In part 3 of our Glossary series, we dive headfirst into RAPM (Regularized Adjusted Plus Minus) and its use in hockey skater evaluation. When we say headfirst, we really do mean it. This part in the series has a lot packed into it. It’s not required, but listening to parts 1 & 2 ahead of time will make things a bit smoother. Let’s go.
Hockey Stats Glossary Series: Part 3 – Regularized Adjusted Plus Minus
Good explanation
Hello Kash
Great job guys. I had a bit of trouble understanding the stint-level approach to the regression, but I *think* that looking at your RAPM write-up has mostly cleared that up. Are there any issues associated with the fact that most of the time you’re predicting a [insert variable]/60 of 0? Or does this not matter?
Also, as an experimental researcher, independent and dependent variables are definitely the appropriate names when you’re doing some sort of a manipulation, but I definitely agree that in this case other terms are probably more appropriate (input-output, predictor-target, feature-outcome, etc). I find that settling on one that’s my favorite makes explaining things easier.
Also also, idk if I’d call this kind of work inferential as much as exploratory. You’re not making any hypotheses or drawing many conclusions about the underlying population parameters, other than, “is this coefficient bigger or smaller than this coefficient”. I guess that’s inferential to some degree, but it still feels more exploratory imo.