Tuesday, March 14, 2006


My friend Jeremy sent out an email recently where all NHL teams with PP% + PK% was over 100. His idea was that teams with PP+PK>100 will make the playoffs. Unfortunately, that list has a couple of weird teams in it. Minnesota is 106. Boston and Phoenix are over 99. And LA is third from the bottom, with 94.1. I tried to put the table in here, but I can't get blogger to put it in right. If I put in the html, it's looks like crap. I uploaded a gif of the table from Excel, but it shrinks it to nothing. If you want the data, email me.

This has all of those stats , plus a few more. The bottom line is how each column correlates to each of the states. 1 means a perfect correlation, 0 means no correlation. Wins has an almost perfect correlation to the number of points a teams has (duh!), and the number of games played has almost no correlation to the number of points (also duh). But lets look at the other stats. Wins and losses mean points, so seeing that those strongly correlate isn't really suprising or useful either. But let's look at PP%, PK%, and PP+PK. PP% correlates stronger than PK% to the number of points a team has, so that implies that it's better to have a good power play than a good kill. But PP+PK has a higher correlation still. Also, if you rank the teams by their PP and PK, and have a rank for the best PP+PK, that's an even better correlator than strictly PP%+PK% (but very close). Even though there are some outlier team in the PP+PK stat, it's still the best correlator to points.

The rightmost column is a little experiement of mine called Pythagorean Points. Rob Neyer and others have used the Pythagorean theorem for years to try and predict the number of wins and losses a baseball team has based on their run scored and runs allowed. This is a similar thing that I modified to try and apply it to hockey. Ties make it hard. I calculated the average number of points a teams scores in each game (1.11), then times that by two, then that by the pythangorean percentage. It's good to see which teams have been "lucky" this season. It stands to reason that if you score 200 goals and give up 200, your winning percentage should be about .500. If it's not, that means you've been somewhat lucky. Philly is a good example. Scored 213, given up 212, and are 35-20-10. Ottawa is somewhat unlucky. They've scored 252, only given up 152, and have 91 points. According to my results, on average a team with that GF/GA ration should have over 100 points already.

Anyway, something to think about and comment on.

No comments: