Real Time Super Stats: Part II

JustinM suggested that I repeat the analysis I did in my first article, but that this time I should only use road stats to eliminate rink bias. I initially didn't want to do this because I thought it would be a pain to separate out the home and road data, but NHL.com actually already does that. So with that in mind, I'm going to look at road-only stats to see if rink bias was driving the results in my previous article. 

I repeated the analysis that I described in my first article, but I only used road stats this time. Here are the numbers over the past five seasons (regular season points were only for road games):

I then did a correlation analysis between reg. pts. and each individual stat. The results:

In the first analysis, Hits had an "r"coefficient of +0.01. Since there was basically no change in this analysis (meaning rink bias did not affect the total numbers), this is very clear proof that how many hits a team racks up has nothing to do with winning. 

The "r" coefficient in the first analysis for BS was -0.47. While the "r" for road-only stats was weaker, it's still relatively strong on the whole. The conclusion regarding blocked shots is still the same: if you're blocking more shots than average, you're not playing good hockey. 

Giveaways saw a slight increase in significance in this analysis, as the previous "r" was +0.09. While this number moves it out of the "no relationship" category and into the "small relationship" category, it's still too little to draw anything meaningful from it. Giveaways don't have a real correlation with winning, yet the sign on both of these "r's" indicates that our intuition regarding giveaways is a bit off. Ignoring the fact that the "r" is too small, the positive signs tell us that teams with more giveaways are earning more points. This makes sense, as teams with more giveaways have the puck more, and teams that have the puck more tend to win more games than average in the long-run. 

Takeaways also saw an increase in this analysis, as it jumped from -0.01 in my last piece. However, the "r" is barely in the "small relationship" category, so it would be ill-advised to try and draw anything meaningful from these numbers. 

Overall, these results were what I suspected. I didn't think rink bias would have skewed the original data too much, mainly because I think it's an overblown issue in hockey. These real time stats have roughly the same correlation with winning on the road as they do when looking at home and road stats combined. I'm not sure if these numbers are a knock against the rink bias claim, but they're interesting, at least to me.