Goalies and the Playoffs Part 1: Why are Goalies Better in the Postseason?

Joel Short
September 23 2016 09:16AM

A few years ago, I started looking at the impact of goalie performance on success in the playoffs. 

One of the first things I noticed was that the league average Sv% tends to be significantly higher in the playoffs than in the regular season.  Here are the averages going back to 83-84 (the first year Sv% was recorded):

RS:P Graph 1

You can see a few things right away:

  • Average Sv% has been rising over time. This is widely recognized, and the causes are pretty well understood.

  • The playoff average is more volatile (to be expected, since it’s a much smaller sample).

  • The playoff line seldom dips below the regular season.

  • The 80’s were nuts.

Over the last 32 seasons, the playoffs have averaged an 8-point higher Sv% (.008) than the regular season. That’s a big gap – only 3 goalies in the regular season this year out-performed the league average by 8 points or more. 

So it begs the question... Why are goalies so much better in the playoffs? The most obvious explanation is who’s playing (or rather, who’s not). Teams tend not to play their backups in the playoffs, if they can avoid it, and teams with poor goaltending are less likely to make the playoffs at all. So let’s remove the goalies who didn’t play in the playoffs from each year’s regular season averages, and see if that accounts for the difference.

RS:P Graph 2

As expected, goalies who see playoff action do (on average) post better regular season numbers than those who don’t. Removing non-playoff goalies accounts for a good part of the gap (about 5 points on average), but not all. The playoff Sv% is still higher in most years, at least until recently.

But if we want to know how much of the Sv% gap is due to who’s playing, we can do better than the yellow line. A lot of backups do see playoff action, but don’t log significant minutes. And we’d expect teams with better goalies to go further in the playoffs, thus having a larger impact on the playoff average. So lets take the regular season Sv% of each goalie who plays in the playoffs, and weight it according to the number of shots they face in that playoff year.

RS:P Graph 3

The orange line – the average of goalies’ regular season Sv% weighted by the number of playoff shots they face – represents what I’ll call Expected Playoff Average Sv% (EPA). If playoff hockey is just like regular season hockey, with the only difference being which goalies are playing (and how much), this should look a bit like a trend line for the playoff average Sv% (the only difference being normal variance due to the smaller sample of a playoff year).

The weighted average is only slightly above the unweighted average – about 1.5 points – but it’s pretty consistently higher. This fits my expectation that teams with better regular season goaltending are somewhat more likely to advance in the playoffs, although I’m not sure how I’d go about quantifying that advantage.

Again, it’s expected that the playoff average is more volatile, because of the smaller sample. I’m not sure that you could find causes for individual peaks and valleys in the playoff line – we could chalk it up to random variance, or just a handful of hot or cold goalies wrecking the curve.

One anomaly we can account for is the drop in all lines in 06 – it’s clear that post-lockout rule changes (especially smaller goalie equipment) depressed Sv% for a while. It’s less clear to me what caused the crazy swings in the 80’s playoffs. 84, 86 and 87 have the largest gaps on record between regular season and playoff Sv%. Then 88 had the largest negative gap. Was there a goalie flu that spring? It’s bizarre.

95 is also unusual enough that I think there should be an explanation. It’s the biggest anomaly since 88, and breaks up what would otherwise be a 20-year run of playoff Sv% outperforming the regular season (and an 18 year run of outperforming the Expected Playoff Average).

Looking at broader trends, I’ll propose that we can divide this Sv% graph into 3 eras:

  1. The Wacky 80’s (~84-88): super-volatile playoff averages.

  2. The Playoff Bonus Era (~89-07): The playoff average Sv% out-performs the EPA Sv%. 95 is the only exception.

  3. The Playoff Deficit Era (~08-present): The playoff average generally under-performs the EPA Sv%. Only in 09 and 13 does it climb above.

Because the playoff average is volatile, it’s hard to pinpoint the start/end points of these eras. For example, it may be that whatever instigated the third era began with the 05 lockout. But I do think these trends are more than a coincidence. The fact that better goalies play more in the playoffs than in the regular season somewhat obscures the trends, but if we accept that the orange line represents a reasonable expectation for the playoff average Sv%, there should be some as-yet-unidentified factor(s) causing a) a couple decades of better-than-expected playoff goaltending, and b) a more recent trend toward worse-than-expected playoff goaltending.


To sum up, here are a few unanswered questions:

  1. Why did average playoff Sv% tend to be higher than expected for so many years?

  2. Why is it now lower than expected more often than not?

  3. What was going on in the late 80s?

  4. Can we explain the very poor goaltending in the 88 and 95 playoffs?

I have a couple theories for the first two questions, but I’d be interested in readers’ thoughts as well.

F7b9382bda050115ee88dfee5421244b
Joel is an Oilers fan who took an ill-timed trip to Africa in the spring of 2006. So he's a bit sadder than the average Oilers fan.
Avatar
#1 Todd Bengert
September 23 2016, 10:48AM
Trash it!
0
trashes
Props
0
props

Great start, Joel. Lots of ideas for future analysis here. I have some comments to encourage discussion, but I'll wait till part 2.

Avatar
#2 mattyc
September 23 2016, 12:50PM
Trash it!
0
trashes
Props
1
props

Nice work. Sure looks to me like there isn't any difference in SV% in the last decade at least. Of course you could formally test this with a t.test, and see if there is any statistically significant difference. Likely it will be difficult to detect any significant difference given the variability/sample sizes.

Even if there was a significant difference in SV%, you still need to make the connection from that to goaltending performance. Alternate explanations could include "Defence box out better/shot location changes", "Team parity is less", "Playoffs subset for teams with better defense"... etc.

Comments are closed for this article.