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.
Last week I took a look at the Canucks’ underlying numbers, which show that the team is far better than anyone expected heading into the 2017-18 season. Unfortunately, what everyone expected is that the Canucks would be one of the worst teams in the NHL. The Canucks have so far surpassed those low, low expectations by being thoroughly average.
The Canucks’ underlying numbers — specifically their adjusted corsi-for percentage (adjCF% and adjusted expected goals-for percentage (adjxGF%) — have actually improved over the last week, as they dominated puck possession in two one-goal losses to the Dallas Stars and New Jersey Devils.
They have improved to ninth in the NHL in adjCF% at 51.94%. They’re even better in adjxGF%: 7th in the NHL at 54.04%.
A quick recap: corsi is shot-attempt plus-minus. We can also refer to it as shot-attempt differential and it is used as a proxy for puck possession. It is frequently expressed as a percentage: getting over 50% means your team had more than half of the shot attempts at 5-on-5.
Expected goals (xG) is like corsi, but it includes a measurement of shot quality. Each shot attempt is given a value based on the estimated probability of that shot becoming a goal. This is based on shot type, location, angle, and a host of other factors from seven years worth of data. The number you get in the end is an estimate of how many goals your team should expect to score given the shot attempts they have taken.
Adjusting those numbers means taking into account the effect that the score of the game and the venue have on these two statistics.
So, the Canucks are doing better than average in both corsi and xG. That’s a good sign for the rest of the season; corsi and xG are two of the best predictors of future success.
What I want to look at now is how the Canucks are getting to those numbers. Which forward lines and defence pairings are pushing puck possession and which are costing the Canucks the most?
What I specifically want to see if which lines and pairings are outperforming their corsi and contributing to the Canucks’ solid xG. In other words, which lines and pairings are having an impact on the shot quality they create for themselves and prevent for the opposition?
The easiest way to compare a line or pairing’s overall corsi and xG: just look at the percentages. We can pull the corsi percentage and xG percentage for each line from Corsica.Hockey’s Combos page and put them together in a chart like so:
I’ve spotlighted the highest and lowest results in each column with blue for good and red for bad.
There are a couple things that immediately pop out: the 99-B Line of Baertschi, Bo, and Boeser are very good, with an xG percentage far beyond their corsi. And the three lines featuring Alex Burmistrov that have played more than 10 minutes together are all at the bottom. They may each have a corsi percentage of 50+%, but the xG numbers suggest they perform poorly when shot quality is taken into account.
Also, keep an eye on that Granlund, Sutter, and Dorsett line. They seem a smaller jump up from corsi to xG, but they’re still quite a bit better than you might expect. Their xG numbers as a line suggest that they should be out-scoring their opposition at 5-on-5, rather than their offensive production being dependent on Dorsett’s good luck.
Let’s look at the defence:
Ben Hutton and Chris Tanev appear to be for real, not only putting up positive corsi percentages, but also outperforming those already strong numbers when shot quality is taken into account. It’s alarming, however, to see Hutton and Tanev at the bottom of the chart when paired with the Canucks’ two injured defencemen, Alex Edler and Troy Stecher. Those are from smaller samples, but the way Hutton and Tanev have played together would make me hesitate about splitting them when Edler and Stecher return.
Derrick Pouliot and Alex Biega appear near the bottom, but only because it’s ranked by the difference in corsi and xG. While their xG as a pairing is a little lower than their corsi, both are still exceptional: they’re the Canucks best pairing by corsi and third best behind Hutton/Tanev and Del Zotto/Tanev in xG. Usage has to be kept in mind — they’re not being used as a shutdown pairing like those two — but they’re still performing very well together.
So, that basic comparison gives us some small insights, but what if we want to explore a little more in depth? For instance, we might assume that the 99-B Line has a significantly better xG percentage than corsi percentage because of their offence, particularly the way they create chances off the rush, but how do we check? Or we might assume that the difference for Hutton and Tanev comes in the defensive zone, limiting the shot quality of opponents by keeping them to the outside, but we can’t know that for sure from just the percentages.
We can look at corsi and xG by their components: for and against. The problem is that comparing shot attempts to expected goals doesn’t really work. They’re on two different scales.
For instance, the line of Granlund, Sutter, and Dorsett has been on the ice for 85 shot attempts against (CA) at 5-on-5. Their expected goals against (xGA) is 3.21. Is that good or bad? Is that higher or lower than what we would expect given that amount of shot attempts? We don’t really know.
But we can find out.
Over the last three NHL seasons, there have been a total of 322,494 shot attempts at 5-on-5. Of those 322,494 shot attempts, 13,127 have gone into the net. That gives us an idea of how likely it is, on average, that a shot attempt will turn into a goal: 4.07%. That’s ignoring any elements of shot quality: location, type of shot, rush, rebound, etc.
What this gives us is a way of comparing a team/player/line’s corsi and a team/player/line’s expected goals. We can multiply 4.07% by shot attempts and get a number that we might call corsi expected goals (CxG).
Think of it as a simplified version of expected goals. Whereas xG treats every shot on its individual merits — location, shot type, angle, whether it’s off the rush or a rebound — giving each attempt a unique likelihood of becoming a goal, CxG treats every shot attempt as having the exact same likelihood: 4.07%.
The upshot is that this simplified metric allows us to directly compare corsi and xG. Let’s go back to the example of the Granlund, Sutter, and Dorsett line: they have been on the ice for 85 shot attempts against at 5-on-5. 4.07% of 85 gives us their CxG: 3.46.
All that means is that if we took 85 randomly selected NHL shot attempts, we would expect 3.46 of them to go into the net.
That line’s actual xG is 3.21, so it would seem that the shutdown line that Travis Green has deployed against the top forward lines of the opposition actually does limit opponents’ shot quality, if only by a little bit.
Let’s look at all the forward lines:
There are some surprises here, specifically when we look at the 99-B Line. Their biggest impact is actually on shot quality against, not on shot quality for like you might expect. The difference between their corsi expected goals against and their actual expected goals against is more than one goal. That’s huge and suggests they’re much stronger defensively than people think.
The Sedins and Virtanen are the line that sees the biggest jump between corsi for and expected goals for, which makes a lot of sense. Not only do the Sedins still create good chances in the offensive zone, but Virtanen adds more of a threat off the rush as well.
What might be truly surprising is the Granlund, Sutter, Dorsett line, whose shot quality for is legitimately quite a bit higher than expected. They’re not far behind the Sedins and Virtanen in terms of creating higher quality chances than their corsi would suggest.
What this suggests is that the shutdown line isn’t just limiting the quality chances for the best forwards the opposition has to offer, but also creating quality chances of their own. In fact, their ability to create quality chances seems greater than their ability to limit them.
The line of Vanek, Burmistrov, and Gagner, on the other hand, looks painful by this measure. That line has looked a little bit better recently, so they might turn things around, but they currently give up higher quality chances and produce lower quality chances than their (quite good) corsi together would indicate.
Now for the defence pairs:
Del Zotto and Gudbranson momentarily stand out here for a good reason rather than a bad one. They seem to legitimately limit the shot quality of their opponents as compared to their corsi. Here’s the issue: they give up so many shot attempts against that their xGA is still very, very high. They also have a lower xGF than you would expect from their corsi.
If you were looking for proof that Gudbranson limits shot quality against with his defensive play, this would support that argument. But when you’re giving up that many shot attempts, the quality of those shot attempts starts to matter less and less.
On the plus side, there’s Hutton and Tanev, who surprise by improving their own team’s shot quality as much as they limit their opponents’. That’s unexpected, but can be looked at as both Hutton’s influence on Tanev and Tanev’s commitment to improving his offensive game over the off-season.
Together, Hutton and Tanev’s offensive and defensive numbers suggest a swing of 1.3 goals to the positive via shot quality so far this season.
Other notable elements of this chart: all three pairings that have featured Derrick Pouliot this season have given up higher than average shot quality against. It matters the least on his pairing with Biega, as they have such excellent corsi in their sheltered minutes that shot quality makes minimal difference, but it’s something to keep an eye on.
So what conclusions can we draw here? To be honest, I’m not quite sure. I don’t know whether this difference between CxG and xG is repeatable, ie. whether the 99-B Line or Hutton/Tanev will continue to influence shot quality in this way over the course of an entire season.
It does provide some numbers to back the eye test for how the Canucks are creating and preventing quality chances and it provides some ammunition for a potential argument for keeping Hutton and Tanev together long term.
It also makes the shutdown line of Granlund, Sutter, and Dorsett look better than just pure corsi numbers, as they manage to create quality chances as well as limit them for their opponents.
In terms of defence, keeping the puck in the offensive zone and out of your own zone entirely is preferable, but if you are going to allow shots, it’s best for them to come from low-danger areas. So far this season, the Canucks have been good at doing both. They have the lowest rate of shot attempts against in the NHL and the second-lowest rate of expected goals against.
That kind of team defence should make the Canucks a tough opponent for any team in the NHL.
