Goalies and the Playoffs Part 4: How to Win a Cup

Updated: May 24, 2018 at 2:37 am by Joel Short

Previously, I looked into the effect that PDO (a team’s Sv% plus its Sh%) has on the results of individual playoff series (tl;dr: a big effect).  Next, I wanted to track that effect over multiple playoff rounds.

The average PDO for best-of-7 series winners since 19841 is pretty consistent round to round:

Round 1: 1.024
Round 2: 1.029
Round 3: 1.027
Round 4: 1.027

Those are high numbers.  To put them in perspective, only 6 teams in the past 8 seasons have posted a Regular Season PDO (what NHL.com calls SPSv%) of 1.024 or more, and just 3 have reached 1.029.  We can confidently say that if playoff series were longer, these averages would be lower.  (The average PDO of series winners in Round 1 from 1984-86 – which were best of 5 series – was a stratospheric 1.051.)  Of course, the shorter a stretch of games you look at, the more normal variance will influence the results.

The chart below shows the average PDO for teams in each round of the playoffs, grouped by how far they advanced.  The error bars show the highest and lowest team PDO in each category.


You can see that on average, when teams lose a playoff series they post a dramatically lower PDO than they did in previous (winning) rounds.  In fact, the relationship between a team’s PDO from one round to the next is basically nil.  This is clear when we compare the PDOs of series winners to their PDOs in the following round:

PDO Rd n+1

Each dot above represents a team that won at least 1 round.  Its horizontal position shows the team’s PDO in a series they won, and its vertical position is their PDO in the next round.  No team is to the left of .965 – the minimum PDO that every series winner since 1984 has achieved – and a third are right of 1.035 – the point beyond which no team has lost.2  The area between the blue lines is what I’m calling the Competitive PDO Zone.

There’s nothing magical about the exact borders of the Competitive PDO Zone (no doubt, a team will someday win a series with a PDO below .965) and of course competitiveness is a matter of degrees – it’s a lot easier to overcome a PDO of .990 than .970.  But I find it helpful to have an idea of where a daunting disadvantage approaches insurmountable.

Note that “approaching insurmountable” isn’t a rare event.  There have been 438 best-of-7 series in rounds 1-3 since 1984.  146 of them have featured a team with a PDO of 1.035 or more – exactly a third – and every one of them won.

I’ll say that again: in one out of every three playoff series, the winning team’s PDO is better than any series-loser ever since 1984.  Sounds like goaltending wins championships, right?  Well, of those 146 teams, in the next round:

  • 54% lost, perhaps because some weaker teams were carried through by heavy PDO luck
  • 49% posted a PDO below 1.000
  • 16% were above 1.035 again, essentially guaranteeing victory
  • 16% were below .965, essentially guaranteeing defeat

All of which reinforces the point that PDO is hugely influenced by luck in samples as small as 4 to 7 games.  Short runs of PDO dominance do not predict future PDO dominance.

Why Teams Lose in the Finals

It’s interesting to consider what went wrong for teams that won three playoff series before losing in the finals.  Did you guess that PDO had something to do with it?  The chart below shows teams’ finals-losing PDOs in red rings, with PDO from each previous round in solid red.

It’s striking that in 30 out of 33 years, the PDO posted in the finals by the losing team was worse than their PDO in each of the 3 previous rounds.  Only 9 of 33 runners-up had a PDO below 1.000 in any of the first 3 rounds, but in the finals, only 7 had a PDO above 1.000.  In fact, 11 of 33 (one third, again) fell beneath the Competitive PDO Zone in the finals.

When a team is eliminated from the playoffs, commentators like to ask what they need to do to get to the next level.  If their PDO dropped steeply in their losing round, most of what they need is better luck.

It’s interesting to compare this chart to one for Stanley Cup winners.  As we saw in the first chart, Cup winners and runners-up post similar PDOs through the first 3 rounds, but champs (like winners in all rounds) tend to keep up that luck in the finals.

More Cup winners than runners-up were below 1.000 in at least one of the first 3 rounds (15 vs 9), but 16 of 33 Cup winners were at 1.000 or higher in all four rounds.  It’s fair to say that these teams benefited from some of good fortune (which doesn’t necessarily mean that they would have lost without that good fortune).  The other side of that coin is that 52% of Cup Winners overcame a PDO below 1.000 in at least 1 round, demonstrating that they could best a strong opponent even when luck turns against them.

I should also note that the tendency of PDO to regress heavily to 1.000 is not an iron law, but largely an effect of the parity of the modern NHL.  The dynasties of the 70s and 80s posted PDOs above 1.020 for several seasons in a row.  So the ’85 Oilers (whose lowest series PDO was 1.032, but who also went 15W-3L and outscored their opponents by 40 goals) were more good than lucky.  That said, here are some interesting facts about Cup Winners:

  • In 33 years, there’s never been one with a total playoff PDO below 1.000.  The 09 Pens came closest (1.006).
  • The Oilers were over 1.040 for 4 of their 5 championships.  Colorado is the only other team to crack 1.040 twice (Roy might have had something to do with that).
  • Of the past 10 Champs, only the ’13 Blackhawks had a lower PDO in the playoffs than the regular season.  The average champions over that span had a Regular Season PDO of 1.005, and a Playoff PDO of 1.021.
  • 7 of the past 10 Champs were within 10 points of par in the regular season (including 3 in the .990s).

The fact that recent Cup winners average a PDO of 1.005 in the regular season suggests that many were skilled enough (particularly in net) to have a “PDO equilibrium” above 1.000.  No one is saying PDO is 100% luck.  But as a rule, if you’re playing after the 80s, your goalie isn’t Patrick Roy, and your PDO is above 1.020, you’re getting pretty lucky.

Quantifying the PDO Advantage

I’ve talked a lot about insurmountable PDO advantages, and their surprising prevalence in best-of-seven series.  But it’s interesting to look at the impact of more modest PDO advantages as well.  The following chart shows how often teams with PDOs below 1.000 won a playoff series, broken down into .010 chunks.  Keep in mind, a difference of .010 translates to an extra goal scored per 100 shots.  The number of shots per team in a playoff series varies quite a bit, but the average is 170.  So in an average series, a goalie who bests his opponent’s Sv% by about .006 gives his team a 1 goal edge.  (In an average 4 game series, ~.008 difference is worth 1 goal.  For 7 games, it’s ~.005.)

  PDO Range  Incidence
(% of all series)
  Lower PDO Won 
1.00-.990 112 (24%) 45%
.990-.980 94 (20%) 20%
.980-.970 79 (17%) 13%
.970-.960 52 (11%) 8%
<.960 134 (28%) 0%

You can see that in a playoff series, PDOs less than .010 below par represent a slight but meaningful disadvantage.  But as PDOs fall below .990, your chances of winning the series drop rapidly.

What does it all mean?

It’s clear that PDO has a very large impact on the outcome of playoff series – a much larger impact than any comparable stat.  I struggle a bit with interpreting this.  I don’t think anyone believes that hockey is a pure meritocracy, but I do think we tend to look for explanations in skill or heart or determination for a lot of the randomness in sports – often a single bounce the other way could completely change the post-game narrative.  It’s been established that luck plays a bigger role in the outcome of hockey games than other major sports (although baseball’s more random in the playoffs), and a part of me finds that disappointing.

On the other hand, the large impact of luck on the outcomes of NHL games and playoff series is due in part to the parity of the cap-era league.  And while there’s always a chance of an upset in the modern NHL, you’ve got to be both lucky and good to win the Cup.  The alternative is a league like the NBA, in which a team can be given a 65% change of winning the title before the regular season’s even finished (as of April 1 2018).

So I say: hockey fans, embrace the randomness.  Celebrate the inextricable mixture of skill and luck that carries your team to victory.  And when you lose, go ahead and blame bad luck and the refs.  But don’t tell me your team “just wanted it more” if they win with a PDO of 1.040.


1 As before, I’ve picked 1984 as my starting date, mostly because that’s when the NHL began officially recording Sv%.  (Also, I’m an Oilers fan.)  However, I’ve excluded the first round of 84, 85 and 86 because it was then a best-of-5.

2 To be precise, the highest PDO by a series loser was 1.0352, by Chicago vs. Detroit in Round 3 of 1995.  The pedants among you are welcome to mentally add that fourth decimal place any time I mention the number 1.035.

Goalies and the Playoffs: Series Home

, ,