NHL Stanley Cup Champions

What does it take to win it all in the NHL? Where do NHL Stanley Cup champions stand with respect to their peers on key offensive and defensive regular season metrics? Where does each season’s champion fall in the seasonal distribution of these metrics? How large an effect do generational players (Gretzky, Lemieux, Roy, maybe Crosby, and, yes, Hasek) have on their team’s performance? As lifelong fans of the Edmonton Oilers, we also really, really wanted to know how long teams spend in the wilderness eating locusts and wild honey. So we created this data-driven history of (almost) three decades of hockey. It shows the fundamental shifts in the way the game is played, the teams that have innovated and defined the modern game (thank you, Scotty Bowman!), the teams that have languished, and the teams that have won it all in different eras. You should really play around with the visualization. Lots of interesting patterns. Explanations and observations are below.

Other Visualizations You Might Like: A Data-Driven Baby Name Finder

What are all these little blue dots?

The big blue dots are teams that won the Stanley Cup. The little blue dots are all the other teams. The vertical axis represents different metrics (e.g., shot differential, save percentage, etc.). Because the dots show where teams stood on a given metric in a given season, you’re seeing the distribution of the metric oriented vertically. The little horizontal bars represent the median value for a given metric in a given season. So, taken together, you’re seeing how the metric (and teams) evolved over the past two decades.

Where are the Atlanta Thrashers and the Minnesota North Stars ?

We use the current city and name for each team. The Winnipeg Jets became the Phoenix Coyotes who became the Arizona Coyotes. The Atlanta Thrashers became the Winnipeg Jets. The Minnesota North Stars became the Dallas Stars.

How can I interact with the visualization?

You can click a team name to highlight its performance over time; you can also highlight a team by clicking on the dots. Hovering over a dot or circle will show you the value of a team’s metrics in a season. You can explore other metrics categories by clicking on the tabs at the top of the visualization.

What does this tell me about Corsi/Fenwick and possession?

Shot differential, (Sf – Sa)/Gp, is higly correlated with Corsi and Fenwick, stats that the hockey analytics community uses to measure possession, or how well a team drives the play toward the opponent’s goal. (The analytics community views possession as a repeatable skill because team Corsi and Fenwick tend to be modestly correlated from season to season. Shooting percentage and save percentage are seen more as luck rather than skill factors because they have poor correlation from season to season.) The shot differential distributions show that, after the end of the Oilers era and after Lemieux’s first cancer scare (1992-93), Stanley Cup champions were often (but not always) high possession teams. So, a definite yes: possession matters.

Is possession everything?

Maybe not everything. The Wings and the Devils rocked possession numbers while carrying middle of the road shooting percentage and save percentage numbers. They were content to win the shot battle and saw off the rate at which shots were converted to goals, because, you know, the math works out. Possession worked for the Wings and the Devils but not for all teams that followed their lead. High possession teams that couldn’t manage to stay near or above the middle of the pack on shooting and save percentage didn’t have as much success (check the Bruins in the early to mid-90s or the Hurricanes following their Cup win, the always good but never good enough San Jose Sharks). Whenever Stanley Cup winners had below the median numbers on the sum of shooting and save percentage, they had to make it up by either having stupid good possession numbers (a couple of the Devils and Wings teams, the first Hawks team) or by being really good with save percentage (the Kings teams). Now, with more of the top-tier teams getting into the high possession game, recent champions (the Kings and the Hawks) have started to target save percentage as the next variable to optimize. The game evolves.

Are shooting percentage and save percentage measures of skill or nothing but luck?

The argument from the stats community is that these metrics don’t correlate from season to season and that teams tend to bounce around the league average. Any team that comes out of nowhere and posts unexpectedly high shooting or save percentages in one season is likely to regress in the next season. You’ll see many instances of this phenomena in the visualizations (the Carolina Hurricanes of 2005-06 are a great example), but you’ll also see examples of teams setting a new level and staying above or below the median over a number of years. These are teams that fans would agree had high level, even historical, talent or teams that lagged in talent and were going through a building or a rebuilding phase. Looking over the whole range of data, you can easily make the case that shooting and save percentage are skill or efficiency measures, because when teams had exceptionally high (low) skill, they maintained a level of excellence (despair) on percentage metrics for a number of years. We’ll be digging more into this somewhere down the road.

Have the goalies won?

Sure looks like it. Now, we’re not sure to what extent save percentage is a team or individual metric. But goalies are bigger, wear more equipment, and they have better technique. And median save percentage keeps pushing higher. There’s much talk about addressing this with rule changes. The league probably realizes that the way the game is played has limited its appeal to the casual U.S. fan. We say give the shooters the posts and the crossbar, that is, increase net size by exactly those dimensions.

So what’s the story the data is telling us?

In the beginning (or starting in the 1987-88 season), there was offense. Lots of goals, high shooting percentages, and median save percentages well below 0.90. Possession mattered less to championship teams like the Oilers and the Penguins because they had generational players who were more efficient in converting chances — even the Gretzky-less 1990 Oilers were pretty efficient. During this time, Mike Keenan was in Chicago coaching a high possession game, seeing success, but unable to break through against the great offensive teams of the era.

1993-94 was an important year. Lemieux played only 22 games after undergoing his second back surgery and Gretzky’s Kings missed the playoffs. Keenan took over the Rangers, immediately converted them to a high possession team and led them to the Stanley Cup. Meanwhile, Scotty Bowman took over the Red Wings and Jacques Lemaire took over the Devils; both coaches implemented high possession games. (Bowman did a nice pivot here when he no longer had the offensive genius of Lemiueux to work with.) In Buffalo, Dominik Hasek was elevated to the starting job, transformed his game and (along with Patrick Roy, who had won the Cup the year before with the Canadiens) transformed the way goal was played. And, in 1993-94, total scoring took a big step down; within 2 years, total goals per game had dropped from above 7 to below 6.

The rest of the 90s and early 2000s were dominated by the Wings and the Devils, playing their possession game and clogging the neutral zone. The Avalanche and later the Stars provided a nice counterpoint by consistently playing a more efficient style (higher save and shooting percentages). Following the strike season in 2004-05, goals spiked under new rules designed to open up the game, and it looked like the possession-efficiency battle could tilt the other way when the Hurricanes came out on top in 2005-06. It didn’t happen. The possession game returned in force, with the Bowman-influenced Wings and Hawks, and the Kings (the new New Jersey Devils) all finding success in the following years.

Why credit Scotty Bowman?

Scotty Bowman’s Detroit Red Wings were the prototypical high possession team, appearing in 4 Stanley Cups and winning 3 during his tenure as coach. He became an advisor to the Chicago Blackhawks prior to the 2008-09 season. The Hawks almost immediately started to resemble the Wings in their emphasis on possession. They’ve won 3 Stanley Cups since. The New Jersey Devils of the 90s were another strong team that consistently won the possession battle. They were coached by Jacques Lemaire and Larry Robinson, who played for Bowman with the great 70s Canadiens teams. So we’re OK thinking that Lemaire and Robinson’s philosophies to the game may have been shaped by Bowman as well.

Is there only one way to win the Stanley Cup?

There are 3 dimensions of team performance: shot difference, representing possession, and shooting and save percentage, representing skill or efficiency. Bowmanesque teams think winning means possession because you can’t count on an efficiency advantage over your opponent. They’re content to be mediocre in efficiency as long as they can establish a large enough margin on shot difference, because they know that the math will work out. But the math can also work out for high efficiency teams that are mediocre on possession, like Colorado’s Roy, Sakic, and Forsberg-led teams. They were an elite team for many years and won two Stanley Cups. And Crosby’s Penguins, the Hurricanes, the Bruins: all high efficiency, mediocre possession teams that hoisted the Cup in recent years.

Not everyone can be Hasek, but what if you have Hasek?

Generational players undermine the shooting percentage and save percentage are luck argument. Look at the Sabres with Hasek, the Penguins with Lemieux (and to a lesser extent with Crosby), the Kings once they had Gretzky, Roy with the Habs and the Avalanche. These players brought sustained success for their teams on efficiency metrics. The throwback game of Sidney Crosby’s Penguins (high scoring efficiency, low possession), at the beginning of the current possession-focused era, is a further testament to the system-breaking effect of generational talent. (But the Penguins have moved to the possession mainstream since then and their offensive efficiency numbers have tanked.) In a league where the high end talent is evenly matched, there may be no efficiency edge; you’re flipping coins. But what do you do when you have a coin that you know comes up heads more often than anyone else’s coin? Over to you, Peter Chiarelli.

How about a glossary?