I love hockey for many reasons, including how statistically interesting it can be. That's why I spend so much time blogging about hockey and studying stats. But to call myself a stats-guy and not call into question some of my peers would be remiss.
There is a thread of conversation going around that is something like this: "shucks, have you noticed the Thrashers win more games when they're outshot?" Graphs and other representations abound. To make things worse, these graphs and representations are coming from some of the bloggers that I admire most. Admittedly, it's often indirect, but they leave open the door for readers to make the wrong conclusion.
Folks, let's set the record straight: despite what you may have seen, getting outshot is not a good indicator of success. As it would turn out, we do score more goals than our opponent (the necessary condition to win a game, thanks Mr. Madden) when being outshot, but there just isn't enough data to establish causality.
Here's the data:
Two important things:
- The trendline does show that the team has a better goal differential with a worse shot differential. In a prototypical "even-game," we'd be outshot by 5. This is the conclusion that you might have drawn if you didn't read point number 2.
- The R-squared value of that regression is 0.083. This means that our model is not a very good one at all. Something else must be causing this trend. If we throw aside statistics for the moment, though, and just look at the data, it's pretty clear: there's not really any relationship in the data at all. It's almost completely random.