Why are the hockey playoffs so unpredictable?

The NHL playoffs have many more upsets than the NBA. Adrian the Canadian tells me that this is ruining their product, since the most exciting teams often get unlucky and bow out early. I can’t help but agree — I stopped watching this year after my favorite team (the Red Wings), my local team (the Bruins), and probably the best team (the Penguins) got bounced. The NHL wasn’t always so unpredictable — the Canadiens, Islanders, and Oilers won 13 of 15 cups between 1975-76 and 1989-90. Adrian’s theory is that the the rise of the butterfly goalie has increased save percentages, which makes outcomes more random.

It’s pretty easy to show that increased save percentages do indeed muddy up the result. I generated 1,000 simulated games for three sets of parameters. First, the 1980s (before the butterfly):

  • Both teams: 89% save percentage
  • Team A: 32 shots per game on average
  • Team B: 28 shots per game on average

Then, for the late 90s/early 2000s (butterfly goalies, slightly fewer shots on average perhaps due to popularity of the neutral zone trap):

  • Both teams: 92% save percentage
  • Team A: 30 shots per game on average
  • Team B: 26 shots per game on average

For the post-lockout period (butterfly goalies, rule changes favorable to the offense):

  • Both teams: 92% save percentage
  • Team A: 32 shots per game on average
  • Team B: 28 shots per game on average

With the 80s setup, Team B (the underdog) wins 41.4% of the simulations. For the 90s/00s setup, Team B wins 43.0%. Post-lockout, Team B wins 43.3%. In a seven game series, a 41.4% chance of victory in each game translates to a series win 31.7% of the time. This number jumps to 35.6% for a 43.3% game-by-game winning percentage. So, better goaltending will indeed help underdogs get through (slightly) more often.

We’ll also see more playoff upsets if seeding doesn’t accurately reflect quality, since then the lower quality team benefits from an extra home game. The addition of shootouts in 2005 and the switch to three divisions in 1998 both screw up seeding. For example, the Panthers and Coyotes should not have been three seeds this year. And, good shootout teams are losing at a remarkable rate in the playoffs but benefit from their shootout wins in the standings.

Regardless, I think Adrian has a point. Fans love dynasties. They build identities for stars that generate lots of interest. The NHL can’t do much about improved goaltending (a smaller stick or smaller pads?), but they could eliminate shootouts and change the seeding rules to help the best teams get through.

Note: my parameter choices above come from statistics posted on Quanthockey.com, which has a lot of cool stuff.


8 responses to “Why are the hockey playoffs so unpredictable?

  1. Nice work Tyler, but you only get my theory half-right. I think that the rise of the butterfly goalie has made goals themselves more random. In other words, the 1980s Oilers could set up quality shots that were likely to lead to goals while today’s goalies (with their large pads and butterfly style) must be beat with more “random” shots – deflections, etc.

  2. Yes, indeed – that (potentially) makes things even worse. This might be measurable by looking at the spread in regular season standings, but lots of other things affect that (distribution of talent being the main one).

  3. Nice write up and that is a creative way to look at things, but I don’t think that the sv% is changing the landscape of the NHL playoffs all that much. If the increased sv% increased the odds of an upset by only 4% wouldn’t that only explain 1/25 of the increase in upsets in today’s game? That’s not all that significant. I think the much more likely reason is that a salary cap was implemented which made it difficult to keep 6 or 7 future HOFers on a single team for that long. Also, I think the casual fan may want to see the penguins play the red wings every year, but isn’t the point of having 16 teams in the playoffs so that teams have a chance to ride a hot goalie deep into the playoffs? If people didn’t want this chance then the top teams would just get a bye.

  4. Hey Mark – thanks for the comment. I totally agree that the salary cap likely contributes a lot to the upset rate in the NHL. There’s a salary cap in basketball too, though maybe it doesn’t matter so much since 2 or 3 basketball players can dominate a game (and season). I still think the goaltending improvement is important though. Imagine that the playoff upset rate used to be 31% in the NHL in the 80s and now it’s 40%. If 4 percentage points of the change is due to goaltending, that’s almost half the total change (9 percentage points). And, I think 9 percentage points would be a noticeable increase.

    Anyway, I think you’re right. An even spread of talent has to matter. However, the statistical difference between basketball (where many actions affects the score) and hockey (where the score almost never changes) must be contributing too. There’s just a lot more meaningful chance variation in final hockey scores.

  5. Also – I’d like to test the hot goalie theory. My hunch is that most goalie streaks are an illusion (just chance variation). Most evidence about hot shooters in basketball points that way, and, in general, psychology research says that we underestimate the likelihood of streaks when there isn’t any real momentum or persistence (they tend to happen just by chance). However, I haven’t seen starting goalie data aggregated anywhere yet.

  6. Pingback: Why are the hockey playoffs so unpredictable Causal Sports Fan| swimming Pool Contractors Ottawa|Pool renovation Ottawa|

  7. Excellent article. I think the NHL and those of us that love the game need to examine these trends more rigorously. My wife and I have been fans for many years and still enjoy watching the spectacle on the ice but individual game and playoff results have become almost meaningless. The strategy seems to be to throw as many pucks at the goal as possible and hope for a deflection. I think one way to parse out the confounding factors of parity due to salary cap versus changing goaltending styles and equipment is to do an analysis, controlling for quality chances (you would assume those should be more even now if teams have similar talents), and looking at the percentage of chances that actually result in goals. As long as you have enough data to reach statistical significance this should be doable. Of all these factors, the variables that can most easily be changed are the standards for goalie equipment and possibly the size of the net.

    • Definitely worth a try to check the conversion percentage on total chances. I have to get the data first (ideally stretching back in time a ways) and hope for consistency in the way chances have been marked down. For example, if goals are scored in more random ways now, it might affect what’s considered a “quality chance.” Still, it would be a good thing to check.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s