Tuesday, March 14, 2017

Ranking the Top QBs in Total Statistical Production

By Brad Oremland

Author's Note: This is a very long post, but I discourage you from skimming it. Wait to read it until you can go over it without feeling distracted.

Two years ago, I wrote an exhaustive series on the greatest quarterbacks of all time. That was a subjective ranking, but I also discussed the formula for Quarterback Total Statistical Production, QB-TSP. This post concerns that stat, QB-TSP, so you may want to read that link if you haven't already.

I've made three minor adjustments to the formula since that writing:

1. Eliminated the +1 for completions and decreased all the constants by 0.5. This has minimal effect on scoring but de-emphasizes completion percentage.

2. Made the AFL's constants more charitable by a factor of 1/3. I think TSP was underrating AFL QBs.

3. Increased the exponent to 2. That's probably a shade too high, but it's easy to work with.

I'll quickly review seasonal values, then the blind spots of TSP, and then we'll do a list: the top 100 Modern-Era QBs, as ranked by TSP.

Let's start by distinguishing [1] raw TSP, [2] era-adjusted TSP, and [3] yearly value points.

Raw TSP is the basic formula: Passing Yards - Sack Yards - [ constant * (Attempts + Sacks) ] + (20 * Passing Touchdowns) - (40 * Interceptions) + (0.5 * Rushing Yards) + (20 * Rush Touchdowns) - (20 * Fumbles). You can calculate this with pencil and paper, or even in your head, which I like.

Era-adjusted TSP is modified by a 5-year rolling average. This is the appropriate figure to use when comparing across different years.

Yearly value points indicates the exponent-modified stat. This is the one you should use when adding up career values, so you don't overrate compilers.

So what do all these numbers mean? From this point on, all references to TSP signify Era-adjusted TSP.

* Zero TSP (0 pts) indicates replacement-level performance. 2016 example: Brock Osweiler. I have data going back to 1946, and I believe Osweiler is the only player with 100+ attempts to score exactly zero. Jared Goff scored lowest this season, -422.

* 500 TSP (0.25 pts) is an inconsequential season, a bad starter or a good part-time player. 2016 examples: Brian Hoyer, Matt Moore.

* 1000 TSP (1 pt) is an average season. The player had some value to his team, but he wasn't a Pro Bowl-quality performer. 2016 examples: Alex Smith, Sam Bradford.

* 1500 TSP (2 pts) is a good season, a top-10 season, a borderline Pro Bowl season. This is a clear positive contribution to any player's résumé. 2016 examples: Derek Carr, Matthew Stafford.

* 2000 TSP (4 pts) is a great season. It's a top-5 performance, the player almost always makes the Pro Bowl, and he'll usually generate some all-pro support. If you have more than two or three seasons like this, you're probably going to the Hall of Fame. 2016 examples: Drew Brees, Aaron Rodgers.

* 2500 TSP (6 pts) is an exceptional season. These only occur about twice every three years. Most of them were first-team all-pro, and about half were named league MVP. 2016 example: Matt Ryan.

* 3000 TSP (9 pts) is a legendary season. The player always wins MVP, and these are seasons that educated fans know about: Graham in '47, Fouts in '82, Marino in '84, Young in '94, Peyton in '04 and '13, and Brady in '07.

I will refer to these numbers a lot, so I recommend re-reading them if you've been skimming.

Last item before the career rankings: what do these get wrong? First of all, this list is purely statistical. It doesn't know the offensive scheme, how good the linemen and receivers were, or anything else. It's just stats, and it differs from my opinion, sometimes significantly.

Second, these are Modern-Era (1946-present) rankings only. I did a work-up on pre-Modern QBs whose careers stretched into the late '40s, like Sammy Baugh, but only looking at seasons from 1946 onward; they're not technically ranked. Actually, Baugh and Tommy Thompson are the only ones who made the top 100 anyway.

Third, I only used data from the NFL, AFL, and AAFC, so players who had good seasons in other leagues are underrated. The USFL was a major league, and its players are underrated, most notably Jim Kelly. The CFL is not a major league, but readers who remember my rankings of Warren Moon and Doug Flutie will recall that I believe it's inaccurate to exclude CFL seasons from a comprehensive analysis of their careers. Nonetheless, they don't count here. Moon is underrated by TSP, and Flutie's not in the top 100.

Fourth, these are regular-season rankings, so outstanding postseason quarterbacks are underrated. Fifth — I know this is heresy to some folks — I think compilers are underrated here. Exceptional seasons score a little too high, while above-average and good seasons score a little too low. Anyone who was great in a partial season is probably underrated (e.g. — Tom Brady in 2016). I think rushing is probably a little underrated, so great scramblers may rank one or two spots too low.

TSP aims to balance efficiency with production, so that neither one alone will yield a good ranking. If a QB passed 600 times but kind of sucked, he's not going to get a good score. If he was really awesome in both games he played, that's not going to do much for him, either. That aligns with my philosophy, it's what I want from this stat. But there are some cases where it goes astray: good QBs on run-oriented offenses, good players who got hurt or took over partway through the season, etc. Good QBs stuck on bad teams sometimes generate massive volume with poor efficiency, because [1] they're the best player on the team, so the coaches call pass plays, [2] they're usually playing from behind, so the coaches call pass plays, but [3] the defense knows they're going to throw, so it sends in extra DBs, dials up exotic blitzes, and won't bite on play-action, and [4] the blockers and receivers aren't very good. Those seasons — "Archie Manning seasons", let's call them — are probably underrated.

Player — Career TSP — Career Value

1. PEYTON MANNING — 34,057 — 76.8

Oddly, I think Peyton Manning playing so poorly in 2015 and still winning a Super Bowl ring helps his best-ever argument more than a great season and Super Bowl would have. If Manning can play like a bum all year and still win the big one, maybe Super Bowls are a better indicator of team quality than of individual performance?

2. DAN MARINO — 31,474 — 66.9

In his first 10 seasons, Marino scored 52.0 TSP, the highest for any QB in any 10-year stretch. That includes nine seasons in a row (1984-92) with at least 1750 TSP. Peyton Manning had eight in a row (2002-09), and no one else has more than five. It seems like Marino's name comes up almost every Sunday, because — despite the new rules that help QBs and receivers, and the explosion of passing stats — Marino still holds a bunch of records. He was really incredible.

3. TOM BRADY — 26,529 — 52.7

Two years ago, Brady ranked 9th in career TSP. Since then, he's had two more great seasons (4.2 and 3.5). I stand by my subjective ranking of Brady two years ago — I think a lot of people were projecting his career forward, rather than ranking what he'd already accomplished. The most consistent feedback I got was that people felt active players were too low: I heard similar, albeit less accusatory, complaints about Drew Brees, Aaron Rodgers, Ben Roethlisberger, etc. I think a lot of folks got about two years ahead of things on all those guys. I prefer to rate active players conservatively; I'd rather adjust their rankings upward when we've had some time to establish a context for their accomplishments, rathing than make a premature proclamation of greatness.

4. JOHN UNITAS — 25,981 — 52.3

Most 2000-TSP seasons: Peyton Manning (10); Dan Marino, Joe Montana, Fran Tarkenton, Unitas (7); Drew Brees (6); Tom Brady, Otto Graham, Roger Staubach, Norm Van Brocklin (5).

5. DREW BREES — 26,083 — 51.4

Like Brady, he's still moving up the list. Subjectively, today I'd rank him 10th all-time. I think his numbers are somewhat misleading, several of the players below him have excellent postseason résumés, and he's the pretty clear third-best QB of his generation. I'm skeptical that in the 70-year Modern Era, three of the top five QBs were continuously active from 2001-15.

6. FRAN TARKENTON — 28,550 — 51.2

Tarkenton ranks 6th in career value points, but 3rd in total TSP (28,550). He never had the one really big season like Manning in '04, or Marino in '84, or Brady in '07, or Unitas in '59, or Brees in '11 — his best statistical season, in 1975, earned 5.6 points and an MVP award — but for consistently valuable play, there's no one in history who's obviously ahead of him. Tarkenton led in QB-TSP from 1964-68, 1965-69, 1966-70, 1967-71, 1968-72, 1969-73, 1970-74, 1971-75, and 1972-76 — nine five-year blocks in a row. Most 5-year leads: Manning, Tarkenton, and Unitas (9 each); Otto Graham and Dan Marino (6); Steve Young (5).

Tarkenton is probably the most underrated player in the top 10, seldom named among the very best of all time.

7. OTTO GRAHAM — 21,285 — 49.7

The constant in the TSP formula approximates replacement-level production per attempt. It prevents volume passers from dominating the yearly rankings. The constant is 2.5 throughout the 1950s, compared to 3.0 for the NFL from 1946-49, and 3.5 for the AAFC. That is not a generous approach to AAFC seasons, and Graham is the only AAFC quarterback to rate in the top three for any year from 1946-49. Most value points from the AAFC:

1] Otto Graham, 28.9
2] Frankie Albert, 5.9
3] Y.A. Tittle, 5.2

The AAFC doesn't score particularly high; it's just Otto Graham.

8. JOE MONTANA — 24,331 — 48.6

I thought squaring the yearly values might push Dan Marino ahead of Montana as the top-rated QB of the 1980s, because of Marino's monster 1984 (3688 TSP, 13.6 points), the highest-scoring season of all time. But Montana held on. He's an underrated regular-season QB. Highest-rated QB by decade:

1946-55: Otto Graham, 49.7
1950-59: Norm Van Brocklin, 30.3
1955-64: Johnny Unitas, 40.4
1960-69: Sonny Jurgensen, 32.4
1965-74: Fran Tarkenton, 34.0
1970-79: Roger Staubach, 31.7
1975-84: Dan Fouts, 37.3
1980-89: Joe Montana, 40.3
1985-94: Dan Marino, 42.5
1990-99: Steve Young, 42.6
1995-04: Peyton Manning, 33.1
2000-09: Peyton Manning, 50.4
2005-14: Peyton Manning, 43.7

9. BRETT FAVRE — 26,386 — 45.3

I think a lot of people in the analytics community underrate Brett Favre. He broke the career records for yardage and TDs, and he wasn't a compiler. He had four 2000-TSP seasons, as many as Steve Young, and he led the league in TSP three seasons in a row (1995-97). Young had Jerry Rice; Favre had Antonio Freeman.

10. STEVE YOUNG — 20,573 — 44.2

Otto Graham and Steve Young are the only quarterbacks with four exceptional seasons (2500 TSP). From 1992-94, Young scored 2652 TSP, 2650 TSP, and 3063 TSP. Those are all 7-point seasons. Young accumulated more value points in those three years than Jim Kelly did in his entire NFL career. The most similar three-year stretch is Dan Fouts from 1980-82, with at least 2390 TSP each year.

11. DAN FOUTS — 21,885 — 43.0

Led all QBs in TSP during the 1975-84 decade. He played with exceptional receivers, for a gifted offensive coach, and he doesn't have an outstanding postseason résumé, but he was the dominant QB of his generation.

12. KEN ANDERSON — 20,133 — 38.2

Anderson was essentially an elite game manager: high completion rate, low interception percentage, not really a downfield bomber. When those QBs play for great teams, like Bart Starr's Packers or Troy Aikman's Cowboys, their greatness gets recognized. When they play for second- and third-place teams — the Bengals won only two division titles in Anderson's 13 years as the primary QB, and went 2-4 in the postseason — there's a sense that the quarterback needs to do more. If you had to run a two-minute drill, you didn't want Kenny Anderson, you wanted Staubach, Stabler, Bradshaw.

In my subjective rankings two years ago, I rated Otto Graham as the second-greatest quarterback of the Modern Era. I am a strong proponent of methodological pluralism, and that's what sets Graham apart. Whatever your preferred approach, he ranks among the very best of all time. There is probably no other QB in history who excelled so dramatically by every measure we use to evaluate quarterbacks.

Anderson represents the other end of spectrum: statistical analysis is the only method of evaluating greatness that suggests he was a Hall of Fame-caliber QB. Anderson made four Pro Bowls and won an MVP, which isn't too shabby. But so did his successor, Boomer Esiason. Rich Gannon made four Pro Bowls and won an MVP. Roman Gabriel made four Pro Bowls and won an MVP. Daryle Lamonica made five All-Star Games and won two MVPs. Jack Kemp made seven All-Star Games and won an MVP. Matt Ryan has four Pro Bowls and an MVP, and I doubt many of us are ready to put him in the Hall of Fame yet. Anderson's awards and honors are consistent with a lot of guys you'd probably rank around 30-45. By other measures, Anderson fares even worse, and doesn't look like a special player.

Ken Anderson is underrated outside the analytics community, but if you rank him among the top 20 QBs of all time, you're essentially rejecting methodological pluralism, saying stats tell the whole story. I love statistics, and I find them immeasurably useful, but I would really discourage you from such a credulous approach.

13. SONNY JURGENSEN — 20,228 — 36.8

Jurgensen (1957-74) and Unitas (1956-73) were contemporaries, but Jurgensen's TD/INT differential (+66) is substantially better than Johnny U's (+37). If you rate Unitas in the top 10 — which you should — you need to rate Jurgensen in the top 20, at least.

14. JOHN ELWAY — 22,258 — 36.7

In 1987, Elway led the NFL in QB-TSP, 2582. With Vance Johnson, Ricky Nattiel, and Clarence Kay as his leading receivers.

15. NORM VAN BROCKLIN — 19,204 — 35.2

Van Brocklin is TSP's highest-rated QB of the 1950s, 30.3. That's the lowest high-score for any decade. Furthermore, if you sum the top five for the '50s, the total comes to only 103, which is also the lowest of any decade, followed by 1946-55 (111) and 1955-64 (112). The top 10 for the '50s sums to only 142, which is the lowest of any decade, again followed by 1946-55 (154) and 1955-64 (170). The '50s featured two Hall of Famers in their primes (Van Brocklin and Bobby Layne), as well as prime years from three others (Graham, Unitas, and Tittle), and Charlie Conerly is sometimes mentioned among the greatest quarterbacks of all time.

I know some people refuse to consider older players in their all-time rankings. I think that's misguided, and — by definition — not an all-time ranking. But it's true that passing is more important now than it was in the past. QB-TSP reflects that: players score higher today than they did in previous decades.

16. Y.A. TITTLE — 19,087 — 35.0

By this method, still the greatest quarterback in New York Giants history. He scored 18.7 in four seasons with the Giants, including back-to-back years with over 2700 TSP (joining Dan Fouts as the only players to do so). But tenure is so critical to team identity, I prefer accumulated TSP to value points when determining franchise leaders. Here are the Giants' top five, through 2016:

1. Phil Simms — 12,333
2. Eli Manning — 11,831
3. Charlie Conerly — 11,431
4. Fran Tarkenton — 8,995
5. Y.A. Tittle — 7,263

I suspect that more closely approximates how Giants fans feel: Simms and Eli pretty close at the top, with Conerly also in the mix for older fans. I don't think Giants fans want to hear that their best QBs were a Viking and a 49er — and I'm inclined to agree with them.

17. AARON RODGERS — 16,976 — 34.7

One of the nice things about this system is that it's easy to calculate a player's climb up the leaderboard. One more great season (2000 TSP, 4 pts) would vault Rodgers to 12th all-time.

18. ROGER STAUBACH — 15,770 — 31.7

Most years leading the league in QB-TSP:

1] Otto Graham, 6
2] Peyton Manning, 5
t3] Roger Staubach and Steve Young, 4

Staubach was a brilliant QB in a short career, essentially just eight seasons. Why does he rank below similar players like Graham (10 years), Young (8 years), and Rodgers (9 years)? Staubach didn't have exceptional seasons. His highest TSP was 2294, and no major organization ever named him first-team all-pro. Paul Zimmerman, an admirer of the '70s Cowboys, lamented in The New Thinking Man's Guide to Pro Football, "Every time Roger had a good year, someone else picked that time to have a slightly better one." TSP sees it the other way around, but it's true that Staubach never posted the statistical production that normally suggests a first-team all-pro.

Staubach was great, and his fine postseason record doesn't count toward TSP, but he wasn't as outstanding as his four years leading the league would imply.

19. WARREN MOON — 19,322 — 31.0

20. JOHN BRODIE — 17,839 — 30.7

I would take Warren Moon's consistency over Brodie's erratic excellence — the 1961, 1965, and 1970 seasons account for over 60% of his 30.7 value points — with zero hesitation. Moon had more 1000-TSP seasons, more 1200-TSP seasons, more 1400-TSP seasons, more 1600-TSP seasons... Moon was good every year, and he did have a 2500-TSP season in 1990. But as it stands right now, the value formula overrates seasons like Brodie in '65 and '70, and underrates seasons like Moon in '87 and '97. There's a comparison to be drawn between Ken Anderson and Brodie.

21. PHILIP RIVERS — 15,819 — 26.2

I think there's a growing sense, in both mainstream and analytic circles, that Philip Rivers has had or is rapidly approaching a Hall of Fame-caliber career. I believe his HOF chances are directly linked to Eli Manning's. If Rivers and Roethlisberger are elected, the voters aren't going to leave Eli out, or at least not forever. Eli's induction would probably boost Rivers' chances as well, if it comes to that, but this was the first year I got the sense that Rivers might have a better chance of enshrinement than Eli.

22. BOOMER ESIASON — 15,881 — 25.8

23. ROMAN GABRIEL — 15,982 — 24.0

24. JOHN HADL — 15,238 — 23.24

25. JIM KELLY — 15,425 — 23.18

26. BOBBY LAYNE — 15,510 — 23.17

27. JIM EVERETT — 13,937 — 23.1

28. KURT WARNER — 12,202 — 22.8

Beginning in 1998, Kurt Warner's TSP by year: 1, 2730, 1770, 2381, -62, -57, 381, 751, 312, 1176, 1856, 963. That's two exceptional seasons, two very good seasons, two essentially average seasons, and six years that don't really add anything to his legacy. Is that a Hall of Famer? Maybe, but I understand why people see it both ways.

29. JIM HART — 16,076 — 22.5

30. BEN ROETHLISBERGER — 16,024 — 21.5

31. JOE NAMATH — 13,253 — 21.3

32. MATT RYAN — 12,753 — 20.9

As I mentioned at the beginning, the career values are derived using a ^2 exponent, when the ideal exponent is probably something like 1.85. That means great seasons are overrated, while average and good seasons are underrated. Normally this isn't a big deal, but there are a few players, and types of player, for whom it distorts their rankings.

Matt Ryan's 2016 season scores higher (6.8) than Joe Flacco's career (6.1). That seems excessive to me. Flacco scores 500-1050 TSP in seven of his nine seasons; such seasons are somewhat underrated, I think. I alluded earlier to Archie Manning seasons and the Bob Griese problem. Let's add a couple more:

Joe Flacco seasons — 500-1050 TSP, classic compiler seasons. All-time leaders: Vinny Testaverde (10); Dave Krieg, Carson Palmer (8 each); Joe Flacco (7).

Ben Roethlisberger seasons — 1100-1600 TSP, good years that score a fraction of the value of great years. All-time leaders: Ben Roethlisberger (10); John Elway, Bob Griese (8 each); Jim Kelly, Donovan McNabb, Phil Simms (7 each).

If I used a ^1.85 exponent, Krieg would rise four spots, Simms three, everyone else two or fewer. So the difference isn't enormous, but it's there. Flacco would pass Kirk Cousins (6.3), which I think I agree with for now. My sense is that the ^2 adjustment works well in the top 25 or so, ^1.85 is better for 26-75, and something like ^1.75 the best below that. Around the border of the top 100, one-year wonders are competitive with compilers, which seems wrong to me. Someone who held a starting job for a decade was better than someone who fluked into one great season. With the ^2 modifier, Scott Mitchell (2376 TSP in 1995) ranks ahead of Michael Vick, Steve Beuerlein, Jim McMahon, and Neil O'Donnell. At ^1.85, he doesn't.

Archie Manning seasons — Above-average QB on a bad team, with high volume and low efficiency. Examples: Archie Manning, Drew Bledsoe, Tobin Rote, Norm Snead.

Bob Griese seasons — Good QB on a run-oriented offense, with low volume and high efficiency. Examples: Bob Griese, Troy Aikman, Ben Roethlisberger, Bart Starr.

The issues regarding Joe Flacco/Ben Roethlisberger seasons are purely mathematical; I could "fix" those if I wanted. I didn't, because the top of the list is more important to me than the middle (and because I can square values in my head, but ^1.85 requires a calculator; simplicity and accessibility are more important to me than the illusion of precision). The Archie Manning/Bob Griese problem is a little dicier, because it's inside the TSP formula; it involves the balance of efficiency vs. production. Furthermore, helping one group would hurt the other. And a mathematical formula can't tell the difference between, say, Norm Snead in 1961 and Joey Harrington in 2003, or Bob Griese in 1973 and Alex Smith in 2011. You can't differentiate those seasons with math, or at least not in any way I'm aware of. You have three choices: conclude that players like Griese were wildly overrated, conclude that players like Smith are wildly underrated, or look beyond statistics and apply subjective adjustments when evaluating these players.

This is an area where I differ from many statistical analysts. Football is such a complicated sport, and quarterback such a complicated position, that I don't believe stats can ever tell the full story. So when I consult a statistic, I assume it to be imperfect. It gives me a frame of reference, a starting point, from which to ask, "What's missing? What is distorted or misrepresented?" This is why I find a methodologically pluralistic approach to be of such importance.

I don't want to misrepresent my position on QB-TSP. I have been refining this formula for over a decade, and I think it's the best stat out there for evaluating QBs. I would caution readers, though, to be judicious in using stats — any stat — to make sweeping judgments about pro football. I've done a subjective ranking that incorporates not only stats, but also a lifetime of research on pro football history, to create a more cohesive evaluation of great quarterbacks. If that's what you're looking for, please read that series, not this article. This isn't a list of the greatest QBs, per se.

33. BERT JONES — 10,551 — 20.6

34. RANDALL CUNNINGHAM — 12,557 — 20.4

35. TONY ROMO — 13,052 — 20.0

36. DARYLE LAMONICA — 10,940 — 19.4

37. RICH GANNON — 12,258 — 19.21

38. DONOVAN MCNABB — 13,717 — 19.19

39. TROY AIKMAN — 13,375 — 18.9

40. JEFF GARCIA — 11,605 — 18.68

41. TRENT GREEN — 11,087 — 18.67

42. LEN DAWSON — 14,225 — 18.4

Does TSP underrate AFL seasons? Certainly Dawson ranks lower than one would expect. In 1962, when The Sporting News named Dawson AFL MVP, he scored only 1646 TSP. That year, the constant that penalizes attempts and sacks was -2.5 for the NFL, but -3.033 for the AFL. The highest-scoring QBs in AFL history were Dawson, Namath, Hadl, and Lamonica, all bunched together between 13.3 - 14.6. Hadl, with significant production in the '70s as well, has the highest career mark, 23.2.

TSP rates Hadl (24th), Namath (31st), and Lamonica (36th) about where knowledgeable fans would expect. Rather than a systemic error in TSP's treatment of AFL quarterbacks, this suggests a problem particular to Dawson. He had a long career and impressive counting stats, but he didn't have big seasons. His passer rating was fantastic, but his sack and fumble rates are among the worst of all time. Of the top 100 QBs in TSP, Dawson ranks 88th in sack percentage (9.3%) and 92nd in fumble rate (1.92%). Don Meredith is the only player worse in both categories. Everything that doesn't show up in the passer rating formula, Dawson was below average. I don't believe Dawson is substantially overrated; he's in my top 30. But I do believe that a casual look at his statistics makes him look much more accomplished than a careful look does. At this point in my career as an analyst and historian, I have little doubt that Joe Namath was better than Dawson.

43. TERRY BRADSHAW — 13,342 — 18.1

Bradshaw, like Bart Starr, is a Hall of Famer largely because of his fantastic postseason performances, which don't count here. Any regular-season ranking inevitably underrates him.

44. DAUNTE CULPEPPER — 8,772 — 17.6

Culpepper only had five seasons in which he scored at least 175 TSP. You can view him as a fluke who had a few good seasons, playing mostly with Randy Moss, or as a Bert Jones-type whose career never fully recovered following an injury. I thought Daunte was a good QB, but not as outstanding as his stats would imply, so I lean more towards the first theory.

45. CARSON PALMER — 12,917 — 17.5

46. JOE THEISMANN — 11,069 — 17.33

47. NORM SNEAD — 12,784 — 17.32

48. BART STARR — 13,829 — 17.30

How on earth does Norm Snead have a higher rating than Bart Starr?

Starr had a lot of seasons that the system reads as compiling, and Snead only had one or two. I'm not sure fans realize how pedestrian Starr's regular-season stats usually were. Consider 1960, when the Packers made their first NFL Championship appearance under Vince Lombardi. Starr passed for just 1,358 yards and 4 touchdowns. Yeah: four. He threw 8 interceptions. TSP scores that season at 875, a little below average, which doesn't seem especially unfair to me.

Well, that's one season. But it's not atypical for Starr. He passed for at least 2,000 yards only five times, never for 2,500. His passed for double-digit TDs just seven times, and his career high was 16. He had a very short prime, basically just 1961-66. Look, Bart Starr was a great quarterback. He doesn't have great stats.

Norm Snead passed for 2,000 yards eight times, including four years over 2,500. His passed for double-digit TDs 10 times, and his career high was 29. He had 1700-TSP seasons a decade apart, 1962 and '72, and was a productive starter in at least eight seasons.

Snead threw way too many interceptions (5.9%), and Starr was the best of his generation at avoiding INTs (4.4%). Snead's INT% is about one-third higher than Starr's, a massive difference. But because of the way we use statistics, QBs who throw interceptions are underrated, and QBs who hold the ball, leading to sacks and fumbles, are overrated. Snead's sack rate was about 7.5%, with a 1.2% fumble percentage. Starr's sack rate was about 10%, with a 1.7% fumble percentage. His sack% is about one-third higher than Snead's, and his fumble rate is almost 50% higher.

Even considering sacks and fumbles, Starr was a more efficient quarterback than Snead. But Snead was far more productive than Starr. Other than 1966, Starr was probably never the best player on the Green Bay offense. Jim Taylor and Paul Hornung were the motor, not Starr. Snead often was the best player on his offense, and TSP values his productivity, as I believe it should. Starr (1956-71) and Snead (1961-76) played at more or less the same time, and they made an equal number of Pro Bowls (4 each); I'm actually not proposing anything too radical here.

I don't want to give anyone the wrong idea. Bart Starr was a better quarterback than Norm Snead, and it's not close. But depending on how you balance volume and efficiency, their regular-season stats are very close.

49. VINNY TESTAVERDE — 14,297 — 17.1

His 14,297 TSP ranks 29th all-time, but he only had four average (1000 TSP) or better seasons.

50. STEVE MCNAIR — 12,716 — 17.0

He's underrated by this method. McNair was an excellent runner, he struggled with injuries in some of his best seasons, and he had good years that don't score a lot of value points. The Titans went 13-3 back-to-back seasons in 1999 and 2000, and McNair scores just 2.5 for those seasons combined. That strikes me as conservative. To some extent, it's a Bob Griese problem, since Tennessee's offense ran first and foremost through Eddie George.

51. MILT PLUM — 9,869 — 16.9

52. KEN STABLER — 11,234 — 16.4

53. CHARLIE CONERLY — 11,431 — 16.1

54. MARK BRUNELL — 12,138 — 15.9

55. DREW BLEDSOE — 12,406 — 15.4

56. CHARLEY JOHNSON — 11,910 — 15.33

57. NEIL LOMAX — 9,279 — 15.27

58. BERNIE KOSAR — 9,914 — 15.21

59. BRIAN SIPE — 9,456 — 15.19

60. BILLY WADE — 10,535 — 15.1

* SAMMY BAUGH — 7,027 — 14.8

This is where Baugh would rank based on 1946-52. That time frame excludes most of his best seasons.

61. PHIL SIMMS — 12,333 — 14.7

62. ELI MANNING — 11,831 — 14.6

At this rate, he may never reach the top 50.

63. TOBIN ROTE — 9,153 — 14.4

64. CRAIG MORTON — 12,033 — 13.4

65. BILLY KILMER — 11,407 — 13.3

66. STEVE DEBERG — 11,424 — 13.2

67. EARL MORRALL — 10,848 — 12.92

68. BOB GRIESE — 12,311 — 12.88

It's incongruous to find him below his backup with the Dolphins, Earl Morrall. I don't believe this ranking reflects Griese's quality or accomplishments as a quarterback, but he didn't have any statistically outstanding seasons. Slow-but-steady is underrated by this methodology, which rewards big years more than consistency.

69. JOE FERGUSON — 10,288 — 12.72

70. STEVE GROGAN — 11,946 — 12.68

71. MATT HASSELBECK — 9,710 — 12.52

72. RON JAWORSKI — 10,309 — 12.49

73. DAVE KRIEG — 12,590 — 12.48

74. MARK RYPIEN — 8,023 — 12.3

75. JEFF GEORGE — 9,067 — 12.2

76. DON MEREDITH — 8,336 — 12.0

77. GEORGE BLANDA — 9,433 — 11.9

A Hall of Fame kicker; a backup QB who had one exceptional season in a marginal major league.

78. RUSSELL WILSON — 7,602 — 11.8

79. FRANK RYAN — 8,929 — 11.743

80. CAM NEWTON — 7,942 — 11.738

I named Cam Newton MVP in 2015. These are the top four in QB-TSP that season:

Carson Palmer, 2235
Tom Brady, 2038
Russell Wilson, 1962
Cam Newton, 1953

Did I get suckered into Newton's narrative, subverting what TSP told me about Palmer and Newton? I don't think so. It was a very close call — I wouldn't argue with any of the four — but I thought Newton accomplished the most relative to his support from teammates: "Working with an undistinguished group of receivers, Cam made plays you have no business making in the NFL, like the ghost of Randall Cunningham."

Football Perspective's Chase Stuart put it well in preseason: "After Greg Olsen, the Panthers top targets are second round draft pick Devin Funchess, Corey Brown, Jerricho Cotchery, Ted Ginn, Jr., and Ed Dickson. So when the Panthers go 6-10, it's because Cam Newton isn't a true leader." They didn't go 6-10, of course; they went 15-1 and led the NFL in scoring. That wasn't because Funchess turned into A.J. Green; it was because Newton played like Aaron Rodgers. Newton passed or ran for 45 of the team's 54 offensive touchdowns. In context, I thought Newton was more impressive than Palmer, and created the most on his own.

81. BRAD JOHNSON — 9,493 — 11.6

82. GREG LANDRY — 8,041 — 11.44

83. MATTHEW STAFFORD — 8,033 — 11.39

84. TOMMY KRAMER — 9,152 — 11.1

85. DOUG WILLIAMS — 8,231 — 11.0

86. JIM ZORN — 8,229 — 10.79

87. KEN O'BRIEN — 8,342 — 10.76

88. STEVE BARTKOWSKI — 8,020 — 10.72

89. JOHNNY LUJACK — 6,022 — 10.69

90. DANNY WHITE — 8,631 — 10.4

91. JAKE PLUMMER — 7,968 — 10.3

92. ARCHIE MANNING — 7,139 — 10.1

93. MATT SCHAUB — 7,549 — 10.0

94. ED BROWN — 7,826 — 9.9

* TOMMY THOMPSON — 5,365 — 9.7

Like Sammy Baugh, he's rated only on what he did from 1946 onwards. Unlike Baugh, that mostly tells the story of his career anyway.

95. BOBBY HEBERT — 8,163 — 9.5

96. KERRY COLLINS — 9,029 — 9.2

97. CHRIS CHANDLER — 8,533 — 9.1

98. LYNN DICKEY — 6,713 — 9.0

99. CHAD PENNINGTON — 6,267 — 8.8

100. ANDREW LUCK — 5,909 — 8.5

101-110: Jim Plunkett, Jay Cutler, Bill Kenney, Marc Bulger, Jim Harbaugh, Scott Mitchell, Michael Vick, Steve Beuerlein, Andy Dalton, Bill Nelsen.

It surprised me to see Andrew Luck (8.53) and Andy Dalton (8.09) separated by less than half a point. But I think that's a function of popular narrative. Luck has been hyped and beloved since Day One, with Dalton the red-headed stepchild. Dalton has more passing yards than Luck, and more net yards per attempt. He has more passing TDs and a higher passer rating. He has more rushing TDs and fewer fumbles. Luck had a great season in 2014, and he rates ahead because of that, but the gap is not large.

Contents copyright © Sports Central 1998-2017