Heading into the 2017-18 season, both hockey pundits and analytical models alike were agreed: the Canucks were going to be bad. They were absolutely right.
What’s remarkable is how right they were. Last year I looked at eight different projections — a mix of traditional hockey publications like The Hockey News and new-fangled analytics like Micah Blake McCurdy’s “Edgar” model. On average, they predicted 75 points for the Canucks in 2017-18.
The Canucks finished with 73 points.
The 2018-19 season isn’t that far away — the start of the Canucks Young Stars tournament, which kicks off their training camp, is just two weeks off — so we’ll start to see some predictions rolling in pretty soon. In fact, we don’t have to wait. Emmanuel “Manny” Perry, the creator of the indispensable Corsica.Hockey, released his projections for the 2018-19 season this week.
NHL regular season team points projections for 2018-2019 (result of 10,000 simulations): pic.twitter.com/IgjUaa3nB8— Emmanuel Perry (@manny_hockey) August 22, 2018
To clarify, these are not Manny’s final projections. His projections will be updated as rosters become more clear heading into the season, but these initial projections are still intriguing, primarily because of how accurate his projections were last season.
His article on Corsica about his 2017 version of his model, nicknamed “Salad,” provides a lot of insight into how he puts together his models, though some significant knowledge of mathematics and statistics is necessary to make sense of some of the higher-level aspects of the article.
Manny takes his model and simulates the season thousands of times. His projections are the average, both mean and median, of those simulated seasons.
Ultimately, his “Salad” model was significantly more accurate than other projections of the 2017-18 season, both those of industry experts and those of other statistical models. On average, his predictions were 10.9 points off from the actual results. The next best model was nearly a point worse, which is a fairly significant margin.
In terms of accuracy, it doesn’t get much better than his projection for the Washington Capitals: 105 points, their exact number of actual points. The model also had them as the best team in the league and, while that wasn’t true in the regular season, it turned out to be so in the playoffs.
The model was dead-on with three other point projections and within five points on 13 of 31 teams. It also had some significant misses — it thought the Montreal Canadiens and Ottawa Senators would be 20 points better than they ended up being and whiffed on the massive improvements made by the Winnipeg Jets, New Jersey Devils, and Colorado Avalanche — but even with those in mind, the model turned out to be remarkably accurate. If you were a betting man and used Manny's model to bet on the NHL, you would have made a tidy sum last season.
That’s bad news for the 2018-19 Canucks, as Manny’s new and improved model has them once again well outside the playoff picture.
The biggest issue is that the news isn’t quite bad enough.
Manny’s model projects that the 2018-19 Canucks will finish with 78 points, good for 27th in the NHL and second-last in the Western Conference. That’s a far cry from the playoffs, but also depressingly far from last place and the best odds for the first-overall pick. Instead, it would have them lined up for the fifth overall pick and a potential fall to eighth overall.
That sounds about right.
Of course, hockey is incredibly difficult to predict accurately, as demonstrated by the big misses from Manny’s model last season. Hockey is subject to more variance than most sports, meaning that the results of any given game are subject to significant randomness. In many ways, that’s what makes hockey great: the “objectively” best team doesn’t always win, leaving plenty of room for an underdog story.
Could the Canucks be one of the teams that outperforms their projections?
Manny’s model projected 77 points for the New Jersey Devils last year; they finished with 97 and made the playoffs. Sure, the Devils' improvement required an MVP season from Taylor Hall, but the Canucks have significant young players looking to make the roster, from Elias Pettersson and Jonathan Dahlen, to Petrus Palmu and Olli Juolevi. Those kinds of untested rookies can wreak havoc with statistical models, as their performance is so difficult to predict.
Does that mean the Canucks will be in the playoff picture by the end of the season? Let's just say, don't put any money on it.