Using analytics to figure out the second round of the Stanley Cup Playoffs
Playoff hockey is a constant reminder of how quickly momentum can change. One second a team is firing on all cylinders and before you know it the car is a smoldering husk on the side of the road.
We’re not talking about the Blackhawks or Bruins, we’re talking about the Department of Hockey Analytics’s first round predictions. On April 20, our model’s record was perfect: The Penguins clinched their series in five games and the Ducks and Predators had won in sweeps. Not only were we one of the very few to pick the Preds over the mighty Blackhawks, but our seemingly insane prediction that the Maple Leafs would take the series over the Presidents’ Trophy winning Capitals in a “coinflip” series seemed to have legs, with the series tied at two (including three games going to overtime and the teams’ stats lines standing virtually identical.
Unfortunately, the Blues, Oilers, Rangers, Senators and ultimately Capitals all refused to cooperate and won series they were supposed to lose, leaving us with a first round record of 3-5. So much for such a promising beginning.
But one round does not a playoffs make. What happens next will be interesting. We’ve been saying for a couple years now that just like any other competitive game, hockey strategies will change, teams will adapt and sometimes things that were once successful aren’t any longer. Maybe that’s what’s starting to happen to core analytics metrics like Corsi, and maybe that’s why the models are struggling so much this year. Or maybe the first round was just the kind of aberration you’d expect to happen every so often whenever you’ve got a small number of series between pretty evenly matched teams. Too early to tell. But what we can say is that we’re all looking forward to Round 2.
Crunch Time: Playoff hockey means it's time to get physical
Blues (99 points) vs. Predators (94 points)
Prediction: Predators (53.5%)
This matchup features our model’s greatest first round success against its greatest first round failure. In Round 1, we gave the eight-place Preds a solid 65% probability of eliminating the heavily favored top-seeded Blackhawks. The Preds swept. But our self-satisfaction was short-lived, as the Blues promptly and deservedly dispatched our model’s Stanley Cup favorite, the Minnesota Wild, in just five games.
Our model predicts this series to be one of the closest of the playoffs. The two teams are almost equally matched on three of the six stats in our model. The Blues have slight edges on total points (99 vs. 94) and points over the last 25 games (34 vs. 32), and the Preds have the better Event, Score and Venue Adjusted Corsi (51.4 vs. 50.6). Somewhat surprisingly given the Predators’ generally staunch defense, the Blues’ biggest advantage is their penalty kill, which was third at 84.8%, compared to Nashville’s pedestrian 80.8%. But the deciding factors appear to be Nashville’s dominance in high danger scoring chance differential (+35 vs. +1) and our DOHA luck factor, which tells us that in the regular season the Blues were by far the luckiest of the eight remaining teams and the Predators were by far the unluckiest. History tells us that kind of luck doesn’t carry into the playoffs. We predict Nashville will win in a hard fought series.
Ducks (105 points) vs Oilers (103 points)
Prediction: Ducks (59.7%)
The Ducks were the hottest team in the league over the last 25 games of the regular season and continued the momentum by sweeping Calgary in the first round. On the flip side, the Oilers, under wunderkind Connor McDavid are by far the best team Edmonton has seen in at least a decade, and bouncing last year’s Cup finalists the San Jose Sharks was well earned. But the Ducks are bigger, nastier, and arguably better than the Sharks. The stats our model uses favor the Ducks almost across the board, with the exception being the Oilers’ sizeable advantage in high danger scoring chance differential (+72 to the Ducks’ +22).
Most signs point to a decisive win by the Ducks, but that’s exactly what worries us. In the past two seasons no team has confounded our model more than Anaheim. Two years ago we predicted that they’d go down in flames in the first round and instead they made it through to the Conference Finals before losing to the Blackhawks by the narrowest of margins. Last year, we predicted that the Ducks would make it to the Final and they managed to get bounced in the first round. Add the fact that we’ve been vocal critics of Anaheim coach Randy Carlyle since his days in Toronto and we’re pretty tempted to pull the plug on the model and just go with our gut (or grit or whatever other voodoo non-analytics pundits rely on).
Unfortunately, the economist among us vetoed what the rest of us thought was a pretty solid idea. So, grudgingly, we predict not only that the Ducks will win the series, but that as of the start of the second round they have the highest probability of winning the Cup of any remaining team, at 25.6%.
Capitals (118 points) vs. Penguins (111 points)
Prediction: Capitals (56.4%)
We mentioned before that the Leafs were actually the toughest possible opponent for the Caps out of any of the playoff teams in the East and so it should now be clear sailing for them right to the Cup Final, right? Well, the defending champs might have something to say about that. Sure, it’s not predicted to be as close as the Leafs series, but 56.4% is hardly an easy victory. The difference in regular season points is not huge, and even smaller is the performance over the last 25 games (33 points for Washington, 31 for Pittsburgh). They were also both moderately “unlucky” during the regular season as they both slightly underperformed in one-goal games compared to other games. The Capitals had the better Event, Score and Venue Adjusted Corsi (54.9% to 51.5% for the Pens), but the areas where we see the biggest divergence between the teams were the two variables that contributed the most to the model – High Danger Scoring Chances and Penalty Kill Percentage. The Caps held a reasonable lead in penalty killing during the regular season (83.8% to 79.8%) but were demolished by the Pens in their high danger scoring chance plus-minus (-64 to +112).
Q and A: In matchup between Capitals and Penguins, Brooks Orpik brings unique insight
At the end of the day, however, we still want to see Ovi versus Sid the Kid. Hockey may be a team game, but the spotlight will be front-and-center on these two as we all watch to see if Ovechkin can shed his “choker” label and overcome his media-appointed arch-nemesis. The theatre will be great, and we expect the hockey will be, too.
Senators (98 points) vs. Rangers (102 points)
Prediction: Rangers (60.6%)
This is the matchup that our model thought would be the least likely out of all possible second-round series. Yet, here we are. And since we don’t learn from the past (or at least, the recent past—the model is based on the nine previous playoffs), the model still doesn’t like the Senators. Which is probably fine with them as they seem to have embraced the underdog role. In many ways, this is a matchup of underdogs, however, as these two teams had the two worst records in the last 25 games out of the 8 remaining playoff teams (30 points for the Sens and 27 for the Rangers), as well as the two worst penalty kills at 79.7% and 79.8% for Ottawa and New York, respectively (although the Rangers were tied with Pittsburgh), and the two worst ESVA Corsi at 48.2% for Ottawa and 49.4% for New York. Ottawa also had the worst high danger scoring chance plus-minus at -68, although the Rangers were positively average at +27.
We also will be treated to one of the rare occasions where the team with more regular season points does not have home-ice advantage. Home-ice is something that gets talked about a lot, but we have yet to find any statistical evidence for there being any advantage to it. (At least in the playoffs; it seems to be very important in the regular season. Go figure.) At the end of the day, one of these teams has to make it through to the conference finals, and we definitely don’t like their chances there. Of course, we’ve taken more of a beating so far than either of these teams, so don’t mind us.
Like us, Ian’s daughters have soldiered on despite a lackluster first round. If you want to see who they like for Round 2, you can find out here.
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, a lawyer, former player agent and Wharton Business School graduate; Dr. Phil Curry, a professor of economics at the University of Waterloo; and IJay Palansky, 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 is a professor of economics at the University of Waterloo.