Free: Rethinking NFL Age Curves for WRs – The Wrong Read, No. 33
Originally published on April, 18, Rethinking Age Curves is part of our Memorial Day weekend free look at some of the best of RotoViz.
Welcome to the 33rd installment of the “The Wrong Read.” This article series started as one that reflected on recent podcast episodes and extended the ideas discussed there to logical conclusions with broader applications. Since then it’s become a space for me to write about whatever I want, with irregular references to various podcast episodes. Nevertheless, I’ll link to the episode that started my train of thought if applicable.
Over the last few weeks I’ve been examining the effects of draft age on career NFL production.1 The results have been illuminating, to say the least. Although we knew age was important before, even I was surprised at just how important it appears to be.
This got me thinking, though: what effect does age have on production after entering the NFL? That is, how does an NFL player’s current age affect production? Rich Hribar has done some foundational work on age curves at each position. It’s often reported that he found all positions tend have specific peak ages — for most positions it’s somewhere between the ages of 25 and 28.
Age Curves and the Elusiveness of Peak Production Age
However, it’s important to understand how to interpret these age curves before using them to evaluate players for dynasty. Taking wide receiver as an example, Rich looked at each season meeting certain thresholds since 1970 (top 36, top 24, etc.) and broke those down by age. So, for instance, about 14 percent of the top-12 WR seasons since 1970 were played by WRs who were 27 years old at the time. His findings give rise to the concept of an age curve, as a quick look at his graph will attest:
Dynasty (and redraft) players often look at this curve, find the age of a specific player on the x-axis, and then try to determine whether he is past the age at which WRs tend to reach their peak production. If he is, typical dynasty owners might try to sell. If not, they will try to buy. The problem is, this curve isn’t measuring a typical WR career arc.
Let me put my point bluntly, lest any confusion arise: using age curves to make buy and sell decisions in this way is wrong. That is not to say that the curves themselves are wrong. Rather, the idea that an individual player’s career tends to fit this curve is wrong. To see my point, here are actual career arcs for a random(ish) sample of fantasy-relevant receivers since 2000:
Few of the players here experienced a either a gradual increase into their prime or a gradual decline after reaching peak production. The careers that do resemble curves, namely, Vincent Jackson’s and, to a lesser extent, Reggie Wayne’s, do not exhibit what we might call a peak age. Instead they show a slow increase with many years of sustained productivity before a gradual decline.
But even this career arc is relatively rare. Rather, most reached their peak production quickly (often in their rookie seasons), and then maintained it for a few years before falling off a cliff — either going straight to zero (injury or retirement), or to well below their career averages before going to zero the next year. Players like Steve Johnson, Calvin Johnson, Anquan Boldin, and Marques Colston ended their careers while they were still producing at or near their peak level of production. Players like Deion Branch and Lee Evans stopped playing after one year of production well below their career averages. Aggregating all of these plateaus and cliffs into a single line is what creates a curve.
That point bears repeating in different words: the age curve you see above is partly the result of aggregating production across multiple player seasons. Again, that’s not to say that it’s wrong or that it doesn’t provide extremely valuable information. It is to say that if you are applying that age curve to individual player evaluations and making buy and sell decisions on that basis, you’re misinterpreting what the age curve is communicating. And you’re also probably making poor dynasty decisions. In reality, you should probably be selling your dynasty assets a lot sooner than you think.
How Does Age Affect Production?
So then how should we value age in dynasty? Or, to put that another way, how does age impact a player’s expected production? The way I’ve chosen to try to get at this problem is to measure how fantasy production changes from one year to next.2 How much can WRs who are, say, 21 years old expect to see their production increase (or decrease) in the next season? The chart below shows my findings:
This data includes every WR season from 2000 to 2016. League-wide, 21-year-old WRs saw their fantasy scoring increase by almost two points per game the next season. Twenty-two-year-old WRs, as a group, saw their scoring increase by about half a point per game. For WRs older than 23, fantasy scoring tended to decrease in the following year. These numbers are irrespective of years in the league. So the average 23-year-old rookie WR will never outproduce his rookie season.3
Age Curves and Failure Rate
The way to read the above chart is to realize it’s not picking out a peak age of production, or predicting that all WRs at age 23 or older will see their production decline. Like Rich’s charts, it too is an aggregation of multiple WR seasons. Rather, it’s important to note that these numbers are the result of what are more or less a series of binary outcomes: either WRs are successful, or they aren’t.4 After age 22, more WRs fail than succeed.5 So the decline from age 23 is the result of the increasing rate of failure at older ages. Here’s another perspective on the same data:
Failure here equals zero fantasy points per game. In other words, we’re looking at only total failures — utter fantasy death. Whereas less than 10 percent of 21-year-old WRs have zero fantasy points in their next season, by the time we get to age 24 that number is above 30 percent. Every age above 22 sees WRs fail completely at a rate higher than 20 percent. And older WRs generally fail more often than their younger counterparts.
This also explains why Rich’s chart appears to show peak WR production around age 26 or 27: by the time you get to age 27, if you’re not talented, there’s a good chance you’re already out of the league. So that means that WRs in the NFL around age 27 almost by definition have the requisite skills to be a fantasy-relevant WR. It’s no wonder, then, that the ages between 25 and 28 dominate the top-36, top-24, and top-12 ranks in the fantasy world. We already know that most of the players who are still in the NFL at those ages are good.
To turn this into something actionable, the optimal strategy would likely be to buy the older survivors when they are cheap — i.e., past the so-called peak production window6 — and to buy young players hoping they become survivors, while (perhaps more importantly) selling those who will fail before they have failed (which may often mean selling during the so-called peak production window or even well before it).7 Shawn Siegele’s work on WR breakouts goes some way toward explaining how to identify those who are likely to survive and distinguish them from those who will likely fail.
Failure Rate and Draft Position
One discovery worth noting here is that the rate of failure looks very different for early-round picks than it does for late-round picks.
Basically, early-round picks are given much more early opportunity, not only in the form of snaps or targets but also in the form of opportunities to prove that earlier poor seasons were the result of bad luck rather than a lack of talent. Late-round picks are, paradoxically, held to much higher standards while getting less opportunity to show what they can do. Therefore, late-round picks bust at nearly a 40 percent rate by age 24. Early-round picks do not bust at even a 20 percent rate until reaching age 30.
What’s interesting to note, though, is how the failure rates for early-round and late-round picks start to converge around age 24 and actually cross by age 30. Here’s what I think is going on: If you were an early pick who didn’t break out, your likelihood of busting continues to rise with every year. Early-round picks, as mentioned, tend to get offensive snaps much earlier than their late-round counterparts, and if they are consistently unable to perform with those snaps, before long their careers are over.
If, on the other hand, you were a late pick who’s hung around the league for a few years without getting many snaps, the likelihood that you will be able to break out later in your career increases slightly. So late-round picks see their rate of failure decrease between the ages of 24 and 28.
Most of this is due to survivorship bias. If you’re a late-round pick who’s managed to stay in the league despite not getting many snaps, that’s potentially a positive signal. If you’re a late-round pick who lacks the talent to contribute in the NFL, you’re probably already out of the league by age 25. But of course we can use this to our advantage. These late-round survivors are usually free in dynasty leagues before they break out. And, especially if we have reason to believe they might actually be talented, they could make for high-upside, end-of-roster options.
Takeaways for Fantasy
There’s still a lot to unpack here, but my initial attempt at a key takeaway would be to say that you should be selling players in dynasty much earlier than you probably currently are. And you should also be buying players much earlier than you probably currently are. Buying players when they are in or leading up to their so-called peak-age window and holding them throughout is about the worst thing you could do for your dynasty team; it’s suboptimal not only from a value standpoint but also from a fantasy-scoring standpoint. You would be holding an asset that is decreasing both in trade value and production value. Better to buy 21- and 22-year olds, who at least tend to see an increase in production.
- This exploration was inspired, originally, by RotoPat’s appearance on RotoViz Radio, and then re-invigorated by John Terry’s appearance on the High Stakes Lowdown. (back)
- Josh Hermsmeyer looked at this using a slightly different method. Although he draws a different (but not inconsistent) conclusion from the data, using both analyses in concert yields a coherent picture that provides actionable information for dynasty owners. (back)
- This fits with what Pat Kerrane discovered regarding trade value: WRs are never more valuable than at age 21. What my analysis adds to his is the realization that 21-year-old WRs are probably being valued correctly in dynasty. Pat made a helpful distinction in his piece between winning the value game and winning the production game. But my findings indicate the two might not be mutually exclusive. (back)
- That’s a gross oversimplification, but in terms of explaining basically why the numbers look the way they do, it will suffice for now. (back)
- Or at least the degree to which the failures fail is, on average, greater than the degree to which the successes succeed. (back)
- This was Josh’s takeaway from his examination of the relation between age and production. (back)
- Or perhaps the optimal strategy is renting rookies for a year before selling them for a profit, as Brian Malone recommends. The only way in which my findings disagree is that it appears 22-year-old WRs in their second year are perhaps undervalued relative to the production gains they can expect. In other words, there’s a case to be made that you should wait one more year before selling JuJu Smith-Schuster, Chris Godwin, Curtis Samuel, or Josh Malone. (back)