NEW PERK: Skater 2021 RAPM (Patreon)

It's been a long time coming, but the first step towards getting the isolate models ready to go is here: Patrick and I have now run the RAPM model and the results are available exclusively to you guys. 

Here is the link. 

RAPM is the most important part of Wins Above Replacement - it's the way we use regression to estimate how a player impacts on-ice events when they're on the ice, isolated from teammates, competition, game situation, venue, and more. The RAPM represented here is estimated impact on Expected Goals (scoring chances) for and against at even strength. 

The RAPM used here is what we call a "vanilla" RAPM model - that is to say that it does not use priors from previous seasons. So it is solely measuring performance in this season so far. WAR itself will use a prior-informed model but because of the limited sample size the results would still be dominated by last year at this point. That smaller sample size does mean that these results are by no means definitive or a perfectly accurate assessment of performance or especially skill, but they should provide a solid indication of how players are impacting their teams.

There are four dashboards included:

1. Scatter Plot

This chart allows you to see all players at once, with filtering options based on TOI and position and highlights options for teams and players. The quadrants should give you a good sense of where players fit in overall.

2. Rankings

This shows the rankings for RAPM xGF, xGA, and total xG +/- with multiple filters to narrow down the results by time on ice, position, player, team, and division.

I've had a lot of people ask about player cards, understandably. Those are coming as soon as possible. A key difference between EvolvingHockey's WAR model and Patrick's is that theirs uses indicators of isolated impact (mostly relative goals for) while his uses the real deal. The benefit of the former is that there isn't a regression component so you can basically put out the stat all year long, but the benefit of the latter is that we believe it will more accurately represent a player's own impact. This is particularly important on the PP and SH where sample sizes are very small. 2021 Player Cards will definitely be out well before the trade deadline and I'll give updates on that as things become more clear. In the meantime, I've been prioritizing putting out as many perks and visualizations as possible for you guys using this season's data and I hope you've been enjoying it. 

Another Note on Tableau Viz: The "date updated" at the bottom of the Tableau page refers to the visualization itself, not the data. The data automatically updates every morning.

Cheers,

Jack