As I grind away on a blog trying to come up with a clear evaluation of Keith Yandle's true worth on the UFA market coming up in July (yes, I know I'm getting ahead of myself here) I continue to branch off into other analyses.
Another quick 'gorilla math' blog here, but you know those are my favorite (as always, shout out to Gramma in ROUNDERS).
The quickest way to accomplish this feat was to run a regression analysis. Using a data set of defensemen who are currently playing under a contract they signed with UFA status, I seriously hampered my sample size. Not to mention, to get a fair barometer of skill, I also limited the player pool to skaters who have seen more than 1000 minutes on ice combined between this current season and last season. This was also a way to get a lot of data on these players.
Ultimately, there are currently 91 d-men in the NHL playing under a contract they signed as a UFA. At first, this seemed pretty low to me, but considering that averages out to 3 per team, it makes a bit more sense. Of these 91 players, 76 of them have played more than 1000 minutes between this season and last.
Not a significant sample size. But we press on!
All stats used in this blog were provided by corsica.hockey
Figuring out what correlated to a player's salary turned out to be a bit more of a task than I had expected it to be. Off the bat, I went with the basics (stat[correlation to AAV]): Goals[.18], Primary assists[.29], Relative Corsi For %[.12], Relative Scoring Chances For %[.06], Individual Corsi For[.26].
Turns out, GMs don't really care about these things when they pay for defensemen. Well, what do they care about?
NHL GMs are currently paying for guys who can log minutes, and are on the ice for shots for and goals for their team. There were no other factors (unrelated to the ones above) that scored over a 0.5 (or under -0.5) to AAV.
Finally, these metrics were thrown into a regression, and the model was built!
Are you thinking those p-values make this useless? Me too. But since the ASA just put p-values on notice, and this is more for a little bit of fun than anything else, we press on.
The final move here was to graph the expected annual average value of players relative to their actual annual value, and highlight some notable players (including Keith Yandle).
The model actually does us some pretty good justice here. Keith Yandle is paid what the model expects him to be paid, the residual coming in at $9,000 more than he is actually earning. Further, the model finds Alex Goligoski underrated in terms of value (a fair assessment), as well as the Andrew MacDonald contract to be a tragedy (it is).