Skip to content
Join our Newsletter

Numbers Never Lie: Have the Canucks been lucky to start the 2017-18 season?

PDO and why Derek Dorsett has been unlucky.
Numbers Never Lie Banner

While numbers never lie, it’s sometimes difficult to figure out exactly what truth they’re telling. To help you out, Numbers Never Lie is a weekly look at the Canucks’ advanced statistics, and figuring out exactly what they have to say about the Canucks’ season and players.


One of the oddest complaints I’ve seen about analytics recently was that the analytics community treats sports like a math equation and doesn’t account for luck. Really, luck is one of the things that hockey analytics have spent the most time talking about, because it plays such a big role in the outcome of games.

In fact, what hockey analytics has generally argued is that luck plays a much larger role in hockey than most hockey fans realize.

Michael Maubossin, author of The Success Equation, looked into how much luck and skill impact regular season results for all the major North American sports and found that hockey is the closest to random. According to his work, skill explains less than half of the results of the NHL regular season.

A big part of that is the number of scoring plays in a game. In the NBA, where teams score nearly 100 points in a game, the better team wins the game almost every time. In the NHL, however, there are far fewer scoring plays, which introduces more chance for randomness to impact results.

Last season, there was an average of 5.45 goals scored per game. Considering how the puck — a relatively tiny piece of vulcanized rubber — can bounce, deflect, roll and skip around the ice, several of those 5.45 goals per game are likely to be lucky goals. Just look at the Canucks’ last game against the Los Angeles Kings: Henrik Sedin’s goal, the one that started the comeback, was a centring pass that deflected off Nick Shore’s leg and past Jonathan Quick. It was a lucky goal.

In some ways, however, teams can create their own luck. If you take more shots than your opponent, you give yourself more opportunities for the luck to fall in your favour. That’s one of the reasons why the analytics community gravitated to a statistic like corsi that measures the quantity of shot attempts.

The more shot attempts you can take over your opponent, the more you can take luck out of the game: picture a hypothetical game where one team takes 50 shot attempts and the other takes 25. Not only is the team with 50 shot attempts more likely to have a bounce go their way, they’re also more likely to have their skill result in a goal or two.

Of course, there’s a lot of parity in the NHL, so many games are a lot close to a coin flip, with both teams having a more-or-less equal opportunity for luck to go their way. Honestly, that randomness is part of why hockey is so much fun to watch: the underdog always has a chance in hockey, whereas in basketball, underdogs have almost no hope of an upset.

So, how do we measure luck in hockey? One of the earliest steps in accounting for luck is still one of the most intuitive and useful: PDO.

PDO’s origins make it hard to explain — PDO doesn’t stand for anything but was just the username of the person who came up with the concept — but what it measures is not. It’s simply shooting percentage and save percentage added together.

We could call it Percentage Driven Outcomes if you want PDO to stand for something, because none of the alternative names for PDO have stuck.

Here’s how PDO works: every shot on goal in the NHL results in either a save or a goal. There are only two outcomes. Therefore, if we add together the shooting percentage and save percentage for the entire league, we will get 100%: all shots accounted for.

That gives us a baseline or average with which to compare every team in the NHL. If a team’s PDO is higher than 100%, that means a higher-than-average number of shots have gone in their favour: either more of their shots than average have gone in the opposing net, or more of the opposing team’s shots than average have been saved, or both.

We tend to only look at 5-on-5 for PDO, as both team systems and randomness play such a big role in special teams play that it can skew results. Over the course of the season, however, a team’s 5-on-5 PDO tends to regress towards 100%. If it’s higher or lower than 100% early in the season, it will tend to even out as the season progresses.

It’s not a perfect regression to 100%, of course. There’s still a little bit of skill involved. A team with above-average goaltending or a lot of shooters that score on a higher percentage of their shots than average will naturally have a PDO a little bit higher than 100%. But it’s not by much.

The Washington Capitals last season finished with a 102.9% PDO thanks to the best goaltending in the league and a team full of some of the best finishers in the league. At the lower range was the Colorado Avalanche at 97.0%, a combination of the worst goaltending in the league and well-below average shooting percentage.

That gives us a range to work with: approximately 3% above and below that 100% average. If we see a team above or below that 3% mark, we can be nearly certain that they will regress over the course of the season to somewhere within that range. But even teams that are still within that range are more likely to regress to near 100%.

This works for players as well, though there is a wider range. The highest PDO among players who played at least 500 minutes last season belonged to Jason Zucker: 105.6%. The lowest was Patrick Sharp at 94.0%. That’s closer to a 6% swing from the 100% league average.

Again, if an individual player is above or below that 6% mark, we can be almost certain they will regress, but even if they are within that range, they’re more likely to regress to near 100% by the end of the season.

What does this mean for the Canucks this season?

We can look at PDO for the Canucks at both the team and individual level to get an idea for whether the Canucks have been lucky or unlucky this season and whether that is likely to continue.

The Canucks’ 5-on-5 PDO is currently 101.9%, which is above average, but not unreasonably so. There is a chance, though not a high chance, that they could maintain these percentages over the course of the season.

When we break it into the individual percentages, we get a clearer picture of where the Canucks have been lucky and unlucky.

Team SH% SV% PDO
Vancouver Canucks 7.69 94.18 101.9%
Los Angeles Kings 8.35 93.02 101.4%
Vegas Golden Knights 9.76 91.58 101.3%
Calgary Flames 7.19 93.57 100.8%
Anaheim Ducks 8.18 92.06 100.2%
Edmonton Oilers 5.94 92.66 98.6%
San Jose Sharks 6.03 92.38 98.4%
Arizona Coyotes 6.94 90.15 97.1%

 

The Canucks 101.9% PDO is the highest in the Pacific Division, with most of it coming from their league-best 5-on-5 save percentage. That’s concerning, as their goaltending is likely to regress, not just because Jacob Markstrom and Anders Nilsson are untested as number one goaltenders, but because that level of play is unsustainable for pretty much any goaltender.

The good news is that their shooting percentage is a little below league-average. That means their 5-on-5 goalscoring is unlikely to get any worse and could, in fact, improve.

All this is to say that on a team level, you shouldn’t be surprised if the Canucks’ goaltending falters at 5-on-5 in the coming months, making it more difficult for the Canucks to out-score their opponents at even-strength. That makes sorting out their special teams more vitally important.

On an individual player level, PDO can help us determine who has been lucky, as well as who has been snakebitten and is likely to turn things around. In fact, we just saw a trade that gives us an excellent example: Mike Cammalleri for Jussi Jokinen.

Cammalleri had a decent start to the season, with 7 points in 15 games, while Jokinen has been part of the problem in Edmonton, with just one point, an assist, in 14 games. Except, Jokinen hasn’t been that bad, really.

Jokinen has a 58.3% corsi on the season — the Oilers have taken 58.3% of the shot attempts at 5-on-5 with him on the ice — but his PDO is 93.2%, one of the worst marks in the league. Cammalleri, on the other hand, has a 43.6% corsi and a 102.4% PDO.

The Kings got the better puck possession player whose on-ice percentages are likely to regress to near 100% by the end of the season. Jokinen is likely to start putting up points, whereas Cammalleri is more likely to see his production dry up.

It’s not that cut-and-dry, of course. Cammalleri could line up with Connor McDavid, for instance, and have the better player prop up both his corsi and his PDO for the rest of the season. But it becomes a lot easier to see why the Kings made a trade while at the top of the Pacific Division: they’re banking on Jokinen’s luck changing.

For the Canucks, let’s start with their forwards.

Player On-Ice SH% On-Ice SV% PDO
Bo Horvat 10.28 96.46 106.7%
Loui Eriksson 6.67 100.00 106.7%
Sven Baertschi 11.32 95.00 106.3%
Jake Virtanen 7.58 96.77 104.3%
Brock Boeser 11.49 92.13 103.6%
Alexander Burmistrov 6.94 95.35 102.3%
Thomas Vanek 6.45 95.45 101.9%
Daniel Sedin 7.34 93.06 100.4%
Sam Gagner 3.92 96.26 100.2%
Brandon Sutter 7.61 92.50 100.1%
Henrik Sedin 7.02 92.96 100.0%
Derek Dorsett 8.24 91.35 99.6%
Markus Granlund 5.26 91.51 96.8%

 

The thing that immediately jumps out at the top of this list is who isn’t there: Derek Dorsett. With his hot streak to start the season, you might expect Dorsett to lead the Canucks in PDO, but he actually has the second lowest PDO among Canucks forwards. You could argue, in fact, that he’s been unlucky.

Of course, this shows his on-ice shooting percentage, ie. the shooting percentage of the entire team when he’s on the ice at 5-on-5. His individual shooting percentage is still an unsustainably high 29.2%. So, Dorsett’s been lucky to score as many goals as he has to start the season, but arguably a little unlucky that he hasn’t picked up more assists.

Right down at the bottom is Markus Granlund, whose on-ice shooting percentage is just 5.92%. A combination of playing in a shutdown role with a healthy dose of bad luck has limited him to just 3 points this season. With Brendan Gaunce’s return to the lineup, Granlund has been moved to a more offensive role; combine that with a little regression and he should start putting up a few more points.

Another unlucky player is Sam Gagner, who has a team-worst 3.92 on-ice shooting percentage. Gagner has been doing a lot of the right things — taking the puck hard to the net, setting up scoring chances, cycling the puck down low — and hasn’t been rewarded. He is second on the Canucks in shots on goal with 43 and is a likely candidate to see his luck turn around.

It’s concerning to see Bo Horvat and Sven Baertschi so high on this list at 106.7% and 106.3%, respectively. Brock Boeser also has the highest on-ice shooting percentage on the team. Their 5-on-5 production is likely to take a bit of a hit in the coming months, which is where they’ll need to make up ground on the power play. The new first unit with Boeser at the left faceoff circle and Horvat parked in front of the net could help.

As for Loui Eriksson, the Canucks have yet to concede a goal when he’s on the ice at 5-on-5. That’s not likely to continue, but his low on-ice shooting percentage should regress a little bit upwards. He’s looked effective with the Sedins and the points may start to come for the much-maligned Swede.

Now a look at the Canucks defence.

Player On-Ice SH% On-Ice SV% PDO
Christopher Tanev 13.33 91.43 104.8%
Alexander Edler 7.50 96.97 104.5%
Ben Hutton 10.24 93.43 103.7%
Derrick Pouliot 6.36 95.40 101.8%
Erik Gudbranson 4.12 97.14 101.3%
Alex Biega 5.00 95.95 100.9%
Troy Stecher 7.55 92.16 99.7%
Michael Del Zotto 5.96 92.68 98.6%

 

There are fewer extremes here among the defence. The most interesting result for me is Chris Tanev finishing at the top of the PDO charts while having the lowest on-ice save percentage. Both his on-ice shooting and save percentages should regress towards the mean, which should result in both fewer goals for and against when he’s on the ice.

Leading the way in on-ice save percentage is Erik Gudbranson, which is a concern. There are some that argue defencemen can influence save percentage, so perhaps some of this is skill, but a .971 5-on-5 save percentage is unsustainably high. Perhaps it’s luck paying him back for the unsustainably low .897 on-ice save percentage he dealt with last season.

Alex Edler has an unsustainably high on-ice save percentage as well at .970, so look for that to regress and a few more pucks to go in when he’s on the ice.

When it comes to bad luck, Gudbranson shows up again with the lowest on-ice shooting percentage among Canucks defencemen and second-lowest behind Sam Gagner on the Canucks in general. As his on-ice save percentage regresses down, the shooting percentage should regress up.

The unluckiest defenceman has been Michael Del Zotto, though not so unlucky that he’s guaranteed to regress upward. Travis Green has relied heavily on Del Zotto — he leads all Canucks defencemen in ice-time — and that combined with his PDO have led to a team-worst minus-8 plus/minus.

There’s a reason analytics have dismissed plus/minus, and it’s partly because of luck and PDO and partly because plus/minus doesn’t limit itself to 5-on-5 situations, including things like shorthanded and empty net goals. Del Zotto doesn’t really deserve to be minus-8 on the season, although his 44.87% corsi suggests that he shouldn’t have a positive goal differential either. His 5-on-5 goal differential of minus-3 makes a lot more sense.

So, that gives us some candidates for regression. The Canucks as a whole are likely to fall back a bit at 5-on-5. Bo Horvat and Sven Baertschi’s 5-on-5 play could take a hit, while Markus Granlund could bounce back. Chris Tanev will likely see his on-ice save percentage recover, while fewer goals go in for the Canucks with him on the ice.

The biggest question mark for me is Brock Boeser. His 11.49% on-ice shooting percentage may or may not be sustainable. There were players who maintained an on-ice shooting percentage that high last season, but they were few and far between.

Just four forwards played 500+ minutes last season and posted an on-ice shooting percentage above 11%: Patrick Laine, Jason Zucker, Mark Scheifele, and T.J. Oshie. Could Boeser be one of those rare forwards who influences on-ice shooting percentage to that extent? It’s hard to say because we have no history to draw on for him. He’s a rookie, so we can’t look at his past results. It’s unlikely, but not impossible.

Nothing’s a guarantee in hockey, of course. There’s far too much randomness involved.