NHL Skater Aging Curves
I'm working on building an NHL statistical model, and one of the outcomes was these aging curves for forwards and defensemen. They seem to largely match the eye test (although there will obviously be some outliers on the elite side like Crosby).
The curves are built using data from the 2011-12 through 2025-26 seasons. The older data is less reliable (some venues counted things differently... Cal Clutterbuck anyone?) so I normalized the coordinates and scaled the counts for various events to try to level the playing field. I fed this data into my xG model and used it to generate a GAR (goals above replacement) stat for each player in each season. I also normalized scoring rates for each season to account for the changing meta over the years.
I used a delta method to look at how each player's production changes over the years to combat survivorship/selection bias. Shout out to Stat Shot (great book) for putting me onto this method.
The players are bucketed by tier using their career highest 4 consecutive season averages. For forwards, this was done using points/60 - elite is over 3.0, top 6 is 2.2 - 3.0, middle 6 is 1.4 - 2.2, and depth is less than 1.4. For defensemen, the tiers are done by TOI/game - elite is over 24 minutes, top pair is 21 - 24 minutes, second pair is 17 - 21 minutes, and third pair is less than 17 minutes.
The "units" obviously don't mean much outside of my model, but I think the general shape of the curves is informative. They also corroborate Ryan Lambert's theory that once a guy hits 30 he's basically toast.