“Milan Lucic Stanley Cup celebration” by Ashley Bayles from Canada – IMG_5526. Licensed under CC BY 2.0 via Wikimedia Commons.
Last time we looked at the relationship between hit and goal differentials. We showed that the outhit team tends to also be the outscoring team.
On Twitter, the subject of playoffs naturally came up. Do physical teams get an edge in the post-season?
I’ve been already pulling some data on the playoffs and doing some studies. I thought this would be a good opportunity to show a few of my early findings.
The playoffs are often considered unique from regular season. The stakes are higher and pressure impacts individuals differently. The level of competition rises with the lower teams not making the cut. The calls are looser and allow for more physicality. The sample sizes are also extremely smaller and a four game cold streak can send even the best team in the league home.
We looked at multiple regular season statistics in how closely they relate to themselves in the post-season and also playoff goal differentials. As hypothesized, hit differentials had a negative relationship with post-season goal differentials. Shot-metrics ended up similar to experiments previously conducted on regular season relationships.
Data on each team’s individual post-season results was pulled, and with it came the regular season results as well. Six separate statistics were selected to be compared: Hit%, FaceOff%, Corsi%, Fenwick%, Shot%, and Goal%. The data was for 5v5 only, and not close nor score-adjusted.
The Pearson’s correlation coefficient was then measured for each statistic twice:
1) Regular season values to post-season values of the same statistic
2) Regular season values to post-season Goal%
RESULTS AND COMMENTS
Interestingly enough, the regular season statistic that has the strongest correlation to itself in the post-season is hit differentials. Face off and goal differentials are the opposite with the weakest correlation with itself.
A team’s regular season Corsi% has the strongest correlation with it’s post-season Goal% and face offs have the weakest correlation.
How well a team outhits or is outhit in the regular season has a pretty strong relationship with how well a team outhits or is outhit in the playoffs, especially given the small playoff sample size. However, teams that outhit their opponents in the regular season tended to have poor success in the playoffs, although the R^2 was only about 2%.
How well a team performed in the face off department typically had no relationship with post-season success at 5v5 and did a relatively poor job in even correlating to face off success in the playoffs.
The shot metrics Corsi, Fenwick, Shots, and Goals fell similar to what we already have seen with predicting future regular season success, given the sample size. Corsi is the strongest of the shot metrics, although places behind goal percentage without accounting for score-adjustments (also note Goals would be grabbing home ice advantage as well).
If you are new to shot metrics and predictiveness, this article on Corsi Tied by J.Likens and this article on Adjusted Possession Measures by M.McCurdy are both highly recommended. My guess is that we would find adjusted-Corsi to actually be lot closer to Goal% in power than we do in predicting regular season success.
The regular season sample is much larger, which has been shown to improve Goal% predictiveness (better indication of scoring and goaltending talent) and decrease Corsi% (which usually peaks between 20-30 game previous sample).
Outhitting for the most part isn’t a very good indication of a good team. In order to hit your opponent (for the most part legally), the other team must possesses the puck. While possession is not the sole causation to generating wins, it is something that stronger teams tend to do more often than not.
The playoffs, despite its reputation and its nature, seems to be no exception to the rule. Teams that generally outhit their opposition in the regular season is still an indication of a team that will likely struggle in the playoffs.
Aside: Whenever trying to predict post-season, keep in mind that you are predicting something that is highly luck (read: variance and outlier) influenced. The median post-season team plays eight games in the playoffs (duh), while the mean and standard deviation are eleven and six. With samples as small as this, you must remember that you are trying to predict success generated by both talent and luck.