clock menu more-arrow no yes mobile

Filed under:

Why You Should Never Take a Statistics Class

Getz: "Does it matter that I struggle with faceoffs?"

Bobby: "I'm not sure I'm the best person to answer that."
Getz: "Does it matter that I struggle with faceoffs?" Bobby: "I'm not sure I'm the best person to answer that."


It's been a few weeks since the Anaheim Ducks' season ended, and I think the mourning period has passed for me. That means it's time to get back on this horse, and find something interesting to write about.

As some of you may know, I'm currently pursuing an MBA. My current class happens to be a statistics class, and we just covered regression analysis. After playing around with Excel, and learning the basics of regression analysis, a light-bulb went off.

Using a simple regression approach, can we determine what statistics/numbers/hockey phenomena correlate with total points?

Now I fully realize I'm probably not the first person to tackle this subject. I also fully admit that I have, at best, a rudimentary understanding of regression analysis. So while I may not be the best candidate to undertake such a study, I figured it would make for some interesting discussion during an unfortunately long summer.

My Methodology

I plan to take ten different variables (more on this later) over the past one and three year periods and perform a basic regression analysis using the Least Squares approach. My hope is to determine whether or not any particular statistic or set of statistics has a mathematically significant relationship to the number of points a team scores in a season.

In order to simplify things, I'm going to use end-of-year, regular season stats. I know that this is imperfect, as trades and injuries can affect a team's personnel (and thus, their overall performance), but there's just no other way to reasonably tackle this project.

Over the next few weeks, I'll compile the data, graph it, analyze it, and present the findings to you. I'll do this in a series of posts, so that the math doesn't get too overwhelming.

What's the Point

Aside from getting Arthur off my back about writing posts, I'm legitimately curious if the things we think matter actually matter. For instance, I certainly think that scoring more goals would get you more points. But if we look at the stats from for the year, we see that two teams in the top ten in goals per game didn't even make the playoffs.

I also think it would help us have a more informed debate about the way to build a hockey team and the kinds of players a team should target. Does it matter if a team wins faceoffs? The Ducks ranked 26th in faceoff percentage this year. Would it be worthwhile to investigate centers that win more faceoffs? Or is this a common misconception among the hockey world? I'm hoping to find out (or at least get a decent hint about) the answers to such questions.

Where You Come In

I'd like to solicit the community's help in coming up with which variables to test. I've thought of the basics like Goals For, PIM, Goal Differential, Blocked Shots, etc. I've also thought of other more esoteric measures like the number of 20+ goal scorers on a team, the number of D-men with a net minus, or the number of players with 50+ points.

As you can see, there's literally hundreds of variables we can examine. While I think certain things should be tested (specifically: Goals For, PIM, Blocked Shots, PP%, and Faceoff %), I'm open to suggestions on what other variables to measure. And if you're convincing enough, I'd be willing to abandon the five I've already specified.

So if you'd like to help, please leave your suggestions in the comments. If you've seen other studies like this, feel free to share any info from them. And, of course, I'd love to hear your predictions about what will or will not statistically correlate to the number of points a team accumulates in a single campaign.

And if you think this whole exercise is a waste of time or somehow flawed, feel free to tell me that too.