Arguably the greatest contribution of hockey’s analytics revolution is its ability to subject longstanding articles of faith to cold hard logic.
Articles of faith have their appeal. Because if you say something often enough without thinking about it much, it eventually feels true.
Occasionally, even those who traffic in pure reason fall into that trap.
One example of this is the cult of depth. Whether you’re listening to young analytics bloggers or old-time commentators, eventually you’ll hear someone talk about how in today’s NHL teams have to be deep in order to win.
Pundits of all stripes routinely laud the Blackhawks and Kings for their remarkable depth and abuse the Penguins for imagining some hanger-on wingers and a bunch of spare parts are enough as long as they have Sidney Crosby and Evgeni Malkin.
The language of old time hockey (“grit!”, “determination!”, “momentum!”) may be different from what the new guard’s peddling (“weak bottom 6 possession!”, “unsustainability!”, “mean reversion!”), but the argument’s the same: deep teams win, shallow ones don’t, and if you don’t already know whether a team’s deep or thin, you must be an idiot.
No doubt everyone would love to have lots of great players. I’ve certainly played video games where I get an NHL all-star team and my nephew gets the Kazakh national team (and still beats me).
But in the real world teams have to make trade offs. For example, they need to ask whether two top 50 scorers are as good as one guy in the top 5? Or how much they value an elite shutdown defenseman like Marc-Edouard Vlasic vs. a one-dimensional forward like Phil Kessel.
Despite the conventional wisdom (or maybe because of it) nobody’s analyzed whether more depth means more winning. In fact, most commentators don’t even bother to define what they mean by “depth”.
One way of thinking about whether or not a team is deep is to look at ice time. Presumably coaches with deeper rosters are willing to allocate ice time more evenly and the best players get more minutes.
There are two problems with this. First, some players log minutes because they’re just that good, while others play because their teammates are just that bad.
Take a guess (no cheating) who leads all NHL forwards in average time on ice (21:15). Sidney Crosby? Ryan Getzlaf? Alex Ovechkin?
Nope. It’s Ryan Nugent-Hopkins.
That’s right – fans who used to watch a one-two punch of Gretzky and Messier can now cheer for Nugent-Hopkins and whichever 3rd or 4th line Centre the Oilers allow to play the 2nd.
The fact that the guy logging the most minutes in the league sits 93rd among forwards in total points says more about how truly bad the Oilers are than how good he is.
But looking at overall ice time can be misleading in other ways.
A team that’s playing from behind is more likely to gamble; one that regularly trounces opponents will roll 4 lines late in the game.
Which is why advanced stats are tracked in all sorts of different situations.
If we’re talking about depth, the most important one is 5-on-5 close play, meaning the game is within 1 goal during the first 2 periods or tied in the third. After all, if a coach is willing to throw guys out in close games, that suggests he trusts them.
Looking at ice time in those situations can yield new insights. For example, you might conclude the Rangers’ “go to guy” is Derek Stepan, who leads the team’s forwards with an average of 18:09 per game. In fact, during 5-on-5 close play, Stepan is 7th, averaging 7:29 per game.
But once you start looking at all of a team’s forwards, you’re very quickly into a jumble of indistinguishable numbers. For example, if Carolina’s Eric Staal (1st on the team) logs 9:17 vs. Zach Boychuk’s 6:19 (12th) is that a big difference or a little one?
To figure that out, we looked at a common statistical measure called standard deviation, which measures the spread in ice time among each team’s top 12 forwards. All things being equal, we’d expect deeper teams to have a lower standard deviation (i.e. less of a spread in ice time between their top 12).
The table below shows each team’s ranking by that measure of depth, as well as its points percentage (total points earned as a percentage of total points available).
Depth Among Forwards (Summary)
There were some surprises for sure.
For example, the Blackhawks rely heavily on their top guys when the game’s close. When measured by the standard deviation of their top 12 forwards’ ice time, Chicago ranks 21st in the league, a fact that suggests coach Joel Quenneville doesn’t think as highly of players who aren’t named Kane, Toews, Sharp or Hossa as others might.
More important, when we looked at the correlation between standard deviation and points percentage, it was positive, meaning teams that were relatively “thinner” did better. The correlation was admittedly small (0.06), meaning the relationship wasn’t strong, but it does mean being “top heavy” doesn’t necessarily hurt if your top guys really are that good.
Not only is “depth” hard to define, it may not be a necessary ingredient to win either.