The Vancouver Canucks have had one of the least successful power plays in the NHL through their first 26 games and yet they’re likely to have one of the best power plays in the NHL over their final 56 games. Confused? Read on.
People have a tendency in sports and life to assume that past performance is the best predictor of future success. This isn’t always true. One of the major focuses of analytical research in hockey has been to sort through those instances in which the past foretells the future and those instances in which it doesn’t. The goal in doing this is to try and isolate differences that permit us to understand when a poor performance is likely to continue and when we should expect a change.
One area in which analysts have had considerable success is in understanding power plays and whether success or failure is likely to continue. There are a couple of simple principles but, before we get to those, a word needs to be said about the typical range of power-play performance.
Discussion about power plays in the analytics community is centered on five-on-four play. The vast majority of power-play time is 5v4, which means that there’s too little 5v3 or 4v3 time to really do much with it analytically. Moreover, the 5v3 and 4v3 time isn’t evenly distributed. The most sensible thing to do is just to leave it out of the analysis. So power-play analysis is really 5v4 analysis.
Goals in hockey are a product of two things: shot volume and the percentage of shots that go in. In order to measure a team’s efficiency at generating shots, rate stats are used—shots are expressed on a per-60-minute basis. Here’s how things shook out over the course of the last six seasons:
Fifty-eight percent of teams during this period generated between 45.0 and 52.4 5v4 S/60, an awfully narrow window. But a few teams stand out as being exceptionally good at generating 5v4 shots: Washington had three seasons at 52.5 5v4 S/60 or better, Detroit had four such seasons and San Jose and Anaheim did it a remarkable five times. At the other end of the scale, nobody can hold a candle to the Edmonton Oilers, who managed the remarkable feat of putting together six seasons in a row with fewer than 45 S/60 at 5v4. (As someone who watched something like 80 percent of those games, I can assure you that it’s more spectacular in the abstract than it is when you’re watching it, game after game after game.)
The range of shooting percentages is also pretty tight. Here’s the range over the past six seasons:
About 62 percent of teams have been within the 11–14 percent range and more than 82 percent are between 10–15 percent.
As of the conclusion of their game on Monday, the Canucks were taking 62.4 5v4 S/60, the second-best number in the NHL. As you can infer from the range of 5v4 S/60 rates above, it’s an exceptional number. They’re 28th in the NHL in 5v4 G/60 though, because they’re last in 5v4 S%, shooting an abysmal 5.5 percent.
This is where the work that’s been done with hockey analytics can teach us something about what to expect going forward. As it turns out, shooting percentage at 5v4 tends to be extremely volatile. Teams are susceptible to extended stretches in which their shooting percentage is quite high or low. We can see this if we compare first and second half 5v4 shooting percentages in 2012–13. As you look at this, keep in mind that the league average was about 12.8 percent.
It’s a dog’s breakfast—the Ducks and Blues had unbelievable shooting percentages in the first half of the season and were below average the rest of the way. The Islanders suffered a large drop. The Sabres and Blue Jackets were abysmal during the first half and then saw their shooting percentages soar. The Maple Leafs were below average in the first half and dominant late. Teams are all over the map.
But 2012-13 wasn’t atypical—shooting percentage tends to work like that, even from year to year. There’s no guarantee that teams with a high shooting percentage in one season will have one the following year.
Shot rates, by contrast, tend to be very consistent from one half to the next and even from year to year. Teams don’t have extended stretches in which they generate 60+ S/60 unless they’re teams that are good at generating 5v4 shots, as evidenced by being able to continue doing it.
Applying this to what we’ve seen from Vancouver this year, the likelihood is that the Canucks will have a shooting percentage that is around league average for the rest of the season. What’s more, because they’re an excellent team at generating 5v4 shots, an average shooting percentage on the power play will result in more goals than average, making them one of the better 5v4 teams in the NHL for the balance of the season.
There’s a broader point here, a philosophical one about process versus results. There are things that are within a hockey team’s control and things that aren’t. One of the key concepts in sports analytics is that the things under a team’s control tend to repeat. If teams that are good or bad at something tend to continue being good or bad at it, then it’s a skill. If they don’t, then it’s an element of luck that’s contained within the sport.
If shooting percentage at 5v4 doesn’t tend to repeat—that is, if the randomness of pucks hitting skates and going in or goalies making great saves overwhelms the skill that’s present—then you have to ask yourself whether evaluating a power play on whether it scores goals makes sense. When you’re judging whether something’s working, it makes more sense to judge it on the basis of those things that are within a team’s control than to incorporate things into your analysis that are outside the control of the team.
Judged by goals (which are driven by shooting percentage), the Canucks have been awful at 5v4 this year. Judged by their shot rate, which seems to be something that’s more within the control of the team, they’ve been very good. The better assessment is probably that the Canucks are a very good 5v4 team that hasn’t had a lot of bounces this year. Look for them to start scoring a lot more goals with the man advantage in the immediate future.