Does NHL Playoff Seeding Matter?

by Kurt Bullard

NHL Playoff season is upon us, and with it, the unpredictability of the chase for the Stanley Cup. In 2012, the Los Angeles Kings not only won the Stanley Cup as an eight seed out of the Western Conference, but did so in dominating fashion, dropping only four games en route to taking home the Cup against, of all teams, the New Jersey Devils—six seed from the Eastern Conference.

Even though the seeding system has changed in the past two years, the parity in the playoffs still persists. This year already, as of Tuesday, four of the eight lower seeds lead their opening round series. Such parity begs the question: is seeding even predictive of playoff performance?

To examine this further, I calculated the wins totals for each seed from 2003-2013 (obviously excluding the 2004-05 NHL Lockout). I then ran an F-test to see if the difference in the means of the win totals per seed was significant. Although an F-test requires samples to be independent, I believe that the fact that there is a decade’s worth of data cancels out any correlation that there may be between seeds who play each other in the first round every year.

The following table displays the average win total for each seed in the NHL Playoffs in the last decade of the old seeding system:

Seed Average Win Standard Deviation

1

1.70

1.341640786

2

1.6

1.569445091

3

0.7

1.031095483

4

1

1.213953957

5

0.6

0.753937035

6

0.95

1.050062655

7

0.55

0.944513241

8

0.5

1.147078669

There is an evident drop off between the first two seeds and the rest of the field. To make sure this is significant, I’ll run an F-Test testing all of the means.

P(W1 = W2 = … = W8) = .003

This statistic is obviously significant, meaning that all seeds are not created equal. I then tried dropping the one seed from the data to test if the difference in the other seeds was significant:

P(W2 = W3 = … = W8) = .035

Again, not significant. But after I dropped the second seed as well, the results completely changed.

P(W3 = W4 = … = W8) = .52

Such a change in the P-value is huge, meaning that there has been no true difference of those seeded three through eight in the last decade.

This is not to say that a team would be indifferent to being a three seed or eight seed. Obviously, there are perks that come with being a higher seed—reseeding working in your favor, home-ice advantage. But that being said, even these advantages have not given the three seed a clear distinction over an eight seed.

There are several reasons why such parity in the lower six seeds may exist—perhaps home-ice advantage isn’t that big of a factor. Or, higher seeded teams could lose players to injuries while lower seeded teams may be getting them back at the right time, or just the fact that often very few points separate the bottom six teams in the standings.

Regardless, the NHL Playoffs are back, and everyone has a chance (even if this year’s four one seeds are slightly favored historically).

About the author

harvardsports

View all posts

2 Comments

  • Perhaps you should at least mention how NHL playoff seedings were structured. The 3-seed was based on winning the division, not record. Looking at the Eastern conference in 2013, the last year of your study, the 3-seed (Washington, 57 pts) and the 8-seed (NYI, 55 pts) were separated by one win during the regular season. It shouldn’t be surprising that they perform similarly in the playoffs.

Leave a Reply

Your email address will not be published. Required fields are marked *