OK, so no one
really asked me to break this ranking down. But it's too big to keep to myself. You're welcome!
So I looked on nhl.com - wow, there are a lot of stats there for every team! It seemed weird that the only numbers the advanced stat community has decided to add together so far has been shooting percentage and save percentage. Then it clicked with me: I could add all kinds of different numbers together!
Once I cleared that hurdle, I briefly pondered what kind of numbers could be added together that sounded vaguely logical. I think maybe I've heard before that special teams (that's both the power play and the penalty kill) are considered "good" if they add up to a percentage of over 100. Say for example, a 16% power play and an 88% penalty kill. That adds up to 104%! Pretty good. Plus it also has the cool side benefit of being able to label which teams are giving more than 100% game in and game out!
Armed with this insight, I hastily copy and pasted the NHL team stats front page into Excel 2003. It was annoying, because I tend to always copy the stats first, and
then open Excel 2003 -- but then the stats paste in as some weird object. So I always re-copy them again, and then Paste Special them as HTML so they kind of paste into normal columns. Another obstacle overcome!
Now it gets to the complicated math part. I didn't want to add the percentages as they're commonly expressed (16% and 88% as before), because the NHL publishes things like Points Percentage and Save Percentage to three decimal places out of 1.000. So I converted the special teams percentages into that format, and then added them together. This gave me some numbers over 1, and some under 1. The numbers over 1 looked kind of funny though, because they had these extra zeroes in them. Numbers like 1.084, 1.016, and so on. Whereas the numbers under 1 looked like normal save percentages and point percentages. So I decided I was just going to eliminate those extra zeroes. Get outta there, zeroes! So for example purposes, the New York Rangers (objectively the worst team in hockey, as shown by my ranking) had a PP% of 23.1 and a PK percentage of 84.1. Those become 0.231 and 0.841. Add those up and you get 1.072, which then becomes 1.72 because of the annoying zero. In cases where the three decimal place number doesn't neatly get to two decimal places by dropping a zero, I just rounded them. Because rounding never hurt anyone.
So now, my list of teams all has numbers between 0 and 2. Which, not coincidentally, is the same range of points a single team can earn in a game. 0 for a loss, 1 for a OTL or shootout loss (which is absolutley .500 hockey, as discussed in another thread), and 2 points for any win. So then it hit me - this new number combination is the number of points that a team should
expect to earn, on average, in every game they play. But I didn't want to simply convert this number into points - team standings already are based on points. Instead, I took each dervied number and divided it by the maximum number of points per game (i.e. 2) to get the teams
expected points percentage.
With an expected points percentage clearly defined, we can now compare it to the team's
actual points percentage in the season to date. Take the actual percentage and subtract the expected percentage. A negative result is just that, negative. You don't want your team doing worse than expected! A number close to zero means a team is more or less doing what they should be doing. The higher the positive number, the better the team is. And the team with the highest positive number is therefore the best team!
All I needed now was to give this number a name. Something that pops, like PDO. I landed on FLQ. It doesn't stand for anything, it's just the first vaguely inappropriate, but not outright offensive, three letter acronym that came to mind.
So there you have it. Here behind the spoiler tags are the ranking I posted above, but with the added insight of how they are definitively ranked from worst to best - via FLQ.
I didn't go back and prove how FLQ works with every past season in the NHLs 100 years of existence, because I got kind of bored half way through writing this.
Thoughts and discussion about how brilliant this analysis is are welcome! Haters check yourselves!
This is Definitely a Real Advanced Stats Discussion Thread
