Welcome to my weekly NHL power rankings. These power rankings were developed with a simple philosophy in mind: the only thing that matters in the NHL is PLAYOFF SUCCESS! You may be thinking, “yeah, that’s great and all, I know no one gives a flying fart about who wins the President’s Trophy, but how do we predict playoff success by watching the regular season?” Good question. Maybe we look at the “grittiest” teams? Or possibly we look for teams with players who “know how to win”?
Nah, sorry folks, you won’t find any narratives here. I’m a data-driven guy, so I took a data-driven approach to this problem. Here is a rough outline of my methodology:
- Quantify which teams were most successful in previous NHL playoffs. I came up with an equation to represent this numerically, but it boils down to this: the farther the team progressed in the playoffs, the more successful they were rated.
- Determine which regular season stats correlated strongest to playoff success. I used the Pearson correlation coefficient for this.
- Build a model using the most correlated stats from step 2 as input variables. I left out standings variables (wins, losses, points, etc.) because I wanted the model to focus more on the on-ice results instead of overfitting to how many wins a team had.
I have a couple of graphics at the end of the rankings that you can reference if you want to learn more about the model-building process, but enough nerdy stuff. Let’s get to the rankings!
I hope you enjoyed the rankings! There is one more thing I should mention about these rankings, it concerns the input dataset – I use the full season data, with each game given an equal weighing. Immediately this may jump out to you as a major flaw, as we can easily think examples of teams that didn’t have strong regular seasons for the full 82 games but ended up having major playoff success – the 2019 St. Louis Blues come to mind as the obvious example. However, I didn’t just pull this decision out of thin air, I crunched the numbers and found that overall the full season dataset had the greatest predictive power.
The below figure shows how the rankings have changed as the season has progressed.
And finally, here are those extra nerdy details I promised.
You can find me on Instagram and Twitter if you want more data-driven insights into the NHL.
