and 6th in the NFC. Since they were a 9-7 regular-season team, this should not come as a great surprise.

The more troubling challenge for Eli-silly-season lies in his interception record, which is "11 bad" -- really unusual and really incongruous for a very good quarterback, as we've seen. I don't think we should lose sight of this negative interception rate for very long. There are, perhaps, however, some factors that plead in Eli's favor.

1. The championship-season postseasons, where he has stayed away brilliantly from interceptions, throwing just one in both years. Using a 3% expected rate, he's saved a total of 7.5 interceptions in those two postseasons. For his other three postseasons, however, he's basically given back 5 interceptions, not helped by a 3-INT-in-18-passes start vs. Carolina in 2005. This ties into point #2.

2. The INT rate is getting better. Of his three best INT-rate seasons, they've all come in the last four years. It's not an impressive INT rate, however -- the league average in 2011 was 2.9% interceptions, and he threw 2.7%. For his 589 throws in 2011, he threw 1 less interception than a league-average passer would have.

Tony Romo is somewhat similar to Manning in terms of his career-YPD-differential, his YPD-differential-per-season, and in terms of not being a great interception-avoiding quarterback, but an improving one. Yet Romo's interception rate in the time of Eli's improvement is markedly better than Eli's: it's 2.3% over the last four years, compared to 3.1% for Eli.

3. In ranking the factors which make Eli a mediocre quarterback vis-a-vis interceptions (and this is wholly subjective), I would rate them this way.

1. Accuracy issues.

2. Occasional foolhardiness. Forcing the ball.

3. Kevin Gilbride's multiple-route offense, which sometimes leaves the ball going where the receivers aren't. There are interceptions that happen which are not Eli's fault, but the receivers' fault, or the fault of the complexity of the offense.

One thing I find interesting: Mark Brunell has been an outstanding interception-avoidant quarterback. But he has one year which sticks out from the others, one year when he threw 20 interceptions, way above his top of 14 in any other season. Brunell's 20-interception season came when he was in just his second year as a starter, but also in one of the only two years when Gilbride was his offensive coordinator.

Jim McMahon

6.8 2.9 (13 good)

As you can see from his career and single-season-best YPD differential, McMahon does not belong in the company of most of the quarterbacks here. Perhaps some people expected him to show a 15 YPD for his career or some such thing, but probably not many; if you're like me, you remember the Chicago offenses of the '80s as grounded, pedestrian units. In fact, though, in the four years from '83 to '88 when McMahon threw the majority of passes for them, their YPD diffential was in the positive zone each year. It never got above 2.9, though, which occurred in the powerhouse season of 1985.

When I say the Chicago units were grounded, the description gives the right idea: subtracting passes thrown from runs, no team in the league was higher in 1984 or 1985 than the Bears, and the '83 and '88 Bears (the other decent McMahon-led offenses) were 5th in the NFL in this makeshift category.

There are going to be several quarterbacks in the remaining profiles whose YPDs are discussed in the context of their coaches' conservatism. I think I have a pretty good handle on the effect, but YPD statistics of individual quarterbacks bring a focus to it, and it's interesting that the McMahon Chicago offenses were good, despite being extremely run-oriented. When I think of just how conservative they were, running 284 times more than they passed in 1984, I think of Mike Ditka's simplistic, tough-guy approach and roll my eyes, but obviously there were some good reasons for Chicago to be so conservative: they were often ahead; they ran the ball well with Walter Payton; and they had a great defense. McMahon's 57.8 completion percentage and 7.4 yards per attempt in his career with Chicago do make the positive YPD-diffentials consistent with his passing record.

Donovan McNabb

16.1 4.8 (50 good)

I thought McNabb was a quarterback who might do better in yards-per-drive than yards-per-pass, because Andy Reid has been said to be a coach who uses short passes in place of runs. That has the potential to hurt yards-per-pass more than yards-per-drive. McNabb made four straight Pro Bowls from 2000-2003, never going over 6.7 yards-per-pass. Seven yards-per-pass is generally average. In terms of yards-per-drive for those seasons, McNabb was more or less average. So he does show better in terms of yards-per-drive than yards-per-pass, but not by a whole lot.

It is also true that he actually has only thrown 500 passes in a season twice, and only been in the top five in the NFL in passes thrown twice. Injuries to McNabb, his number of rushes and sacks, and an inability to keep the ball through good drives all probably play into his seasonal attempts not being as high as I would have guessed.

In 2004, McNabb broke out with 8.3 yards-an-attempt. He threw 31 touchdown passes, far more than he has in any other season, and the Eagles made the Super Bowl. But his YPD-differential was disappointing at only 2.6.

2006 and 2007 are his best YPD-differential seasons, both in excess of 4. The 2007 rating is a head scratcher, with McNabb's statistics appearing unremarkable across the board.

McNabb is tied for 8th in total interceptions-saved with 50.

Steve McNair

10.7 4.8 (26 good)

McNair was a respected player, so perhaps his career total in YPD, which has him tied for 73rd among the quaterbacks listed here, will be a disappointment to some. He would probably rank a little higher in his number of positive seasons, which is six, and all of those seasons were consecutive, which counts for something in my mind. He was good year in and year out for a while. A player who got very beat up, he only started for nine seasons, and that hurts his total, too.

His profile looks a little bit like Terry Bradshaw's: Bradshaw scored 11.0 for his career, with all six of his positive seasons consecutive, and the postive streak nestled in between a number of seasons that he needed to figure things out, and an apparent decline phase. Like Bradshaw, McNair finished off with two negative seasons, although he only required one negative season before becoming a winning player (by which I mean a positive-YPD player). McNair did sit on the bench for most of two seasons before starting, while Bradshaw was thrown right into the lineup -- maybe McNair would have been negative for more seasons at the start of his career if he'd been forced to play. A discrepancy between the two, and the factor that makes McNair a superior player in my mind, is that McNair was a good interception quarterback, while Bradshaw was a pretty bad one: McNair is 37 interceptions ahead of Bradshaw in interceptions saved.

McNair's career total would have been better if his YPD in his MVP 2003 had been MVP-like, but it was just 1.7. As his statistics were excellent in 2003, with a league-leading quarterback rating, the YPD standing is surprising. I don't think we can dismiss what YPD is telling us, but the Titans' 3.3 yards a rush certainly made the job of moving the ball more difficult, and certainly wasn't principally McNair's fault.

McNair's career high, and indeed his only positive score that stands out at all, was his 4.8 season in 1998. Why this season was the Titans' best in the McNair era is likewise perplexing: the team was 8-8, and McNair registered just 6.6 yards-a-pass. McNair's summary statistics simply do not support this as his best year, but he did have his highest yards-per-rush in any season at 7.3, his most rushing yards next to 1997, while throwing the ball more successfully than he did in 1997.

Don Meredith

13.0 5.3 (14 good)

See the Staubach comment.

Scott Mitchell

10.9 4.5 (6 bad)

Starting for the Lions for four seasons during the peak of Barry Sanders' career, Mitchell was actually competitive with Brett Favre during this time. Mitchell's career 10.9 YPD-differential is all in these seasons (he was not a majority starter in any other seasons), while Favre's total from '94-'97 was only slightly higher at 11.8. While you see many tricks of this type (taking a big-time player in his down years, or in a weak area, to prop up another guy), '94-'97 was not only Barry Sanders' peak, but allegedly Favre's as well: he won NFL MVPs from '95-'97. Of the 16 four-year Brett Favre runs that can be used, '94-'97 is his third-highest total. The only quarterbacks I can find who scored higher than Mitchell from '94-'97 were Young (17.9), Marino (16.7), Elway (16.3), and Favre.

Mitchell's high came in 1995, and his statistics were good enough for that year that it's not terribly surprising he scored as high as 4.5: he was among the NFL's top five in any number of categories. Mitchell's yards-per-pass-attempt were under 7 in each of his other three seasons, and way down low at 5.9 in 1994. Although Dave Krieg has one of the least impressive YPD careers out there, Krieg started seven games in '94, and unquestionably greatly outperformed Mitchell: his role in getting 1994 to 1.7 for Mitchell cannot be overlooked -- as Barry Sanders's also cannot be.

It might sound crazy to say, but I just don't see any evidence that Jim Brown added yards per drive to the Browns. The Browns' YPD totals neither increased when he came to the team, nor fell off when he left, and seem to correspond with what their passing games did statistically when he was there. One figures that if Jim Brown was only adding negligibly to YPD, no running back has made a substantial positive difference on the measure. But at least in the '94-'97 period, the Detroit offensive production doesn't add up from the passing game, and the temptation is to say that this was Sanders' doing. Certainly, even open-minded analysts are going to look at the strong numbers compiled by Mitchell with an asterisk because of Sanders.

You might note that Mitchell came out with a negative interception rating. His interception rate jumped all around in his four seasons with Detroit. As eager as fans and commentators were to pin whatever they could on Mitchell, who was highly paid and never gained their trust, his interception rate is sneaky bad, as he never threw more than 17 interceptions in a season. I wonder if we again have a case here where interception rate corresponds with the eyeball and gut test more than just about any other statistic?

Earl Morrall

12.0 6.3 (5 good)

Was a NFL quarterback for 21 years but only a starter for five, and with five different teams. The last two were the great '68 Colts and '72 Dolphins, where he was a combined 22-1 as a starter during the regular season. While these were two of the greatest single-season teams ever, they were not necessarily two of the greatest offenses, although the 4.5 and 6.3 YPD differentials (Colts and Dolphins respectively) were nothing to sneeze at. Morrall averaged over 9 yards a throw with both teams. Going by the statistics, seeing him as a caretaker in '72 wouldn't necessarily be wrong, but in 1968, he was NFL MVP, led in touchdown passes, and came close to leading in yards passing. He was so impressive that people probably forgot he was a journeyman, and that the '68 Colts might not be all that they were cracked up to be come Super Bowl time.

One oddity in Morrall's career is that he made the Pro Bowl with Pittsburgh in 1957. This was just his second year in the league, the team was 6-6, and the YPD differential, -4.1. One wonders how many Pro Bowl quarterbacks have been worse. He did lead the league in terms of having the lowest interception percentage, which, as you can tell from his "5 good" rating, wasn't necessarily a trend for his career.

Craig Morton

Overall, not a quarterback who shows particularly well from a YPD perspective, although his years with Dallas were an exception (see the Staubach comment). Morton's -2.8 score with the AFC champion, 1977 Denver Broncos is particularly interesting: although he was negative in seven of his eight seasons as a starter after leaving the Cowboys, his score in 1977 was actually his worst. It's not downright dreadful, so you can see how he plodded along, starting not just for five years with the Broncos, but for three with the Giants as well.

Until very recently, perhaps until the undertaking of this project, I would confuse Morrall and Morton. Not only are they close alphabetically, but both were starting quarterbacks for the Giants. Both played a role in Super Bowl history; in fact, acting as their teams' primary quarterbacks in the 1970 Super Bowl faceoff between the Colts and Cowboys. And lastly, both Morrall and Morton played for a very long time.

Joe Namath

17.6 4.8 (17 bad)

There's been a lot of attention paid to Namath's statistics recently, so my findings will probably be more a source of interest than disappointment. I'm not sure that everyone who has reviewed his statistics has grappled with era norms, however, instead reacting more like Ugly Americans overseas. Then, too, from listening to the discussion of the HBO documentary "Namath" on the Slate podcast "Hang Up and Listen," the belief expressed in the documentary was that his statistical record contains good and bad elements. So perhaps a measure that balances them out like YPD is particularly necessary.

In any event, it's safe to say that Namath did run one of the better offenses in football from '66 to '69, and also when he returned as a full-time starter in '72. His scores in those seasons were all comfortably positive, ranging from 2.6 to 4.8. But once he hit his 30s, Namath's total score was just 0.5.

His interception ledger for his career ("17 bad": 9th worst among the 117 quarterbacks here) seems to remove any claim he has to having been a very good quarterback. But actually, while never a strong quarterback interception-wise, Namath was "12 good" through 1969. He had thrown over half of his career passes at that point, but would be so bad in his remaining seasons, that he would get to "17 bad."

I can't present it quite in the vein that the good Namath, who could move a team, was an interception avoider, and the ineffective Namath was ineffective on all fronts. In 1972, Namath led the NFL in yards, touchdowns, and yards-per-attempt, and his YPD diffential was 4.3. This is his last very good year, YPD-wise, but also the real start of his INT problem, as he threw 4 more than expected. But for the most part, Namath is akin to the batter who was hitting .300 and 30 homeruns, then went down to .250 and 15 a season. His peak is relatively blight-free.

We've seen a number of quarterbacks like Namath in recent years, who became interception-prone late in their careers, sometimes very late in their careers. From the outside, one might think that interceptions would mostly have to do with smarts, and would not rise late in a quarterback's career. But whether because quarterbacks don't properly account for their loss of arm strength, or because their hand-eye coordination is just not what it used to be, interceptions are a problem of older quarterbacks. Or at least anecdotally, it seems to be this way. If interception rates do indeed follow general decline, this is one more argument for my developing opinion that interceptions are the best indicator of quarterback quality.

Namath's career score was 17.6; Eli Manning's right now is 17.2. Eli has already compiled more consecutive positive seasons that Namath ever did, but Namath's scores year to year in his good period look a lot like Eli's, although they were a smidge better. Namath scored 17.1 in his five-year run, while Manning's best, and last five years, total 15.0. For whatever reason, although the YPD numbers are calculated against the league average, it's clear the best quarterbacks compile higher numbers today than they did in Namath's day, so maybe the comparison of Namath to Manning goes clearly in Namath's favor. The two quarterbacks are also a good comparison because both had problems with interceptions.

Manning never misses a game and has never played better, so he is not a likely candidate to experience a sudden decline. He did, in fact, compile his 2011 season at age 30, and Namath's last good season came at 29. In the space of a couple of years, Manning is likely to rate above Namath, and to make the comparison of the two particularly maladroit.

Bill Nelsen

8.4 3.9 (0 good)

Briefly mentioned in the Frank Ryan comment.

Ken O'Brien

6.7 3.2 (42 good)

At least some similarities with Neil Lomax: traded possible interceptions for sacks, had his best year early on, logged seven years as a starter, but played fewer total years than the average guy who started for seven years. Both players also went to schools (UC Davis for O'Brien, Portland State for Lomax) where they faced less than elite competition. As for differences, Lomax's rushing statistics, while far from senational, suggest that lack of mobility was not the chief cause for all of the sacks he absorbed. Lomax was probably just a better player overall, too, although one could argue about the strength of supporting casts.

Allen Barra was ahead of his time with his Football by the Numbers book in the '80s as well as his Village Voice articles of that time. He argued that yards per pass and interceptions thrown were the key passing statistics, and was straightforward and brave in asserting offense to be almost the sum total of the passing game. He distilled many other aspects of football to their essence, and propounded positions that current analysts like Aaron Schatz have embraced.

For being at least 90% right, then, it's surprising to me how many times, looking back, I think Barra got his quarterback evaluations wrong, specifically when he was championing unheralded players vs. stars. One case he made, at least in '85 and '86, was that O'Brien was better than Marino. I would say that it was always a shortsighted position, and one that certainly looks particularly ridiculous now, with Marino totaling 64.3 for his career-YPD-diffential, and O'Brien 6.7.

Carson Palmer

12.4 4.6 (6 bad)

Babe Parilli

3.9 2.7 (14 bad)

Early AFL: 2.7 of total

What I knew about Babe Parilli before doing this work was that he was the backup quarterback for the Super Bowl Jets, and that Joe Namath had outrageously said he was a better player than the Colts quarterback, Earl Morrall. Although Morrall was the 1968 MVP and entrusted with the Super Bowl instead of Johnny Unitas, this wasn't as outrageous a statement as I thought. My numbers actually suggest that Parilli wasn't any good, but he did start 75 of the Patriots' 84 games from '62 - '67 and was their majority starter every year during that time. Morrall was never a majority starter for consecutive years, even though he played for 21 seasons. From one angle, Morrall could be considered the more inveterate journeyman. If you had asked football fans which of the two they would take first after 1967, I bet many, and maybe even a majority, would have said Parilli. If they didn't say that, it was probably because they were NFL snobs, accepting uncritically that most everyone in the NFL was better than most everyone in the AFL. And as I understand Namath's bluster, that, fundamentally, was the attitude he was attacking.

It is true that Parilli received his opportunity to start because of the AFL coming into existence. He was Green Bay's majority starter in 1952, and then not a majority starter again until 1962. The gap in time between starting jobs certainly makes for an odd profile.

Chad Pennington

10.6 3.6 (13 good)

Pennington is the all-time NFL leader in completion percentage. And whatever his lack of arm strength was, he did effectively move his team down field, showing positive in five of his six seasons as starter, and just -0.9 in the sixth, when I would bet Kellen Clemens starting half the schedule was to blame for the anomaly. Pennington was solid if not special avoiding interceptions. While his resume is not empty of anything substantive, I do think a history of the significant quarterbacks of his era could omit him without committing an injustice.

Milt Plum

16.3 6.2 (16 good)

See the Frank Ryan and Otto Graham comments....I can't remember exactly why Ernie Accorsi was so dismissive of Milt Plum's posting one of the best single-season quarterback ratings ever in 1960, but part of his opinion was probably that Plum was a "system" quarterback. If you accept that Otto Graham and Frank Ryan were very good players, this argument is weakened by Plum having had similar YPD differentials with Cleveland to the ones they had (not that the teams around the quarterbacks were identical; Graham, for instance, never played with Jim Brown).

It is true that Plum didn't distinguish himself at all after leaving Cleveland.
He played for seven years, two more than he played with Cleveland, starting the majority of his team's games three times, and doing no better than 0.3 in YPD differential in those three seasons. He struggled mightily with the Lions in '65, coming in at -6.3.

How does the 1960 season come out in YPD differential? It's actually not Plum's career best, which was the year before, 1959. In fact, Plum 1959 effectively whips Plum 1960, 6.2 to 3.7. This is a hard one to figure, and as big a mystery in its own way as Sammy Baugh's 13.7 1945. 1960 topped 1959 on each of the passer rating categories, decidedly enough that the overall edge in passer rating was 110.4 to 87.2. Plum threw nearly all of Cleveland's passes in both seasons, so it's not the case that other quarterbacks changed the overall YPD. The raw YPD, as opposed to against the league average, is even more in favor of 1959 than 1960. Cleveland threw the ball more in 1960 than 1959, which should have helped 1960's YPD. Cleveland ran for more yards per carry in 1960, which also should have helped YPD. I honestly can't come up with a single, solitary thing that points to 1959 as the better offensive season, but this is the kind of example that makes compiling the YPD numbers worthwhile: if they could always be deduced from conventional quarterbacks statistics, we wouldn't need them, and could combine the other statistics instead.

Jake Plummer

13.2 5.0 (22 bad)

See the Elway comment.

Aaron Rodgers

19.3 8.7 (25 good)

Confining my focus to YPD, Rodgers is off to a good enough start that his ultimate body of work and ranking might please the biggest Packers' fan. However, he does not currently rank among the game's greatest-ever quarterbacks. There is more needed here than simply lengthening the career, or adding quality seasons. Rodgers has not had a run like Kurt Warner did at the start of his career, when Warner scored 8.1, 10.5, and 10.1 right off the bat, or like Daunte Culpepper had in his first five seasons, either. The Packers were 5th in the NFC in YPD in 2008, Rodgers' first year as a starter, then 4th in 2009, 3rd in 2010, and 2nd in 2011. That's an excellent start, but not historic.

The reason I say that Rodgers is still in great position is that, while he sat behind Favre for three years, he didn't turn 22 until December of his rookie year, so he is still young. And speaking of Young, the quarterback with that surname, first name of Steve, scored all 52.8 of his points after the age of 30. Warren Moon compiled 36.6 points from 1989 on, and he turned 33 during the 1989 season. Roger Staubach had a 35.5 career, not starting until age 29. And looking at his last season's performance of 4.6 in 16 starts, he hung 'em up before he had to.

Among active quarterbacks, Peyton Manning, Brees, and Brady have already had all-time great careers: Brady is 3rd among them, and still 7th all-time in YPD-differential. Seeing Rodgers as the heir apparent to these guys is reasonable, but ignores that he does have competition. Rodgers' numbers and age stand out, but so do those of Philip Rivers, who is 9 points ahead of him in career YPD differential. And Rivers has even outperformed Rodgers since Rogers became a starter, totaling 24.9 from 2008-11.

I'll admit, I think we've gotten a little ahead of ourselves with Rodgers. Until you do it, it can't be taken for granted that you will. Granted, coverage of the NFL in past eras fell prey to less hype than it does today, but I think you could look back at the highest rated and most acclaimed quarterbacks of some seasons and be surprised at who they were. I don't think the advance story of history that we're writing now is going to be the story that ultimately gets told.

It's easy to say that with Marino and Elway and Favre we always knew....But did we? I really wasn't sure how Favre was going to turn out until 1995. I saw him as a promising but inconsistent quarterback. Just look at how dramatically Rodgers' image has changed since he became the Green Bay starter; indeed, even since the end of the 2010 regular season, when somehow, Michael Vick and Matt Ryan were voted to the Pro Bowl over him.

All of this said, Rodgers is far from just the product of a media craze. Even in the worst-case scenario, he will live on, and as much more than a Super Bowl-winning quarterback, if not as an all-time great. His very low interception rate bolsters the idea that he is a special player. And he currently has both the all-time highest quarterback rating, and the highest rating for a single season.

Ben Roethlisberger

13.1 6.1 (3 good)

Roethlisberger's never being notably above-average in his first five years came as a complete surprise to me. After all, not only did his team win two Super Bowls during that time, but he was a standout in yards-per-attempt, a statistic that unquestionably does have a strong correlation with yards-per-drive. Roethlisberger led the NFL in yards-per-attempt in 2005, was 2nd in 2004, and 3rd in 2007. He has continued to do well in YPA over the past three seasons, and broke through with a 4.1 YPD in 2009, and a 6.1 YPD in 2011, good for 3rd in the conference behind Brady and Rivers.

There are three facts about Roethlisberger and/or the Steelers that logically depress their yards per drive, although they don't seem to have the cumulative importance to fully explain the YPA/YPD discrepancy. The most important point is that Roethlisberger is among the sack and sack-percentage leaders as reliably as he is among the YPA leaders. In all five years that he's started 14 or more games, and also all five years that he's thrown at least 400 passes, he's been in the top 5 in the league in sacks. But even in net yards-per-attempt, which subtracts sack yards from yards passing, and adds sacks to attempts, Roethlisberger rates 12th all-time. Moreover, he has three top-5 seasonal rankings in net yards per attempt (2004, 2005, and 2007), but has yards-per-drive scores of 0.5, 0.2, and 0.1 in those yards, making the connection tenuous.

Those do happen to be the years the Steelers played ultra-conservatively, as defined by rushes - passes, and that's the second characteristic of the Steelers that separates them from a typical high-YPA offense. But if Bill Cowher and a running emphasis are not compatible with high YPD, how then did the '97 and '01 Steelers lead or co-lead the AFC in yards per drive under Kordell Stewart? Those teams were 5th-to-last and 4th-to-last in the NFL in passing attempts. There are ample other examples of teams that favored runs and did very well in yards-per-drive, including the Bob Griese Dolphins. I modeled the effect of play-type percentage on yards per drive and found it to be moderate; it generally does not serve as a "Eureka!" factor in considering statistical components and yards per drive.

Pittsburgh's YPD in the Roethlisberger years may also have suffered from his occasional missing of games. The four years where the YPA performance greatly surpassed the YPD performance were 2004, 2005, 2007, and 2010. Three of those years involved significant Roethlisberger absence (his three years of three or four game sit-outs). Pittsburgh's backups weren't Roethlisberger, but did about what you would expect in YPA cumulatively over the three years, averaging 6.7 YPA. So there's nothing extraordinary at work in the games missed, although they probably do tend to choke the seasonal YPDs a bit.

Add a poor interception record for a leading quarterback to the unsatisfying YPD record and Roethlisberger's standing slips all the more. He has been under the league interception rate in four of his last five seasons, however, including a stellar 1.3 percent in 2010.

Certainly, if what I said about Aaron Rodgers still having the time to thoroughly affirm his greatness in YPD doesn't quite apply to Roethlisberger, who after all has the scars of having been sacked almost twice as many times as Rodgers, the points about his relatively young age, and the streaks some older quarterbacks have gotten on still apply. And criticism also rings hollow when the player's last season was as good as any Fran Tarkenton ever had, and better than any Roger Staubach or Bart Starr ever had.

But my point is that I don't know Roethlisberger without his statistics. I always assumed they put him among the NFL's elite, at least when we weren't operating in the inanity of total yards and total touchdowns. He's such an unusual player to watch him, that it's also hard to take on faith that he's really outstanding. He's a variant of the running quarterbacks: a high-sack, high yards-per-completion guy, who also looks different from average because of his sheer size. So, while I generally reject this way of thinking, there isn't even a part of me that can say, "Of course. Roethlisberger will be great in YPD in due time."

I think I did learn something about his value from compiling these data, and it wasn't to his credit.

Tony Romo

17.8 6.4 (6 good)

Mentioned in the Eli Manning comment.

Tobin Rote

8.1 6.8 (12 good)

Early AFL 6.8 of total.

Something of a mainstay at quarterback for the Packers before Bart Starr, starting 73 games. His early years included -5.3 and -4.1 seasons (1950 and 1953), but Green Bay was a pass-happy team from '54 - '56. Rote led the NFL in passing yards in 1956, and was also in the top 3 in '54 and '55. Green's Bay YPD diffential for these three years was right around 0, however. In 1963, Rote resurfaced as an AFL quarterback with the Chargers, having an apparent dream season that included All-Pro honors and a 51-10 drubbing of the Patriots (led by his one-time rival for playing time with the Packers, Babe Parilli) in the Championship. The Chargers offense was outstanding in 1963, with a 6.8 YPD Differential. John Hadl became San Diego's starter the next year, and although he was prolific over a good period of time, never scored quite as high in one season as Rote did in '63. Rote was an excellent running quarterback, averaging 4.9 yards a carry, and going over 3,000 yards for his career.

Matt Ryan

17.0 5.8 (15 good)

Averaging a healthy 4.3 a season, with all seasons easily in the black, despite average career yards-per-pass (7.0) and completion percentage marks (60.9). He's been as good as he has in YPD despite not having had any time to learn on the bench. He's only missed two games in his career because of injury, and his interception rate compares favorably with most of the other top quarterbacks in the league, only needing some more years to grow from its current "15 good" to an impressive total. His record really stands up to scrutiny.

I think, if you were to call him the poor man's Aaron Rodgers, you'd actually be selling Ryan short rather than overstating things. The comparison comes to mind because like Ryan, Rodgers started playing in 2008. Their total scores are close (19.3 for Rodgers, 17.0 for Ryan); their interceptions-saved less so (25 to 15 in favor of Rodgers), but both impressive. Ryan was better than Rodgers in 2008, having his best season, and the two players actually put up the exact same scores in 2009 and 2010. So Rodgers had never had a year where he moved his team more effectively than Ryan until last year, which was of course a landslide in Rogers' favor (8.7 to 2.5).

So how does Ryan put up such good YPD numbers despite his completion percentage and yards-per-pass average? For starters, he's avoided sacks: never more than 26 absorbed in any one season. I also know that in 2010, for example, the Falcons were masters of the long drive; I don't want to look deeply into it right now, but Ryan has probably done a good job of converting third downs.

The Atlanta running game hasn't helped their YPD considerably, at least as gauged by their yards-per-carry, which has ranged from 3.8-4.4 over the Ryan years. I actually see teams like the Falcons as typically starting fast and grabbing a lead, then getting complacent, running the ball, and finishing with YPD under their ability. This tye of team normally doesn't overperform -- and yet, under Ryan, the Falcons have.

The 2008 Falcons also belie the idea that conservative teams will necessarily not have YPD data as strong as their per-pass data augurs. Their 5.8 YPD throwing the 29th most passes in the league is another dagger in the argument that Roethlisberger did not have a fair chance to post strong YPD numbers because of the Steelers' approach.

Going forward, I would expect from Ryan at least as much success as we seen so far. He figures to simply get better with experience, and Julio Jones has the ability to catapult an offense.

Mark Rypien

14.0 4.6 (6 good)

If I had had to predict Rypien's profile beforehand, and you told me he had a 14.0 total, I probably would have predicted a brilliant season in there, a 9.0, probably in '91. But instead what emerges is a guy who was never brilliant but steadily good from '89-'92. 1989-1992 is almost his whole time as a starter, but his 1993 score of -3.5, and an even more distressing 4.7 yards per pass that season, does not suggest he was deprived his job unduly. Rypien had an on/off/on/off pattern with his yards-per-pass-attempt from '89-'92, but not with the Redskins' yards-per-drive.

It's something of a mystery how the Redskins were able to score 485 points in '91, about 100 more than they scored in '89 and '90. They had a good turnover season, forcing 41, but that was well behind two teams ahead of them, the Saints and Eagles, who forced 48. Forty-one turnovers was nine more than the NFL average. The Redskins also had a normal number of total drives on the season, and yards per rush of just 3.8. Perhaps great defense just gave their offense good field position. The special teams, at a glance, seems to have been less than stellar.

I remember long bombs as Rypien's trademark. It wasn't his arm strength that was generally credited for these completions, but instead the right touch.

What Rypien also did extremely well was avoid sacks. In his entire career, which included 78 starts, he was sacked just 97 times. Over '90 and '91, he was sacked an average of one time for every two games. In that era, this pattern was probably mostly credited to Washington's highly regarded offensive line, but current thinking gives quarterbacks the majority of the credit for low sack totals

Matt Schaub

19.5 5.9 (10 good)

If we characterize today's offense-dominated football as video-game football, the appropriate appreciation of Schaub probably suffers from video-game-football fatigue. As Schaub has never played in a playoff game, and Houston is very rarely if ever in widely-broadcast games, he's the guy who is unfairly left out of the discussion of the game's top quarterbacks.

I think the growing complaints about excessive offense have foundation, and do not just reflect a need to let off some steam, the way that say, complaints about the modern athlete and his lack of seriousness and proper perspective do. But the fact is that Schaub is 5th among active quarterbacks in career YPD-differential. That perhaps sounds more impressive than saying he's 38th all-time, so the numbers for active quarterbacks are not that inflated.

Now, it's true that Schaub is close enough to a bunch of close pursuers that he could just as easily rate 10th among active quarterbacks as 5th. But why shouldn't Tony Romo or Eli Manning be the quarterback left out of the discussion of the best, rather than Schaub? To a certain extent, I would actually say that Romo is left out; he's discussed a lot, but not always as one of the best. Romo is actually the most similar active guy to Schaub, in terms of his scores, how many years he has played, and how old he is. Schaub has been better than Eli Manning; there's not much question of that. He's completed just short of 65% of his passes for his career, for 7.9 yards a throw. He also doesn't struggle with sacks or interceptions. He's been well over 4,000 yards in passing twice, and led the NFL in passing yards in 2009. I think we need to make room for him on the pedestal.

Where you can fault him most is durability. He started just 11 games in 2007 and 2008, and just 10 this past year. In those seasons where he missed significant time, his backups averaged 7.4 yards a throw, compared to the 8.1 he averaged while in there.

The math works out quite neatly, because Schaub has actually played four full seasons (64 games) with Houston, not the five the YPD approach gives him credit for. But we can't just dock his seasonal average of 3.9 from his score, because his average is actually better than that when he's in there; his backups have almost certainly dragged it down somewhat. If we say that his backups have performed at a 1.5 level, then Schaub's true rate is 4.5 a season. Give Schaub four seasons of 4.5, and his total is 18.0.

That's hardly lower than his current 19.5, and he's also not the only guy in NFL history to have missed games. The Marc Bulger comment may help spur the kind of thinking that is needed to adjust YPD for games missed in different kinds of cases.

Jay Schroeder

7.3 2.7 (1 bad)

Phil Simms

8.6 3.7 (29 good)

As a Giants fan who grew up in the '80s and now admires Simms' honest and independently-formed views about football, I was surprised and disappointed by his profile. The Giants did not win in the Simms' era because of their offense, YPD says. Simms was really a game manager, whose value to the team was that he threw fewer interceptions than most of his contemporaries. Unlike the prototypical game manager, he did make the big play, finishing in the top 10 in yards per completion seven times. But the overall package did not make for good offense. One of the problems was lots of sacks -- Simms was among the top five-sacked quarterbacks in the NFL every year from 1984-1988.

Simms' start was particularly grisly, which I'm sure could be partly blamed on his supporting cast, and perhaps also on coaching. His '79-'81 seasons were -6.8, -5.1, and -6.1. Archie Manning never had three consecutive years like that. Charlie Conerly nearly did, totaling -15.9 from '51-'53, when he was actually in his early 30s. Both Simms and Conerly have very good and very similar interception records, but overall in YPD, they are in different categories, as Conerly was negative nine times in 12 years as a starter, and Simms just six times in 11 years.

Brian Sipe

17.4 5.4 (16 good)

Sipe wrapped up his NFL career just as I was beginning to follow football, and my image of him has been more a fan favorite than a really good football pleyer. YPD begs to differ. Sipe was positive in seven of his eight seasons as starter, including seasons of 2.6, 2.8, 5.4 (the MVP 1980), and 4.2. His one negative season was the strike season of 1982, and he missed three games in addition to the strike, so the performance was probably fluky. Although he twice led the league in interceptions, he got there by throwing a lot of passes more than because of anything else, and saved 16 interceptions more than he gave up over the course of his career.

Norm Snead

14.3 5.1 (2 good)

You can find Snead's year-by-year totals in the introduction. Since 10.0 came from two years, seven years apart, one is sort of tempted to say that if a guy hangs around long enough, he's bound to have some good happen to him. The high years are from a nine-start season with the '65 Eagles, which doesn't appear to have been a very interesting season, and from the '72 Giants. Snead took over from Tarkenton after '71; the Giants were unexpectedly very good offensively during this era, which I suppose could be called the Alex Webster era as correctly as it could be called any other. A neat coincidence is that Snead and Tarketon were rookies in the same year, 1961.

They are also 5th (Tarkenton) and 6th (Snead) in career interceptions thrown. But Tarkenton saved 83 interceptions over his career because he threw his interceptions in over 6,000 passes, while Snead battled interceptions to just about a standstill, with over 4,000 passes thrown. Snead led the NFL in interceptions four times: in '63, '68, '69, and '73. He doesn't seem to have thrown a tremendous number of passes in those seasons, the way that Brian Sipe did in '79 and '81 (Snead's highest rank in pass attempts in the interception-leading seasons was 4th).

How could he have been a zero-sum INT quarterback over the course of his career, then, and not a negative one? My interpretation is that the really bad INT quarterbacks oftentimes do not stay in the lineup long enough to even lead the league in INTs thrown for a season. So, of the guys who have careers of any longevity, the strong majority were either good INT quarterbacks or average INT quarterbacks. The average INT quarterbacks go down in fan's minds as bad INT quarterbacks, although they really are not. The standards are just higher if you're good enough to stay in the lineup.

Kordell Stewart

10.5 6.9 (6 bad)

See the Roethlisberger comment. Stewart was a really interesting player, and I bet people of my analytical persuasion years younger than I am will be even more confounded by his record than I am. At least I had the advantage of living through it.

I wrote that quarterbacks' career-YPD numbers generally have arcs to them, but Stewart was outstanding in '97 and '01 and not a good player in the intervening years. If Steelers' fans had past given up on him at the start of 2001, they could have been forgiven.

The 2001 season was brilliant: not only did Pittsburgh's YPD lead the AFC, but Stewart ranks 24th on this list of players in terms of his best season.

How, then, did Bill Cowher give up on him so soon the next year, after carrying him along from '98-'00? Stewart started just 12 more games for his career, including just five for the '02 Steelers. He was benched in '02, and not injured.

Not only did the Steelers run more in '97 and '01 than most top offenses do, but their effectiveness eludes yards-per-pass attempt, where Stewart topped out at 7.0 in 2001. Although not good offensively from '98-'00, the Pittsburgh offenses were always better than their yards per pass said they should be. Stewart's average YPA in the '98-'00 years was in the danger-zone at 5.8, while Pittsburgh's YPD differential was between -1 and -2 each year: thoroughly mediocre, but not really bad.

The obvious factor to point to in the overachieving is Stewart's running. I have found only weak evidence that quarterbacks who rack up good yardage and yards-per-attempt totals running add to the offense. Stewart was certainly also a cut below the great running quarterbacks -- at least as measured by those statistics. He was tyically around 450 yards with a 5.0 average. But maybe he was timely with his runs, or maybe he added other assets, not obvious in the statistics.

Since the Steelers' 2001 offense was so surprising in rating so highly, it's worth pointing out how singular they were in rushing yards. They ran for 173.4 yards a game and had over 530 more yards than any other team. The difference in rushing yards between the Steelers and #2 was greater than the difference between the #2 team and #17. Yet I would argue giving the passing game the lion's share of the credit for the team YPD is not misguided, because in yards per carry, there is not affirmation of how the Steelers were moving the ball so well. They averaged 4.8 yards a carry -- that's not exactly O.J. Simpson in 1973, or Jamaal Charles in 2010. They averaged 4.8 yards a run, and 6.4 net yards a pass, and so made more impact when they passed. Let me put the issue that way.

Vinny Testaverde

11.9 4.6 (39 bad)

Testaverde's two best seasons by YPD were what you would expect: 1996, when he threw for over 4,000 yards and 33 touchdowns with the Ravens; and 1998, when he went 12-1 as a starter for the Jets, and threw for 29 touchdowns. Both of those were 4.6 seasons.

Joe Theismann

9.0 3.1 (24 good)

Theismann was a towering figure when I started watching in football in '82 and '83. As a game broadcaster, he first seemed cocksure and later, right an uncanny percentage of the time. As a player, he seemed to know just what to do, and the Redskins were always winning, going 28-4 over the regular and postseasons of '82 and '83. But in his other seasons as a starter, only twice was he the starting quarterback for nine or more season's wins, so I had a bit of a skewed view of things. Pro Football Reference also lists him as a Pro Bowler just twice. He probably doesn't have any claim as a great player, although he might have had a small shot had he become a NFL starting quarterback before age 29. His 1983 MVP season, when the Redskins set a record for points scored, comes in at just 3.1. This is the biggest point missing on his YPD resume: a big season. He is, in this respect, much like Steve McNair and Terry Bradshaw, a couple of other seasonal MVPs. Like them, Theismann did have a good streak of positive seasons, which encompassed his final five seasons in the league, but just no big years. The other four positive seasons in the streak all came in from 1.0 to 1.6.

3.1 really isn't a bad score, even for a MVP, but you might question if the Redskins' 541 points mean that Theismann deserves more credit than the YPD system gives him. He was actually on a conservative team that managed 1,347 yards from John Riggins on just 3.6 yards a carry, so perhaps he does. But the Redskins' point total has to largely be attributed to an astonishing 61 turnovers forced, not far from the 1961 Chargers' all-time record of 66.

Theismann was good at avoiding turnovers himself, or at least interceptions, as his career mark is "24 good."

Tommy Thompson

11.7 5.2 (18 good)

The quarterback who so outplayed Paul Christman in the '47 Championship Game but came out on the wrong end. Perhaps the Eagles were determined never to make the same mistake again, because after throwing 44 passes in the '47 Championship game and losing, Thompson threw just 21 combined in the '48 and '49 Championship Games, and the Eagles won both (with no small help from their defense, as they shut out the Cardinals in the rematch, and shut out the Waterfield and Van Brocklin-led Rams the next year).

The Eagles piled up 118 runs in the '48 and '49 Championship games. They had Steve Van Buren, the NFL's leading rusher in both years, and indeed in '45 and '47, too.

It certainly doesn't follow that because the Eagles pulled in the reins, Thompson didn't throw well in his limited opportunities, particularly since the Eagles won the games. But in fact, Thompson was very 1947 Christman-like in the championships, throwing two interceptions in both games, and sporting a horrible 2-12 for 7 yards line in the '48 game. So you have the very odd situation from '47-'49, where if somebody knew nothing about football, and observed the passing lines of the winning and losing teams, they would think good was bad, and bad was good.

Ironically, when it came to regular seasons, Thompson was excellent in '48 and '49, and average in '47, at least in terms of YPD.

I've invested next to no time trying to find out, but it's not quite clear to me how highly regarded Thompson was. A player whose career was interrupted by World War II, Pro Football Reference gives only one Pro Bowl selection for him, in 1942. I would steer away from the traditional categories as the ideal way to rate a player, but Thompson never led the NFL in yards passing, or in completion percentage. He led in touchdowns passing in 1948. Given the Eagles struggles at the highest levels of the game historically, it's hard to imagine any championship-winning quarterback of theirs not being looked upon with some reverence (although Eagles fans were certainly less numerous and less long-suffering in Thompson's day).

Richard Todd

4.5 2.9 (25 bad)

I scored Todd because the '82 Jets were 3rd in the AFC in yards-per-drive differential at 2.9. His overall profile is one of the worst here: he's 108th in total score; tied for 106th in best season; and tied for 111th in terms of interceptions saved/cost.

I confess to being highly satisfied that my approach finds his game to have been essentially without virtue. As a little kid, I remember my father tucking me into bed, and telling me that the Jets had a good team, but there was one concern, which was that their quarterback wasn't very good. Todd was then for me the original player who stunk yet seemed to be exempt from the possibility of replacement. Many other players would follow, with my father doggedly yelling at the television set in the case of each one.

Looking at his record, Todd probably did milk as many seasons as he could have out of his talent. "15 bad" in interceptions over '79 and '80, including the league leader with 30 thrown in the latter season, Todd was then a good INT quarterback over '81 and '82, saving 12 interceptions over those two seasons. If he had just been mediocre in interceptions the whole time, it's doubtful the Jets would have stayed with him for so long. But he seemed to be showing vast improvement, and the Jets also were a playoff team in both later years. In '83 he was back throwing interceptions again, the 2nd most (26) in the NFL, although with the 3rd most passes (518). He was with the New Orlean Saints for the next season, which proved to be his swan song as a starter.

The 1979 Jets, 8-8, had a very odd offense. Todd led the NFL in yards-per-attempt. Completing just 51.2% of his passes, 2.9% below the league average, you certainly would have thought he would have led in yards per completion, too, but the 49%-completing Steve Grogan beat him out (Grogan is also one of the few quarterbacks to have cost his team more interceptions over the course of his career than Todd).

When I saw the high yards-per-completion for Todd, my first thought was "Wesley Walker." Indeed, Walker did average 24.7 yards a catch, but missing seven games, caught only 23 balls. He led the Jets in receiving yards, though. Doesn't that sound like something from before World War II? A receiver leading his team in receiving yards on only 23 catches? I feel very old to have been alive when this was still happening in the NFL. (Footnote: Stephen Baker led the 1990 Super Bowl-champion Giants in receiving yards and had only 26 catches.)

One reason Walker could lead the Jets with just 569 yards receiving is that they seldom threw. In run attempts - pass attempts, they led the NFL, with 265. If you don't know who they could have been giving the ball to in 1979, you're excused. Their horses in the backfield were Clark Gaines and Scott Dierking. Two unlikely subjects around which to revolve the game's most active ground game!

The sum total of these oddities was a YPD of 0.7.

Michael Vick

11.1 4.3 (7 good)

Vick produced slightly positive returns in his Atlanta days. His Eagle teams (I use "Eagle teams" rather than Vick because of the odd start he has missed in both 2010 and 2011) have been good, but not great. The expected improvement from the Atlanta days inarguably does show up in YPD, though. Since Vick has only had one negative Differential in six years as a starter, with below-average yards-per-attempt (6.7) and completion percentage (55.3) records for his career, it seems that his running is legitimately helpful to his teams, although it is not something that should vault him to the top of the league's quarterbacks.

Billy Wade

8.2 4.6 (20 good)

The Rams' 1951 offense (11.0 YPD) was comparable to their 2001 offense (10.1). While Bob Waterfield would retire after 1952, Norm Van Brocklin was only 25 years old, and threw more passes in '51 than Waterfield did. Yet the Rams still made Vanderbilt's Billy Wade their #1 pick in 1952, the first overall. This sounds a little bit like the current Peyton Manning/Andrew Luck drama.

Wade wouldn't actually become the Rams' starter until 1956, a year in which he split time with Van Brocklin. Wade was back on the bench the next year, then the Rams starter from '58-'60.

Despite the team's 4-8 record, 1956 scores as Wade's best year with the Rams (how much Van Brocklin's contributions helped is unclear, although the 4.6 score is better than Van Brocklin did in '55 or '57. Wade also probably had the better stats). From '58-'60, Wade didn't really stand out, even with the legendary Sid Gillman being his head coach in the first two of those years.

The second phase of Wade's career was with the Bears, and he is known as the quarterback of the 11-1-2 champions of '63, who halted the dominance of the Packers. That was of course a fantastic defensive team. The offense was -0.5 in YPD, although Wade's INT rate was 2nd-best in the league (the highest he ever rated, by the way).

To look at Wade's career as average at best (at least when it came to ball moving), with some help early on from the foundation laid by the Rams' juggernaut, seems accurate to me.

Bob Waterfield

14.8 5.9 (11 good)

I feel like I've at least been talking about Waterfield at the edges through this entire history, but to spell the basics out more clearly: he played for eight years, but gets credit for just four years as a starter. The four non-starting years, however, include two years where he split the Rams' passes very close to evenly with Hall-of-Famer Norm Van Brocklin; one year where he threw about one pass for every two that Van Brocklin threw; and one year where he split the passes very evenly with a guy named Jim Hardy. Before Van Brocklin came along, the Rams' strongest years with Waterfield were his first two, where he scored 5.7 and 5.9. These seasons also came before the arrival of the Hall-of-Fame ends Elroy "Crazylegs Hirsch" and Tom Fears (although the Rams did have a receiver in '45, Jim Benton, who not only had a 1,000 yard season, but accounted for 60% of their receiving yards! A receiver accounting for 60% of the 4,000 yard totals we see today would have 2,400 receiving yards).

Waterfield also deserves particular credit for the '51 season, one of the two teams where the Rams won a championship with him on the roster (the other was '45). Their YPD that season was 11.0, as mentioned in the Billy Wade comment. That was one of the seasons in which Waterfield split time almost evenly with Van Brocklin, and also the season when his achievements were most competitive with Van Brocklin, from a statistics standpoint. In the championship itself, he threw 24 of the Rams 30 passes, not surprising since he was the more experienced player. However, his numbers were not good, and Van Brocklin was likely more the hero, breaking a 17-17 tie with a 73-yard pass to Fears.

Waterfield went into the Hall of Fame before Van Brocklin. I get the feeling that he was believed to have that "je ne said quoi" more than Van Brocklin; the older player is always going to have that respect, sort of the way an older brother has it in a family. Sammy Baugh was the first quarterback (at least of the ones listed here) to make the Hall; the quarterback class two years later counted Graham, Waterfield, and Luckman in it. Y.A. Tittle, a more modern player, and Norm Van Brocklin were the next quarterbacks in, but that wasn't until six years later, in 1971.

With Van Brocklin retiring in 1960, I'm not sure if he would have been eligible for the Hall of Fame in 1965, when Waterfield was elected; under today's rules, he wouldn't have been. In any case, he was obviously not elected in his first year of eligibility, but maybe the selectors just sort of forgot about him until they decided to address quarterbacks again, having already produced the Graham, Waterfield, and Luckman class.

Today, it's not just my impression that Van Brocklin has eclipsed Waterfield, though, it's real: if you want to sponsor Norm Van Brocklin's Pro Football Reference page, you're out $55; for Waterfield, you're out just $20. And, although both individuals retired over 50 years ago, I have no doubt that Van Brocklin's somewhat-more-recent connection to fans serves him well when it comes to the sponsorship price. What goes around, comes around, I guess.

Danny White

6.3 2.6 (5 bad)

The Staubach comment is a good start, but I wanted to say some other things. YPD really doesn't like White, and it's certainly a surprise his best season only comes out to 2.6. I say it's a surprise, because while no one really wants to acknowledge it, the Cowboys won an awful lot of games in the early '80s: they were 12-4 in '80, '81, and '83, and made (but lost) the NFC Championship game in '80, '81, and '82. They remind me of the Yankees of the same era, although the Cowboys had been outstanding for many more years consecutively preceding the '80s than the Yankees had been. Not only did the Cowboys win games, but they scored points: the most in the NFL in 1980; tied for the most in the conference in '82; and 479 (more than everyone but the Packers and Patriots this year) in '83. Their defense did create a lot of turnovers, and their number of drives looks highish, but I would still have thought their yards-per-drive would be higher.

I don't think this really applies to the White Cowboys, but maybe with teams that create a really extreme number of turnovers, their YPD are biased downwards because they score a lot of touchdowns after having short fields. Those drives would have gone on and included more yards with longer fields. This analysis would seem to be germane with the '83 Redskins and MVP Joe Theismann, if it ever was.

White was a career 59.7% passer, which certainly looks good given expectations for the era, although he never rated higher than 4th in the NFL in completion percentage for a season. With lots of explosive playmakers, it's not surprising that White's yards per attempt were at least 7.5 every season in the points-laden '80-'83 seasons. What I'm trying to say here is that White's YPD appears an anomaly from the standpoint of his individual statistics, too.

The final point that has to be made about White's YPD profile is that it looks better season by season than it does in the box. Yes, his numbers are lower than I expected, but they are all positive for the first five seasons. This is another guy underrated in the career total because he didn't have anything approaching a big year.

The YPD numbers for White's Cowboys are interesting, food for thought. They seem not to capture the team's true offensive efficiency, however.

Now, how good White really was, and whether he personified anybody's idea of a good quarterback, is another question. The prevailing belief is that he had a lot of help.

Doug Williams

3.4 2.0 (20 good)

I'm sure this must have been part of the conversation about Williams, but it wasn't the dominant part, so I don't remember it. The man absolutely would not take a sack. He qualified for the "sack percentage championship" (by throwing the ball enough times for Pro Football Reference's liking) five times, and he rated 1st, 4th, 4th, 2nd, and 2nd in the NFL in terms of having the lowest percentage of sacks taken. In 1978, his rookie season, he started 10 games, meaning I count that season as one where he started, too, and if he'd thrown enough passes, he would have led in sack percentage that season as well. Taking so few sacks as a rookie goes against what we expect -- are their studies on the subject? I just think it's interesting that Williams was able to come into the NFL with this skill.

If it was something that was coached into him, it did carry over to the Redskins when he was the majority starter in '88, too, and was 2nd in sack rate. Although judging by Mark Rypien's sack rates over the years, the Redskins did have a good pass-protecting line in those days.

Williams almost surely traded incompletions for sacks. With all the talk about the Broncos winning games this year despite Tebow's outlier completion percentage, the Buccaneers made the 1979 NFC championship game, and Williams' completion percentage in 16 starts in the regular season was 41.8 (league completion percentage was 54.1). His yards per attempt were lowish at 6.2; the Buccaneers ran for an average of 4.0 yards a carry; and yet the Bucs' YPD differential was hardly noteworthy at -0.9. The low sack rate was a big aid, and the terrible completion percentage was misleading.

I certainly don't want to give the impression that Williams rates at all as a good quarterback statistically, because he doesn't. Overall, he was awful as a rookie, and fair after that with Tampa, although perhaps improving in his last years before going to the USFL. In his one season starting for the Redskins, his score was his career high, 2.0, but that did place 3rd in the conference. On the plus side, Williams not only avoided sacks, but interceptions to a degree, too, with a career mark of "20 good." He did enough things well, both statistically and impression-wise, that with a good supporting cast early in his career, he might have shown himself a fine player.

Jim Zorn

7.3 3.6 (10 good)

The Seahawks went 9-7 in just their third and fourth seasons, and Zorn to Largent was a big reason why. The offense had YPDs of 3.6 and 3.2 in those two seasons. Otherwise, Zorn's yearly YPD numbers get lost in the shuffle of starting quarterbacks.

How completely are the recent guys dominating this statistic? Lengthy and careful consideration

So, I had intended to finish up with methodological odds and ends, to give those of you interested in more precise rankings some suggestions on how to achieve them. Besides, I felt like I had to have such a section for my own credibility; it's galling and seemingly naive when a writer forges on as if important issues don't exist. But having finished the work, it seems that one issue dwarfs the rest, and that I should devote my energies exclusively in that direction.

Namely, the question is if modern players really do dominate "the statistic" as much as they seem to? I think that has to be established first, and so I've attacked it from many angles. Whether one chooses to adjust the pattern is largely a matter of personal taste, so I haven't gone into that as deeply.

YPD differentials were always figured against the conference average, and against the NFL or AFL average separately from 1960-1969. Therefore, after 1950-1959, there were 20 separate leagues in a decade. Expansion after 1970 has been rather tame, from just 13 teams per conference, to 16 today. So examining the YPD-differential scores of the best and 2nd-best teams in each conference in each year seems like a pretty good way to measure dominance over the years.

The shakiest league for this approach is the AFL from 1960-1967, which had only eight or nine teams. You'll have to accept that these numbers may then be underestimating the dominance in that era a bit.

Compiling the conference leaders year by year, I was struck by the extent individual teams had a hold on the positions in different eras. For instance, the Rams led in YPD five times from '50-'57; the Chargers led five times from '79 - '85, and were 2nd the other two years; the Colts were first four times, and second five times from 2000-2010; and most glaringly, the 49ers led the NFC every year from 1982-1994, except for 1986, when they were 2nd (Montana only started eight games that year).

I think this problem is largely corrected by examining the average for the second-leading time yearly in each decade, but for another perspective, this is what I did. I took the third-highest single-season score by any team in a conference in a decade, with the restiction that the first three franchises all had to be unique. For instance, in the NFC in the '80s, the best four scores were all posted by the 49ers, so the 6th-best score ('84 Cardinals) and indirectly the 5th-best score ('89 Redskins) determine the result. The '84 Cardinals had a YPD diff. of 4.5, so that's what I use. Then the AFC in the '80s produces the '82 Bengals (7.4); 4.5 and 7.4 combine for a rounded 6.0.

I'm not sure this approach captured what I wanted it to; I'm not sure it reached far enough down into the pool of teams, and going with the 3rd-best franchise still seemed to leave the chosen teams at the mercy of the number of outlier franchises in the decade. Oh well.

So here's a starting table with all of this information. For clarification, the "best diff" and "2nd best diff" are the average of the best and 2nd-best diffs year by year, conference by conference, across the 20 conferences -- not the best or 2nd-best of the entire 10-year period. The third statistic does take the entire-10-year-period approach, again only allowing one contribution per franchise.

It didn't make sense to calculate the third statistic for the four-year period involved with 1946-1949, so I exclude it from the table.

Time period(1st_col) Best_diff(2nd_col) 2nd_Best_Diff(3rd_col) 3rd_Best_Unique(4th_col)

1946-1949 6.0 4.7 -
1950-1959 6.1 4.0 6.4
1960-1969 5.2 3.9 6.3
1970-1979 5.0 4.0 5.7
1980-1989 5.8 4.0 6.0
1990-1999 6.8 4.5 7.6
2000-2011 8.1 6.0 9.5

Taking the first four full decades, if you use the average of the best team, you have a parabolic shape, with scores lower in the '60s and '70s than '50s and '80s, although perhaps it could be said that the differences were modest. One could have lived through the time as a football fan, and even have studied the numbers, and not really have noticed the shift, perhaps. The comparison of '70s to '80s, say, might not be apples to apples, but neither is it offbase or academic.

Interestingly, the 2nd-best-team approach has each of the first four decades incredibly close in terms of dominance; the implication is that the Rams of the '50s, the 49ers, Chargers, and Dolphins of the '80s are skewing the "best team" numbers. Certainly, the conference-leading 12.0 for the '82 Chargers and 10.8 for the '84 Dolphins were not like anything else for that decade.

Scores with the independent-teams approach can't be summarized as neatly, but seem to show little change over the first four decades, with perhaps an overall trend of decline in dominance.

No matter which of the three measures you use, offenses were more dominant in the '90s than in any of the first four decades. The gain in the second-best conference offense is still modest, but it can't fairly be said that the best-teams score for the 1990s was driven by outliers; the median for best teams was the same as the mean, and there were seven different franchises among the the top 10 conference-leading single seasons (Oilers, 49ers, Dolphins, Cowboys, Broncos, Vikings, and Rams).

As the table shows, since the turn of the century, the best marks clearly have not looked like they did in the '60s or '70s, or even the '80s. Whenever the growth took place exactly, the cumulative reality of it is unmistakable.

Remember that the differential in YPD differential is with the league average. In other words, even though the mean leading-scores have risen, the sum of all scores in the league is 0 every year. So it's not simply the case that YPD differential scores have risen because passing and offense have exploded. The phenomenon we see here is not quite the same as the overall increase in offense. The YPD differential numbers suggest that not only is the league up in offense (presumably), but the best teams are up even more than the average teams.

Before we can draw that conclusion however, we have to consider that maybe a simple differential between yards per drive and league yards per drive is not the way to go. This theory says that improving 5 yards on a base of 30 yards per drive is not the same as improving 5 yards on a base of 20 yards per drive. Perhaps the differentials started growing 20 years ago because yards became easier to come by. If we were measuring how rich someone was for his era, we probably would do it in terms of multiples of the average net worth, not in terms of raw differential from the average.

Before we can evaluate this theory, we first need to see if the increases in YPD-differential mirror the increases in offense (i.e, yards-per-drive). It's also fun and instructive just to look at the state of offense through the years, setting aside whether the changes dovetail with the YPD-differential changes.

Some people will ask why I break this statistic down for every five years, while I go by every decade for YPD Differential. I felt that meaningful changes in the league average could materialize in a single year, so 10-year averages did not provide enough detail. Conversely, just averaging a group of select teams, as in the case of leading and second-leading teams, required the sample size that a larger block of years provided.

Raw YPD by Period (AFL or AFC averaged with NFL or NFC after 1960)

1946-1949 22.1
1950-1954 21.5
1955-1959 23.3
1960-1964 23.2
1965-1969 22.6
1970-1974 22.4
1975-1979 23.8
1980-1984 26.1
1985-1989 26.2
1990-1994 26.7
1995-1999 26.9
2000-2005 27.4
2006-2011 29.0

While it can be noted perhaps with some justification that

a) average YPD in the second half of the '50s was almost two yards higher than it was in the first half

b) YPD in each subsequent five-year span after 1950-1959 declined until 1975-1979

c) YPD has increased in every period since 1970-1974

these are secondary aspects of the table in my mind. The run of consecutive decreases in the first half of the table, and then increases in the second half of the table, seems like happenstance more than anything else, because the sum of them is so small (a grand 3.1% increase between 1980-1984 and 1995-1999, for instance).

When YPD really increased, at least in terms of five-year intervals, was in the '80s, and to a lesser extent, in the second half of the '70s. Additionally, we do see the big bump we expect in the last six years (I used six years and not five to divide the 2000-2011 period evenly). So the increase in offense generally preceded the incremental gains by the top teams, which occurred in earnest no earlier than the '90s. Not surprisingly, then, it will be shown that "a percentage correction" cannot account for all of the increase we see with the top YPD scores.

Rerunning the YPD differential table for the average leading and runner-up teams of each decade, this time with the percentage difference over average, we get

Time period(1st_col) Best_diff(2nd_col) 2nd_Best_Diff(3rd_col)

1946-1949 27.0 21.1
1950-1959 27.3 18.1
1960-1969 23.0 16.9
1970-1979 21.5 17.2
1980-1989 22.3 15.2
1990-1999 25.4 16.7
2000-2011 28.5 21.1

From the league averages for YPD shown before, it was a mathemetical given that at least some of the superiority for the best teams between say, the post-2000 years and the 1960s would vanish. However, the trends in this new YPD differential statistic remain the same. The best YPD teams from 1946-1959 were better than the best YPD teams in the '60s and '70s; the top teams took their games to a new level in the '90s, and augmented that standard even more post 2000. The fact that the trends are still there, if not quite to the same degree, again tells us that the error in how differential was calculated before did not wholly account for the high scores of recent offenses.

I've grown very used to YPD differential with the +- in terms of yards, of course, so I like to convert the average conference-leading and conference-second-leading scores to that metric. The average YPD over all years since 1946 seems to be just a little under 25. If we use 25 as a base, then a percentage-difference, such as 27.0 for 1946-1949, can be converted to YPD +- by 25*.27. This gives 6.8. The entire table, again with conference-leading teams in the second column, and conference runnerups in the third, in terms of yard differential against a base YPD of 25.

1946-1949 6.8 5.3
1950-1959 6.8 4.5
1960-1969 5.8 4.2
1970-1979 5.4 4.3
1980-1989 5.6 3.8
1990-1999 6.4 4.2
2000-2011 7.1 5.3

There's an advantage to converting the percentages back to +- scores, in addition to just familiarity and stubbornness. The differentials can be compared to the original table. The conference leading teams of 2000-2011 had a 3.1 YPD greater edge over the average team than the 1970-1979 conference leaders. Once we put both eras relative to a 25-YPD base, the edge for 2000-2011 is 1.7 YPD. So, approximately half of the difference between top teams today in YPD-differential and top teams in the 1970s is just the metric, and half is a real difference in dominance relative to the league.

When I said that the trends of the original table stand once a percentage standard is adopted, I was taking some liberties. In raw differential (subtraction), we had an increase for the conference-best teams in the '80s that seemed to signal what happened later on a larger scale. Using percentages, this disappears. The increase in the '80s is simply a function of the improvement in offense, and the devaluation of a yard.

One might also note that, in terms of percentage over average, the '46-'49 period and '50s now rival today's scores. I think the decade-long grouping is misleading on that score, however. The big percentage score for 1950-1959 is wholly a 1950-1954 effect. The average leading team over those five years was 36.1% over the league average, while the average leading team for 1955-1959 was 18.1% over the league average. I was tempted to say that the effect was solely a Los Angeles Rams effect, but the league-leading Browns and 49ers of '52 and '53 had scores of 29.7% and 32.4%, respectively, while none of the league leaders from '55 to '58 reached 20%, including two Rams teams.

In terms of the second-best offense in each conference, the mini-surge in the '90s also disappears once we convert to percentages.

Although it's not strictly relevant to the questions involved with this work, I wanted to delineate the timeline for the league-wide offensive increase that occurred in the late '70s and early '80s more exactly. When one see increases in that general period, one immediately suspects that the rule changes of 1978 were paramount. These were to allow offensive lineman more freedom in pass blocking, and to outlaw contact more than five yards down the field. YPD did increase in 1978 to 24.3, compared to the 22.8 that it was in 1977. That's a good size increase for one year. But 1979 (25.2) was almost as much above 1978 as 1978 was above 1977, and 1980 likewise a healthy new top (26.2) from 1979. It's as if teams didn't really realize how to exploit the liberalized rules immediately, and it took a few seasons for the new rules to have their full effect.

Also looking at the jump from 2000-2005 to 2006-2011 more closely, 2006 was not a year of increase. If there has been a qualitative change recently, a distinct new era in offensive proficiency, it is best understood as beginning in 2007 or 2008. There's a feeling that 2011 was completely unlike anything we've seen before, and it did set a new record for YPD average, exceeding the threshold of 30. But 2011's YPD was only a yard higher than 2010's. 2008's was 0.8 higher than 2007's, and 2007's was a yard higher than 2006's. There was no growth in offense from 2008-2010; the YPD were almost the same in each of those years. The offensive story in this era is in flux and still being written, but it's certainly possible that 2011 will ultimately just be the highest of many high surrounding YPD years, and will be able to be grouped with the rest.

It would take a substantial amount of work to rerun all of the scores to account for the relative difficulty of dominating. But anyone wanting to work in this framework in the future should probably use percentage-above-or-below average, rather than distance. Whether an additional correction should be levied to make scores of recent seasons fit better with historical scores is a more open question. Looking at how the season's best and second-best scores compare with previous eras, multiplying them by anywhere from .67 to .75 would probably level the playing field. Note that this correction will be much less steep if the percentage adjustment has already been made.

Reflections on a racing form

Monday, January 30, 2012

Losing a step? Not the gals in the SM F&M Turf

Sunday, January 29, 2012

The Yards-Per-Drive framework for rating quarterbacks