# Rising and Falling in the Western Conference

The simplest way to win a hockey game is not to “forecheck harder” than the opposition. You don’t get two points for blocking more shots than the other team, and at the end of the day, no one cares if you played more physically or even directed more shots at the net. What matters is the scoreboard, and the simplest way to win a hockey game is to score more goals than the other team.

It’s such an easy concept, and yet, when analyzing the standings throughout the season and looking to see where there’s room for teams to move, the “put the puck in the net, prevent the same” formula is all too often thrown to the wayside.

Baseball writer-turned-executive Bill James, who I regard as the greatest sportswriter of all time, had a very interesting metric to help us determine how many wins a team should “expect” to have. It is called Pythagorean Expectation, and while i sounds like a lot of math, it really just involves plugging a couple of numbers into a spreadsheet and letting it do the work for you.

Pythagorean Expectation, according to Wikipedia, can be described as such:

$\mathrm{Win} = \frac{\text{runs scored}^2}{\text{runs scored}^2 + \text{runs allowed}^2} = \frac{1}{1+(\text{runs allowed}/\text{runs scored})^2}.$

In hockey, we can simply replace “runs” with “goals” and at the end of the season, a team’s goal differential syncs up very well with a team’s record, except in maybe three or four extreme cases. The theory behind this is that winning close games isn’t a true team talent as much as it is luck and random draw. If a team scores 270 goals and allows 200, regardless of the order, it will win 53 games.

Since winning close games is luck-based, we can use it to look at which teams may have enjoyed “good luck” in the first part of the season. What Pythagorean Expectation allows us to do this early in the season, is judge how many games a team should have won, not necessarily how much they have won. We should be able to see who is due for a climb.

First, the current Western Conference standings:

Team GP W Pts Pts/82
Detroit 22 14 29 108
Minnesota 24 14 31 106
San Jose 21 13 27 105
Phoenix 23 13 29 103
St. Louis 24 14 30 103
Chicago 25 14 31 102
Vancouver 24 14 29 99
Dallas 24 14 29 99
Los Angeles 24 12 28 96
Edmonton 24 12 26 89
Nashville 24 11 26 89
Calgary 23 10 21 75
Anaheim 23 6 16 57
Columbus 24 6 15 51

The cutoff between first and eighth, based on simple point percentage, is 99 points. The Western Conference has apparently played well enough against the East so far that a team will need almost 100 points to crack the playoffs. If that number seems a little high, it’s because it is, and it is bound to drop to a more reasonable 94 or 95 points.

If the playoffs started today, the Canucks would apparently be the 7th seed, playing in Minnesota. I’d almost take that.

Why? Because take a look at each team’s expected win total over 82 games:

Team GF GA xW xPts
Detroit 65 49 52 115
San Jose 60 48 50 110
Vancouver 73 60 49 108
St. Louis 59 50 48 105
Phoenix 65 57 46 103
Edmonton 65 60 44 99
Minnesota 57 53 44 98
Los Angeles 57 55 42 95
Chicago 80 78 42 94
Dallas 62 65 39 88
Nashville 60 63 39 88
Calgary 51 60 34 79
Columbus 55 79 27 64
Anaheim 50 76 25 60

(xW = expected wins | xPts = expected points)

Now, Vancouver appears to be on a 108-point pace, as they have outscored their opponents by 13. In essence, their early record is disproportionate to their play. Here, Minnesota, with a 4-goal differential, has simply 98 “expected” points. For expected points, I gave each team the benefit of ten overtime losses. There’s no real scientific basis for this, but it gives us a better ballpark figure for how good a team will have to be to make the playoffs.

Here, Minnesota is 7th, and the Canucks would play a series against Edmonton to start the playoffs.

According to early season goal differential, Columbus, Edmonton and Vancouver are much better than their records indicate, while Chicago, Minnesota and Dallas have earned the benefit of some close wins so far.

• Mantastic

you should use that metric against previous seasons and see how close those numbers are to reality.

• Wanye

One thing to consider is sample sizes. Over 22 or so games a blowout can have a huge impact.

For instance, Edmonton beat Chicago the other night 9-2. Without that game, Edmonton is a -2 and Chicago is a +9.

• Mantastic

How are you calculating the expected points?

• Mantastic

on expected wins… of course.

• Mantastic

Canucks have an xWin of 49 games, but an xPoints of 108.

At 2 points a game, 49 wins gives you 98 points. So how are those 10 OTL’s being calculated?

• Mantastic

Heh, Your post brought memories of something I did many years ago. Damn, now I feel old(er).

http://jesgolbez.blogspot.com/2005/06/hockey-rants-study-200304-nhl.html

So, this was before the shootout, which made the standings even more weird.

Things to consider:

1. Not every game is worth the same amount of points.
2. Year-to-year. I’d love to see how the Plexiglass Principle affects lucky/unlucky teams from season to season. I never did follow up on my original study, and the lockout did shake up things so much that I figured rosters would be far too different.
3. Do some coaches get consistently lucky/unlucky?