Should We Be Age Adjusting NFL Production?
chris baldwin

A question that came up in our email chain among RotoViz contributors was whether it’s reasonable to do some kind of age adjustment for NFL receivers, or whether age adjustments should stop when players enter the NFL.

To clarify the above point, let me just offer that I think we’re on pretty solid ground when we argue that age is an overlooked component for NFL receiver prospects. To illustrate this simply, I asked the question of whether the NFL draft properly accounts for the age of prospects.

Here’s a graph that shows average fantasy points (PPR) for a receiver’s first three seasons in the NFL. This includes receivers who started their careers after the 1999 season, but before the 2014 season. I wanted every player that I looked at to have three years of experience and didn’t want to bias the results by including players who only had one or two years of experience. I also took out the very oldest rookies, any player who was over 25 years old as a rookie. I did this for two reasons. First, those cases don’t apply very broadly since it’s rare for rookies to enter the NFL that old. Second, I didn’t want this work to potentially be affected if the very oldest rookies were creating out-sized effects that wouldn’t translate to rookies in the 23 and 24 year old cohorts.

Rplot199

I used the smoothed trend in that graph to then create a new variable called DEXP (or draft expectation) for each prospect. Then I ran a regression to see if I could improve upon a model’s fit by adding Rookie Age as a variable. I was able to improve upon the model’s fit. You can both see from the scatterplot above, and you’ll see from the results I post below that the fit is not great. But there’s nothing I can do about that. It’s an issue inherent to the nature of football that outcomes can vary so widely. In the model below I’m using RAGE for rookie age because using RAGE makes me seem cool and maybe then you won’t think that I haven’t seen the sun in months.

Capture

Rookie age is a significant variable in this model even when draft position is accounted for. As I mentioned, the r-squared of the model is very modest and reflects how little of fantasy scoring can be explained by even something as powerful as draft position (where presumably a lot of work goes into the process and then the teams that draft players also get to assign playing time).

If regressions aren’t your thing and you want to think about this in another way, I also took each player’s fantasy point average and then subtracted out the draft expectation. So the amount left is the amount they overperformed the draft expectation. Then I aggregated that data by rookie age. Here’s a table that shows the averages:

Rookie Age Average Over Expectation
21 1.16
22 0.29
23 (0.01)
24 (0.74)
25 (1.01)

I’m sure there are other ways to quantify the effects of age for prospects and many are probably better. I did this just to quickly show in a fairly straightforward way that at least before players get to the NFL, it’s probably worth paying attention to age as it relates to fantasy points.

But what about after players get to the NFL? Maybe it’s the case that young players are just good right away, and there’s no need to age adjust at that point because experience level is more important than age? Maybe it doesn’t matter if a player is 21, or 22, or 23 years old, as much as it matters whether it’s their first, second, or third year in the league.

In order to look at this question I decided to look at how much players fantasy points change from one year to the next, but broken out based on both age and experience. I tracked the delta between Season N and Season N+1 average fantasy points/game for any receiver that played at least five games in each season. This includes any player whose Season N+1 occurred between from 2001-2015. I created this heat map:

Rplot198

The way to read the graph is that age is on the X Axis and NFL experience is on the Y Axis. Players that played Season N at 21 years of age, and that was their first year in the league saw an average change of +3.2 PPR points in Season N+1. For players aged 22 that played their first year in the league at that age, they saw an average change of +1.64 fantasy points.

Again, I’m sure there are better or more elegant ways to look at this issue. I wanted to do it this way because the relationship between age and fantasy production isn’t likely to be monotonic. It probably goes up at first, and then goes down later as a player gets older.

But I think that this graph would support the idea that further research into the relationship between age and NFL production is in order, with an eye toward figuring out how much or how little adjustment should be made for a player’s age when they are early in their career. You can see from the graph that those youngest rookies have averaged the most change between their rookie season and their second season. In fact between ages 21 and 23, as well as years one through year two, things look pretty intuitive in the graph. Younger players showed slightly more improvement compared to older players, regardless of experience. But then there are also questions raised by this graph as well. The 24 year old rookies also averaged a decent amount of improvement on the strength of second year breakouts like Mike Sims Walker and Cecil Shorts.

I look at these things from the standpoint of where there’s smoke, there might be fire. So when I do a simple exercise like this, and it shows that those youngest rookies did average the most improvement, I’m interested to look at the issue further. This is a very rough way to look at things considering that I haven’t even adjusted for offensive strength, or made adjustments for availability of targets yet.

At this point we just have some indicators pointing in the direction of age being an important issue, with work still to be done to tease out all of the ways that it could impact how we view NFL players.

What we do
Connect
Support

rotovizmain@gmail.com

Sign-up today for our free Premium Email subscription!

© 2019 RotoViz. All rights Reserved.