Frozen Tool Forensics: Home Sweet Home Part II

Chris Kane


This week we are going to take a second look at the data set from last week and dive into a few more players and specific categories.

As a refresher, last week we used The Frozen Tool's report Home/Away and examine trends in categories that happen more often in home games than in away games. The upshot of it all was that most of the counted categories (goals, assists, points, power-play goals, power-play points, hits, and shots) happen more often at home than during away games, and only penalty minutes happens more at away games. These largely fell on a pretty standard line consistent with their size, meaning that there was a lot more variation (so big swings when comparing a player's home pace to their away pace) in categories that had lower totals (goals, power-play goals, etc.) but much less variation in categories with high volume totals (shots, hits). It essentially means that sample size was the biggest factor in determining how much variation there was, but there were a few categories that were unexpected – Power-play points, power-play goals, and penalty minutes. These categories fell outside of the line and so could actually be more or less likely based on where the game is being played.

Today our goal is to take a look at players who have performed consistently across a couple of years, and to see if largely the home/away trend is repeatable through multiple seasons.

To get started I exported another Home/Away report from Frozen Tools – the 2018-19 season. I ran it through the same analysis as the 2019-20 report to get per player numbers for each season for each category. I did a little r^2 to get an idea of how closely correlated the data sets were. The details aren't particularly important here, but I decided to focus on three categories, points, power-play points, and shots. If a player had a lot of control over their home vs away performance, I would expect the two data sets to be relatively strongly correlated (with likely a stronger correlation between shots than points since shots happen so much more frequently). Given that we learned that power-play points happen more frequently at home than we would have expected and that is likely due to officials calling more penalties for away teams, we might also expect that to be fairly well correlated between years.

Without going into the math, the general result is "not that correlated." Power-play points were the most consistent year to year, but it was still a pretty weak correlation. Next comes shots, and then points. This pattern at least makes sense given what we know about the frequency of those stats.

While the pattern isn't that consistent across the entire NHL, there are definitely some players who, over the course of the sample, were consist