SnuggaRUDE
Registered User
- Apr 5, 2013
- 9,622
- 7,236
The dataset is flat out flawed.
75+% of your data set would be incomplete when collecting data in this fashion, because most of the players did not play those ten years they are graphing for. You can't graph trajectories of players with incomplete datasets. Your results are going to be misrepresentative of what you are trying to measure.
As a universal scoring measuring stick, context is immensely important. Do you believe Girgensons in his 30s would not outscore his age 20 year old self now if he were playing a full season with Thompson and Tuch? I know he would.
The dataset on the majority of young players is also extremely lacking. The majority of early 20s players play incomplete seasons on callups, and are not eligible for this dataset. Most of them end up playing in the bottom six as filler and their inclusion in this graph would dramatically change the curve, yet they are missing. Then there are the young guys with high draft pedigree that only stick around for a couple seasons, but are given top six minutes, but who have no data in their mid to later years as they are out of the league.
Teams rely on older players to play defensive roles. The coaching scheme utilizes them in matchups or energy roles, and they are instructed to play defense first. This also dramatically skews the stats when viewed out of context. Almost every healthy winger in the league will put up better numbers at age 27 when playing with a McDavid or a MacKinnon than they would at 22, but the majority of players that are still in the league at 27 are playing bottom six roles by then, further skewing this data set.
If you want an honest indicator of progression by age then you take out the calendar year restrictions and run the data on individuals by player age. The graph changes substantially. It would still be highly flawed as it doesn't account for the context, but it wouldn't be quite as bad. As Chain pointed out, the best teams in the league are rolling out mostly rosters devoid of a lot of kids. This is definitely a case where bad data leads to faulty assumptions.
And for the record, we were talking about this Sabres team and the age of the players three years ago, mostly 19-22, as not being ready to go "all in on the trade market". We aren't talking about 24-26 year olds.
You don't need a 10 year career to show up on the curve. The average player's career isn't even that long.
All this curve does is graph the % of a player's career year in points for a given age. Missing early data when they're call ups won't change anything. It won't change their peak year, nor will it change the % of their peak when they're 30.
Again, that's not 'context' it's a different idea. Go ahead and graph xG% if you want to see how puck possession plays out over a player's career.
I'm confused by the bolded. That's basically what the curve does. If a player's best year is 100 points and they score 75 points subsequently they're at 75% for that age. Combine all of the players over that data set and you have a curve with confidence intervals (gray).
EDIT: Btw here's the other link from my original post. It plots WAR over age. A New Look at Aging Curves for NHL Skaters (part 1)