Analytics predicts the 2016 NHL playoffs and Cup winner
The playoffs are finally here, and with them come predictions from every analyst, commentator, and regular old hockey fan. Now it’s our turn.
As we did last year, we’ve conscripted our friend, and professor of economics, Mikal Skuterud to help us develop a model to predict how this year’s playoffs will unfold.
Before getting to the fun, a word about our model. We looked at dozens of regular season statistics and the outcome of every playoff series since 2008 (a total of 120). After a lot of mixing and matching, we settled on a technique called factor analysis, which allowed us to take the 12 best variables and combine them in a single model (for those who like math, it’s the same methodology as we used last year—but with different numbers—you can read more about it here.
The 12 statistics driving the model are (in no particular order):
1. Home ice advantage
2. Total regular season points
3. Points over the last 41 games
4. Points over the last 10 games
5. Penalty kill percentage
6. Power play percentage
7. High Danger Scoring Chance Differential (from war-on-ice.com)
8. A measure of goalie quality called xSv% (created by Don't Tell Me About Heart)
9. Penalty differential
10. ESVA Fenwick (from puckon.net)
11. SRS (from hockey-reference.com)
12. DOHA Luck Score (a measure we developed that takes a team’s win percentage in one-goal games and shootouts, and subtracts the average of its win percentages in games decided by two and three or more goals; a higher DOHA Luck Score suggests unsustainable lucky results)
All of the predictions below are based on our model, not gut feelings or new information that we can’t quantify.
We also asked Ian’s daughter, who is now five years old and whose only interest in hockey continues to be based on how late the games allow her to stay up, to make her picks. You can see how she does here.
So what’s our Magic 8 Ball saying this time around?
Capitals vs. Flyers
Prediction: Capitals (69.1%)
These two teams enter the second season on opposite trajectories. The Caps stumbled to the finish line, earning only four points more than the eighth-place Flyers over the last half of the season, and two points fewer during the last 10 games. The teams are also closer than you might expect on some of the stats driving our model. They’re identical on High Danger Scoring Chance Differential and close on xSV%. But unlike past years, under Coach Barry Trotz the formerly uni-dimensional Capitals have become an excellent team at both ends of the rink and enjoy a massive advantage in penalty killing (+4.7%), and ESVA Fenwick (+1.5%). As a result, both the stats and the eye test tell us that the Caps’ superior skill and depth should be more than enough to take care of the upstart Flyers.
Penguins vs. Rangers
Prediction: Penguins (72.0%)
They say necessity is the mother of invention, and this has certainly been the case with the Penguins down the stretch. After howling at the top of our lungs back in November for the team to pair Phil Kessel with Nick Bonino, an injury to All-World center Evgeni Malkin finally forced new coach Mike Sullivan’s hand. The combination of Kessel, Bonino and Carl Hagelin has been outstanding, and when Malkin returns, assuming Sullivan keeps that trio together, Pittsburgh will have three lines that are all monster scoring threats. Not only are the Penguins playing their best hockey in years, they better the Rangers in almost every stat in our model. Pittsburgh has a High Danger Scoring Chance Differential that’s a seemingly impossible 304 chances higher than New York’s, plus a penalty kill percentage 6.2% better, and an ESVA Fenwick that’s 3.3% better. The biggest concern for the Penguins right now is the possibility that (a) Jeff Zatkoff or Tristan Jarry might be their starting goalie; or (b) a recently concussed Marc-André Fleury is between the pipes and still exhibits some serious rust or te tendency to melt down at inopportune moments. Despite the Rangers having the edge in goal (+0.8% xSV%), this projects to be the biggest mismatch of the first round. Add the fact that defenseman Ryan McDonagh is expected to miss most and perhaps all of the series, and it seems clear that even King Henrik Lundqvist won’t be enough to save the Rangers against the powerful Penguins.
Lightning vs. Red Wings
Prediction: Lightning (70.4%)
Before the playoffs started last year our model predicted that the Lightning would narrowly beat the Blackhawks in the Cup final, and building on that we had the Lightning as the preseason favorite to win the Cup. But with a series of injuries throughout the season and a generally underperforming Steven Stamkos (who is now sidelined by a blood clot), the Lightning never realized the dominance we expected. Sure, they are nine points better than the Wings over the last half of the season, but that number is probably misleading given recent injuries to key players such as Stamkos, defenseman Anton Stralman, and forward Tyler Johnson. The injuries will make the series closer than the 70.4% probability our model gives us, but Tampa Bay’s advantage in penalty killing (+2.5%), ESVA Fenwick (+2.1%), High Danger Scoring Chance Differential (+29), and xSV% (+0.6%) should carry the day against a Red Wings team that appears to be the lightweight of this year’s playoffs. Red Wings superstar Pavel Datsyuk has already made his plans to return to Russia for next season known. He likely should be booking travel for two weeks from now. Maybe sooner.
Panthers vs. Islanders
Prediction: Panthers (50.9%)
Our model says this will be the closest of the first round series, with the Panthers coming out on top. Florida set a franchise record FOR points (103), but only edged the Islanders by three, and actually had two fewer over the last half. The Islanders had a slightly better power play (+1.4%) and much better penalty killing (+5.0%), but the Panthers had a better High Danger Scoring Chance Differential (+32) and better ESVA Fenwick (+1.5%). Both teams are nursing injuries to key players. As guys who are in their 40s, we like the “ageless Jaromir Jagr” narrative as much as anyone else, but this series will basically be a coin toss. In the end, our model gives a slight edge to the Panthers.
Stars vs. Wild
Prediction: Stars (63.0%)
No big surprises here. The Stars flirted with the top of the standings for much of the season while the Wild had to fire a pretty good coach midway through and still only barely made it into the postseason. Despite Minnesota’s playoff surge, the Stars outperformed them badly in the second half (earning 14 more points total) and last 10 games (+6). Minnesota’s dismal win percentage in one goal games (0.378) suggests its record was likely worse than it should have been, but Dallas wasn’t exactly lucky either, having fared better in games won by two or more goals than ones decided by one. The Stars also had a better penalty kill (+4.4%), ESVA Fenwick (+4.0%) and were slightly stronger on High Danger Scoring Chance Differential (+10). Their Achilles heel is the fact that the mediocre goaltending tandem of Kari Lehtonen and Antti Niemi does not add up to a single good goalie. As a result, Minnesota enjoys a significant edge in xSv% (+1.2%) and will need Devan Dubnyk to stand on his head to be competitive in this series.
Blackhawks vs. Blues
Prediction: Blues (54.4%)
You never want to count the champs out, but there are lots of reasons for Blackhawk fans to not expect a deep playoff run this year. The teams were close by many measures, but for the first time in a while Chicago’s penalty kill has been mediocre, leaving St. Louis with a significant advantage (+4.8%). The Blackhawks have also been a possession juggernaut for nearly a decade. Not so this season, where they lag behind St. Louis badly (ESVA Fenwick +1.6%). While both teams have had solid goaltending this season, the Blues enjoy an advantage on xSV% (+0.5%). The only thing that suggests the Blackhawks might have been the better regular season team is St. Louis’s 0.658 win percentage in one goal games, which tells us the Blues' record may be somewhat inflated. But our model doesn’t take into account the rust from Corey Crawford’s recent injury (he was shelled by the lowly Blue Jackets his first game back) or the suspension to top defenseman and 2015 Conn Smythe-winner Duncan Keith. Sure it’s only one game, but it makes a tight series that much tighter. In the end our model predicts coach Ken Hitchcock will push his talented St. Louis team into the second round in what surely is a series on which his job depends.
Ducks vs. Predators
Prediction: Ducks (58.7%)
What a difference a year makes. Last season we had the Ducks as our punching bag, and they proved us wrong by winning two rounds before coming awfully close to taking the Western Conference Finals against the Blackhawks. This year Anaheim is a different team in nearly every respect. To begin with, as we predicted, the one-goal game magic abandoned them. Anaheim suddenly was below .500 in them and significantly better in ones that were won decisively. More encouraging, they went from a possession lightweight to a juggernaut. Unfortunately for them, because Nashville has also been strong, the Ducks enjoy only a 0.1% advantage in ESVA Fenwick. Where Anaheim has been really good is on its top-ranked penalty kill—a walloping 87.2%, which is 6.0% better than that of its opponents. Meanwhile, Nashville netminder Pekka Rinne, who for several seasons had thrived behind one of the league’s best defenses, has struggled this season, leaving the Ducks with an advantage on xSV% of +1.0%. A lot of people who had Anaheim on their preseason Cup favorites list had egg on their faces early on. They’re looking a lot smarter these days.
Kings vs. Sharks
Prediction: Kings (53.2%)
Sharks fans are going to hate us for this, but it looks like they’re in for yet another season of good but not quite good enough. After missing the last postseason entirely, the Kings appear to be back with a vengeance. The teams’ regular season point totals were close (+4 for the Kings), but in the second half the Sharks were actually up by seven points as the Kings stumbled down the stretch and conceded the top seed in the division to the Ducks. Neither team has had a great penalty kill, but the Kings still enjoy an advantage of 0.9%. They have also continued their dominance on possession, generating a league-best ESVA Fenwick of 56.9%, which is 3.9% better than San Jose’s very respectable 53.0%. L.A. also enjoys a High Danger Scoring Chance Differential of +150. This looks like yet another Battle of California that will be hard fought but is likely to tilt south for a gem of a showdown between the Ducks and Kings.
And the Stanley Cup goes to …
Although the Eastern Conference doesn’t have nearly the depth of the West, an argument can be made that the it still boasts hockey’s two best teams in the Capitals and Penguins. The Caps’ impressive 120 points were a whopping 11 more than second place Dallas. And the Penguins, who seemed to flip a switch as soon as new coach Mike Sullivan took over on December 12, have the league’s second best record (after the Ducks) over the second half of the season despite missing Malkin for the last month.
Our model predicts not only that the Eastern Conference Champion will be decided by the expected second round match-up between these powerhouses, but that these are the two teams mostly likely to win the Cup.
Specifically, our model gives Sidney Crosby and his Penguins a 16.4% chance of lifting Lord Stanley's shaving mug over their heads when all is said and done, and arch-rival Alex Ovechkin’s Capitals a 12.1% chance to do the same.
Of course since the Pens and Caps can’t actually play each other in thefFinal, we have to pick someone from the brutally deep West.
The team most likely to emerge from that side is the Ducks.
So pencil this prediction in for your hockey pool: Ducks vs. Penguins Stanley Cup Final, with Pittsburgh winning in seven.
The Department of Hockey Analytics employs advanced statistical methods and innovative approaches to better understand the game of hockey. Its three founders are Ian Cooper (@ian_doha), a lawyer, former player agent and Wharton Business School graduate; Dr. Phil Curry (@phil_doha), a professor of economics at the University of Waterloo; and IJay Palansky (@ijay_doha), a partner at the law firm of Armstrong Teasdale, former high-stakes professional poker player, and Harvard Law School graduate. Please visit us online at www.depthockeyanalytics.com. Dr. Mikal Skuterud (@mikalskuterud) is a professor of economics at the University of Waterloo.