That’s usually why I avoid QBs coming off career years. It’s generally counter-intuitive to believe a player coming off a career year will be undesirable, like Gary Barnidge this year. But career years are called such for a reason: They’re almost never duplicated. These are cases of something unusual happening, which usually means the opposite will occur in the near future.

-- 2016's Lessons Learned

Two weeks ago, I started looking into QB "career years". Then, I wrote about what happens to a QB's fantasy numbers the season after a career year. But of course that is only useful if we can actually identify a career year when it occurs. This week, I'm going to start to answer my second question from the previous article: how do we know what just happened WAS a career year?

The previous article had a detailed definition of what I mean by "career year." Here's a quick synopsis:

  • It was a QB's best season in BOTH total FP and FP/G, minimum 200 attempts, 25 attempts per game, and 7 games.
  • It only includes QBs who ended their careers before 2016.
  • Only seasons after the merger are included and strike-shortened seasons (1982 and 1987) don't count.
  • The career year had to be ranked in the Top 12 in FP/G for that season's QBs.

The scoring system I'm using is 25 pass yards = 1 FP, Pass TD = 4 FP, 10 rush yards = 1 FP, Rush TD =6, no turnover penalty.

I found 9 career seasons under these criteria, but dropped 4 from the study as not being fair comparisons to the other 94 QBs:

  • Roger Staubach, retired after his career year in 1979 at age 37.
  • Bobby Hebert, did the same after a career year at age 36.
  • David Garrard, lasted one more year but didn't play.
  • Randall Cunningham, who lasted less than a game into his post-career-year season.

Here's how the 94 career years stack up by age compared to all Top 12 seasons since 1970: 

Relationship of Age to Career Years in Top 12 Fantasy QBs

Age

Total
Top 12 Finishes

Non-career Year

Career Year

% of Career Years

Cumulative% of Career Years

22

5

5

0

0%

0%

23

15

10

5

5%

5%

24

25

22

3

3%

9%

25

42

27

15

16%

24%

26

41

28

13

14%

38%

27

33

24

9

10%

48%

28

29

20

9

10%

57%

29

40

36

4

4%

62%

30

35

26

9

10%

71%

31

34

26

8

9%

80%

32

27

26

1

1%

81%

33

24

20

4

4%

85%

34

23

16

7

7%

93%

35

18

16

2

2%

95%

36

14

12

2

2%

97%

37

10

7

3

3%

100%

38

7

7

0

0%

100%

39

1

1

0

0%

100%

40

1

1

0

0%

100%

41

1

1

0

0%

100%

Age, the first column, is defined as the player's age as of December 31st in the season in question. The 2nd column is how many Top 12 finishes were recorded at a given age, for example, QBs aged 22 years have posted five Top 12 fantasy seasons since the merger. Of those, all five turned out to be non-career years; at age 23, 10 such years weren't "career" and 5 were (3rd and 4th columns). Those 5 career years at age 23 were 5% of the total career years (5th column). And the 6th column tracks the percentage of career years posted at that age or younger, also 5% at age 23. By age 28, 57% of all career years have occurred; by age 31 (Matt Ryan in 2016), it's up to 80%. So while Ryan's age doesn't mean he won't have a better season in 2017 or later, it's also not that common of an occurrence.

Note that no QB has ever posted a career year after age 37. I found age to be one of the strongest indicators of when a career year has occurred.

The next table summarizes the previous one by age groups: 

Career Year Age

Career Year

% of Career Years

22 to 24

8

9%

25 to 28

46

49%

29 to 31

21

22%

32 to 34

12

13%

35 to 37

7

7%

Only a small proportion of career years happen before QBs turn 25. Of those 8 career years, 6 were by "running" QBs (defined last time as QBs who got over 20% of their fantasy points from running the ball). Only Dan Marino in 1984 and Dennis Shaw in 1970 had their best fantasy seasons before age 25 as "pure" passers. You'll notice that both those instances were a long time ago.

About half of all career years occur between ages 25 and 28. This is a little misleading, inflated a bit by using a bigger spread of ages than in the other 4 categories. But if you look at the first table, there seems to be a break in the frequency of career years between ages 28 and 29, so I drew the line there. The age 29 number could just be a statistical fluke. Over two-thirds of all career years occur between ages 25 and 31.

From age 32 on, the frequency of career years drops rapidly – and remember, only very good-to-great QBs keep playing regularly into their late 30s, so the post-32 group is inflated by survivorship bias.

How well a QB did in his career year was a weak indicator:

Career Year Rank

Career Year

% of Career Years

Top 4

39

41%

#5-8

32

34%

#9-12

23

24%

It's not surprising that Top-4 finishes make up a larger share of career years: it's hard to be that good and even harder to do it again. But given that, 41% of all career years isn't that much bigger of a share than 34% for QBs who ranked 5th to 8th. And it's striking that about a quarter of all career years were no better than 9th for fantasy.

I then looked into individual stats to see if I could spot other markers of a career year. I looked at three "counting stats" (attempts per game, yards per game, and total TDs – and yes, "per game" is not a pure "count"), and three "rate stats" (TD% per attempt, yards per attempt, and completion percentage). To post a career year, a QB must have either opportunity, production, or efficiency. Pass attempts per game equate to opportunity (at least for non-running QBs). Yards and TDs are production, especially for fantasy purposes. And the three rate stats are those that directly relate to efficiency for fantasy: improving any of those three stats will result in more yards or TDs, and hence fantasy points.

What I wanted to see was if career years correlated strongly with any of these six numbers. In particular, I wanted to see if QBs had large jumps in any of these stats from their previous maximum in that category.

The way I measured this was to divide each QB's stats in the Top 12 seasons by their previous career high in that category. For example, in 2014 Peyton Manning threw for 39 TDs. His previous best was 55 (the year before, which was his career year). So for his 2014 season, Manning threw for 71% of his previous career best. But in 2013, when he completed those 55 TD passes, it was up from his previous high of 49 (2004), or 112% of his previous "career" number.

I was looking for percentages that I could use to say were markers or indicators that a career year had occurred. For example, did a lot of these career years correspond to a 50% increase in pass attempt per game over a players previous career high?

I think this table shows that I wasn't obviously successful:

Percent Change on Previous Maximums in Statistical Categories by Top 12 Fantasy QBs

Statistical Category

Median Change

Non-CY

CY

Att/G

96%

108%

Yd/G

94%

117%

Total TDs

83%

137%

TD %

79%

96%

Y/A

92%

98%

Comp %

97%

98%

I used medians instead of means for my "average" stat to avoid skewing the data with outliers (and avoid the statistical perils of averaging ("mean"ing?) percentages). Generally, QBs in career years (CY) did have career highs in my "counting stats" while non-career year (non-CY) did not. But note that those non-career year medians imply a LOT of cases where those players did have one or more counting stats higher than their previous best total. And the rate stats were not very indicative of career years at all: on the whole, in the median career years QBs posted TD%, Y/A and completion percentages LOWER than those their previous bests. Said another way, more career years occurred when a QB had a LOWER TD%, Y/A, or completion % number than his previous best year in that category. 

(For these numbers, I had to throw out the rookie seasons since I couldn't compare them to a previous career high; I also discarded seasons when a QB technically wasn't a rookie but had played sparingly, if at all, in his first year. For example, Brett Favre in 1992 and Kurt Warner in 1999 had 4 and 11 attempts respectively in the previous year; while they were in their 2nd seasons in 1992/1999, it would be misleading to use those seasons in these calculations. So there are only 86 career years included in the table above)

Say I want to use Att/G above the median percentage (108%) as an indicator that a career year has occurred. While that will only capture half of the career years (by definition of "median"), at least it would be a starting point to differentiate those years from the non-career variety. Unfortunately, a lot of non-career year QBs had also Att/G above the CY median. This table shows the problem:

Number of Seasons Above the CY Median Percentage Change

Type

Att/G

Yd/G

TD

TD %

Y/A

Comp %

Non-CY

63

53

44

82

97

137

CY

44

44

43

43

41

42

In every case, there were more QBs' non-career years above the defined cut-off: 63 had Att/G over the CY median, compared to 44 for the actual CY QBs. This meant that using any of these stats would give me more "false positives" than actual career-year markers. (This was true if I used other cut-offs based on standard deviations or other statistical measures, I stuck with medians for this article).

A bit more useful was identifying the minimum stat changes for career year performances: 

Percent Change on Previous Maximums in Statistical Categories by Top 12 Fantasy QBs

Statistical Category

Minimum Change

Non-CY

CY

Att/G

55%

75%

Yd/G

45%

83%

Total TDs

6%

81%

TD %

22%

27%

Y/A

49%

48%

Comp %

64%

73%

Obviously, all 86 remaining career years were at or above the minimums. But I could screen out more than half of the non-career year performances by using these numbers: of the 323 non-career years left in my population, only 149 were above all six of the minimums for career years (really five, since all non-career year QBs had better Y/A changes than the career year minimum – 49% vs. 48%). For example, we can be fairly certain that a Top 12 QB who posts a Yd/G number that is only 80% of his previous career best did NOT have a career year, since no career year since 1970 has been under 83%. Records can be broken, so I won't say I'd be certain, but it is a strong marker that a season was NOT a career year if it falls below any of the minimum percentages for these stats. Eliminating the non-career years may not be as good as identifying the actual career years, but so far it's the best I can do.

Finally, what if I combine three things I've talked about: age, median changes from previous best stats, and percent changes below the career year minimums. For the next table, I first eliminated QBs under 25 and over 31. That excludes some career years, but looks at 80% of them. Then I eliminated all the Top 12 QBs who were below the minimum percent changes discussed in the previous paragraph: for example, all Top 12 QBs with Att/G below 75% of their previous career high, below 83% of their previous best Yd/G, etc. Then I counted the QBs above the median change for career years: more than 108% of Att/G, 117% Yd/G, etc. 

Number of Seasons Above the CY Median, QBs Aged 25-31

Type

Att/G

Yd/G

TD

TD %

Y/A

Comp %

Non-CY

42

32

29

40

50

56

CY

37

40

39

39

38

38 

For most of these stats, there are still more false positives (non-CY) than actual career years identified. But Yd/G and TDs standout as stronger indicators, or at least more frequent than for non-career years. So QBs in the age 25-31 "sweet spot" for career performances who post Yd/G better than 117% of their previous high, or TDs more than 137% of their best season are more likely than not having a career year.

If you've made it this far, and aren't totally confused, congratulations! I've gone pretty far in the weeds with some pretty esoteric percentages (the stat work itself is rudimentary, if you're confused it's the writing, not the numbers). Next week, I'll calculate the numbers for the 2016 Top 12 QBs and use what I've talked about here to take a stab at identifying who had a career year. And I'll compare how post-career year performances (from last week's article) compare with the following year fantasy numbers for Top 12 QBs who did NOT have a career year.