Regression of P% vs Various Factors
The regression of team winning (P% - percentage points) vs different factors from 2007 to 2011 is shown in the chart below for
1. P% vs Fenwick Tied Road % (FTR)
2. P% vs Fenwick Close Road % (FCR)
3. P% vs PDO (SH%*1000+SV%1000)
4. P% vs FTR + PDO/10 or P% (FTR% + SH% + SV%)
5. P% vs FCR + PDO/10 or P% vs (FCR% + SH% + SV%)
Fenwick Close Road + PDO/10 gave the best average results for Rsquared with the lowest standard deviation. Team winning is 54% explained by team shooting + team defence + puck possession (that is, PDO /10+ Fenwick).
Effect of Power Play and Penalty Kill
Adding special teams (PK PDO and PP PDO) to the regression for a single year improved the regression a few percent though it seems to be scraping the bottom of the barrel of relevant factors. In that regression PK PDO was 2.5x more important then PP for teams that tend to win.
The Hockey Gods Matter
This implies teams that win tend to based on
- 46% Luck (or some yet unidentified skill) and
- 54% PDO+FCR% (SH% + SV% + FCR%)
Winning appears to be almost 50% luck based.
Which team will regress this season?
P% plotted against (SH% + SV% + FCR %) gives an indication of team that maybe expected to regress.
MIN, CGY, DAL, NYR are expected to regress downward and CLB upwards. The regression accounts for 54% of the results so 46% of the results of winning are either luck (or some yet unidentified skill). Note +-1 and 2 sigma bands are shown in green and blue respectively. The red regression line used is the 2011 best fit.