Quote (Hizkuntza @ 19 Jul 2019 18:17)
I hate it so much that this analysis is paywalled and I have to wait for you to post it here LOL. More that I'd like to understand how these guys arrive at these numbers rather than taking them at face value.
He bases his stats of evolving hockey and created his own model explained below:
How the model works
It’s mostly outlined here in this FAQ posted before our 2017-18 projections, but basically it’s built at the player level using Game Score – a stat I adapted from basketball a few years ago. Working at the player level rather than the team level is one way that my model differs from others that are scaled via team performance only. It offers some challenges in terms of allocating proper credit, but has the advantage of being able to instantly factor for injuries and trades in ways a team-level model cannot.
Game Score is a linear weight model with the weights for each stat within it being derived according to the frequency of goals occurring from them and are as such:
Goals: 0.75
Primary Assists: 0.7
Secondary Assists: 0.55
Shots: 0.075
Blocks: 0.05
Penalty Differential: 0.15
Faceoff Differential: 0.01
5-on-5 Corsi Differential: 0.05
5-on-5 Goal Differential: 0.15
It uses data from each player’s last three seasons, with each component weighted by recency and regressed to the mean individually. That means that the weight for each prior season is different for goals than it is for shots or blocks (and different for forwards and defencemen), as is the regression factor. On top of that, there’s an age adjustment (using methods outlined here) performed at the start of each year that slowly lessens until the end of the season, as well as a small usage adjustment that factors in a player’s teammates and competition based on 5-on-5 Game Score.
From there, each player has a projection for each component going forward and that’s plugged into the Game Score formula to get a projected Game Score going forward. That’s then transformed into a wins above replacement rate (with replacement level being the 372nd forward and 186th defenceman) to create Game Score Value Added, or GSVA. That value is added up for each team based on the players in their starting lineup, and voila: team strength projections.