The reason on-ice shot differential is so important to evaluating hockey is not because it is an end-all and be-all statistic. The reason shot differential carries so much influence is because it’s the best (publicly available) statistic that predicts future goal scoring.
Expected goals arose as a way to better describe the quality of the shots teams take. Expected goals take into account how close to the net a player was when he took a shot, whether the shot came on a rush or was a rebound shot. Expected goals isn’t a perfect number, either, but it can tell us a whole lot about the quality of chances a team generates. Shot differential, combined with expected goal (shot quality) differential, and luck explain about 85 percent of winning in hockey.
Maybe the most beautiful thing of all is that these types of numbers contain the nuances the eye test purports to see. Brent Burns may turn over the puck often, and not all of Erik Karlsson’s stretch passes connect. In either case, those two are doing enough positive things to make up for any lapses we remember seeing while watching the games. How do we know? Because, at the end of the night or stretch of games, we can see that those players help the Sharks produce more shots and chances than they give up, on average.
At 1:21 of this highlight reel, Brent Burns sends a nice pass to Marcus Sorensen, who skates in for a dangerous chance on goal. In the box score, that entire play just shows up as one shot on goal for Sorensen. If we use on-ice shot and expected goal differential, Burns gets a tick mark in both of those columns. Shot differential and expected goal differential do a better job of rewarding players who are catalysts.
We can also see how, on occasion, those figures do too much to detract from a player. At 2:05 in that video, Carolina dumps the puck in. Aaron Dell plays it to Karlsson, who has an immediate decision to make as a Hurricane forechecker comes skating in. He can try to turn up-ice with the puck while his goalie is behind the net, or he can send the puck as far away from the incoming forward as possible, back from where it came. In this moment, Karlsson tries to bank it off the end boards back the other way. The puck takes a bad bounce, Brenden Dillon can’t control it, and it ends up in the back of the Sharks’ net. For trying to do the right thing, Karlsson ends up with a goal against, an on-ice shot or two against and some fraction of on-ice expected goal against.
Does Karlsson deserve some of the blame for the shot and chance against? Yes, at least partly. Does he deserve even the majority of the blame for the goal against? Absolutely not. For that goal to happen, a whole sequence of events must occur. Dell has to play a puck to Karlsson in a dangerous situation. Karlsson’s pass has to take a weird bounce off the boards, and Dillon has to inadvertently knock the puck right into the crease as Dell is nonchalantly sliding back into place, unaware of where the puck is the entire time. Meanwhile, all of San Jose’s forwards have to decide not to support their defenders during the supposed break out.
Goals in the NHL are mostly random. That sequence from the Carolina game is a wonderful example of how much has to go wrong or even just slightly askew for a goal to happen. To help remove improper blame and improper credit for those random events, we look at a larger sample size of events: shots and expected goals. At the end of the night, Erik Karlsson’s shot and expected goal differentials include the sequence we just watched. And, at the end of the night, Erik Karlsson still stood atop the Sharks shot differential leaderboard and held the fourth-best expected goals differential mark. What those numbers say is that, even though Karlsson was on the ice for that weird goal and plenty of shots and chances against, he did more than most of his teammates did at the other end of the ice to make up for it.
In fact, Karlsson has been doing plenty since the season began. And Karlsson paired with Marc-Edouard Vlasic has been one of the most impressive defense pairs so far this season. We can compare the performance of those two with those of defense pairs over the past few seasons to see where they rank in a few categories. One note: in the article cited above written for The Athletic, A closer look at Corsi, how much it matters, and what it tells us about the Jets’ season so far, Garret Hohl describes what “luck” means in the context of winning games:
Even if every team was perfectly equal in talent, effort, or any other tangible and intangible, and each game merely came down to a coin-toss, we would still find some teams outperforming others in the standings. In two separate cases, we have seen that the noise of variance accounts for about one-third of the variation in the standings.
To help control their destinies as much as possible, hockey teams should strive to control shots and expected goals/scoring chances/shot quality. The other key to remember here is that Karlsson and Vlasic have only played about nine games worth of 5-on-5 ice time together. As a result, there is plenty of luck (variance) attached to their performance so far. Their results will likely regress somewhat as they play (hopefully) more games together, but the levels they’ve reached can tell us quite a bit about how they’ve performed together.
During the four seasons including and between 2014-15 and 2017-18, 294 defense pairs played at least 400 minutes together at 5-on-5. That leaves us with about 74 pairs per season and 2.5 pairs per team each season.
Here is how the Sharks number-one defense pair compares to those 294 pairs from the past four seasons. The only major discrepancies are the percentage of defense zone starts relative to their team and the expected goals they generate relative to their team. The d-zone starts might have an impact with this few games but shouldn’t impact their other numbers much in the long run. It’s there for context.
They aren’t generating quite as many dangerous chances as the rest of the Sharks are, but the pair is still doing just fine in that regard. In terms of limiting shot volume against, they are just barely out of the 98th percentile. In essence, Marc-Edouard Vlasic and Erik Karlsson have been better than 98 percent of defense pairs to play substantial minutes together during the four seasons preceding this current season.
Shot and expected goal differential don’t tell us the whole story, but they tell us about half of it. Keeping in mind that variance is playing a role, especially this early in the season, we can see that the shot and chance numbers paint a very positive picture of these two skaters’ on-ice collaboration. They haven’t just been good this season, they’ve been dominant.