What does the new NHL Edge data really tell us about players and teams?

I have good news for hockey fans – sports fans, really – who feel that data and “spreadsheets” are ruining their favourite game: a human element in managing, coaching, and analyzing is still crucial, to the point that it might be more important than ever to have the right humans making those decisions.

Humans have a unique set of strengths. We excel at teasing out subtexts, at combining streams of information to pull out decisions, and at interpreting that information.

These days Artificial Intelligence can do remarkable things, and there’s no shortage of computer programs and algorithms impacting hockey. From chips in pucks and jerseys to optical tracking in each rink, an incomprehensible amount of data is being created around every NHL game. And it’s precisely because of that I talk about humans here – while the computers are spitting out Kilimanjaro-sized mountains of information, creating literally millions of data points on a nightly basis, we then stand beside those mountains of information like a cat that just coughed up a hairball, looking at it and going: “OK then, what do you want to do with this now?”

As humans – nay, as hockey analysts and fans and coaches and managers – our job is to figure out how to answer that question. What does any of it mean, and for those in the game, how do you make those answers actionable in productive ways?

If you’ve followed the public hockey analytics conversation over the past 15 years, it’s looked something like this.

“This player has the most hits in the league, that’s good.”

“I agree.”

“Wait, upon further review that’s because the player’s team never has the puck when he’s on the ice, that’s bad.”

“I agree.”

“But this other player, his team gets more shots than their opponent when he’s on the ice, that’s good!”

“I agree.”

“But wait, it’s because that player plays sheltered minutes against weak lines and only starts in the offensive zone.”

“WELL WHAT DOES ANYTHING EVEN MEAN THEN?”

We’ve often taken things that looked good at first glance and figured out that well actually, it’s not that good. Having lots of hits and blocked shots can mean the other team has the puck and you’re defending too much. Score effects matter (teams that are losing often come out looking good by shot attempts because teams that are winning often settle into defending and conservative play). How players are used matters, too (zone starts and linemates and who they play against).

This is my favourite part of digging into hockey numbers: seeing some strange outlier, and wondering “Huh … what the heck does that mean?” And new information gives us plenty of new outliers.

This week the NHL launched NHL.com/EDGE, which uses their player and puck tracking to give NHL fans a ton of fascinating new statistical tidbits. And frankly, I think these tidbits are the type that most fans will find interesting (at least more interesting than “expected goals” or any stat that takes seven sentences to explain).

For example, maybe it’s not that interesting to know the fastest speed reached this season has been 38.54 km/h by Rasmus Kupari. Or maybe it is. But the league also lists “speed bursts,” including leaderboards for raw totals of 35-plus km/h speed bursts this season (as well as 32-plus km/h bursts), so we’re able to see who regularly hits the type of jaw-dropping speeds that make Alex Pietrangelo audibly mumble “oh boy” as Nathan MacKinnon rips up the ice at him through the neutral zone.

And that’s definitely interesting. I love this stuff below. To me, this is the list of players who are legitimately the league’s fastest. Any of the guys in the fastest skater contest can win any year with a few clean laps. But the guys who consistently get their speeds up in game play are the ones I want.

Here are the leaders in speed bursts over 32-plus km/h:

I was surprised to see Eichel there, a guy who skates so smooth he’s like Ernie Els swinging a golf club, where you can’t believe that relaxed action generates so much force.

Next are bursts over 35-plus km/h, which are far less common.

Funny, you see a guy like Kasperi Kapanen show up here, who we know can absolutely motor, but we can also see above that he doesn’t do it often enough to show up on the list of 32-plus km/h bursts.

Also, holy Marty Necas. Here’s a stat that immediately makes fans who don’t watch the Canes very much feel different about the player. He has double the total of 35-plus km/h bursts as the next fastest guy in MacKinnon:

Also: no McDavid, huh? That means he skates fast often (fourth in the league in 32-plus bursts), but is rarely at full throttle when doing so (he’s also unique because he does it with the puck. I’d bet my head McDavid is top-two in the league with MacKinnon in speed bursts over 30 km/h with the puck).

The hardest shot information is fascinating, too. (Excuse my while I toggle to “imperial” here, as nobody talks shot speeds in km/h.) The same ideas apply, where guys who shoot it hard most often fascinate me more than a guy who had one random hard shot. For example, on the “shots over 90 mph” list, seven of the top 10 are defencemen (you have to shoot it harder when you’re farther away to have a chance to beat a goalie). Of the mere three forwards on the list, Alex Ovechkin has four 90-plus mph shots, Steven Stamkos also has four, while Tage Thompson has 10.

You feel different about Thompson’s game now, don’t you? You didn’t realize how hard he shot the puck or how often?

So there’s tons of stuff that, on its face, is just really cool.

But it’s fun to start asking questions, too, and what I really wanted to talk about today. Do those three forwards really shoot it the hardest of all the forwards in the NHL? Or, do they just so happen to be flank one-timer options on the first power play unit (they are), so they get the chance to demonstrate their great shots more than the other 11 forwards on their team each game?

These shooters obviously rip it, as they’re ahead of other flank options in the NHL on this list. But I’m guessing some other guys can rip it, too, but they’re just watching from the bench.

I highlight this because what you’re seeing is player role having a huge effect on the numbers we’re coming across.

If you look at the leaderboard for top “skating distance” players at even strength (back to kilometres per 60, now), one thing stands out to me: the first three names are centres -- most of them are -- and there are no defencemen at all.

Centres play low in their own end all the way down to the corners at the other end, using the full 200 feet. Of course they show up here.

If you want deeper answers you can sort for position, and strength, and find guys who do a lot in their respective roles. And "role” will always be an important consideration when evaluating these lists. Even some of the players who have numerous speed bursts or high top-end speeds might be in those places on the leaderboard because the structure their team plays with asks them to be F1 on the forecheck on a line that gets plenty of ice time.

The more I thought about the information above, the more questions I had about what it all meant. What occurred to me was that this sort of player tracking information hasn’t been available for very long in hockey, but it has been in soccer.

During my early days writing hockey at theScore, I learned from a brilliant guy (shout-out Richard Whittal) that the sport of hockey and soccer had a lot in common, mostly because events occurred during consistent runs of play in both sports, rather than with the stops and starts afforded baseball and American football.

I was curious then: what does the fact that the Leafs grade out by these new measures as a “slow” team mean for their chances of success this season? Below are the Leafs' skating metrics so far this season, where you’ll see they’re “below 50^th” percentile in both top skating speed and speed bursts over 32 km/h (“below 50^th” is the NHL's way over never calling any player or team “bad” at something, which I can understand them doing). The Leafs are also 65^th percentile in skating distance.

Do good teams cover a lot of distance? Or not very much? What does this tell us about the games they’ve played so far?

I was fortunate enough to get a soccer expert on the phone to help answer some of these questions. Sam Gregory is a Canadian guy and the Director of Analytics for Inter Miami, the MLS team where Lionel Messi currently plays. He helped me come to terms with what these player tracking numbers have meant very generally for soccer, which I believe will apply to hockey as well.

The answer to the questions about distance: the data is much more about systems and styles than they are connected to winning. In fact, he says in soccer, distance run (a lot or a little) has effectively no correlation to winning.

Yet it’s not unimportant information, as it tell us which teams are playing which way, and that in itself helps us better understand the game.

For example: I’d wager man-on-man defensive teams (chasing opponents around) ask their players to travel more total distance than a zone team whose switches involve handing players off to where your next teammate is standing. Hand-offs are great for conserving energy, but miscommunications can lead to defensive breakdowns. (Tired skaters can lead to breakdowns in man-on-man, too.) This isn’t about which system is better, so much as a note: if we understand how a team plays, and knowing skating distance can point us towards that, we can discuss which players would be best suited for their system.

Carolina plays man-on-man D-zone, which they can do because they have great-skating defensive players. But they added Dmitry Orlov, who came from two systems in Washington and Boston that played a zone. He’s gotten off to a slow start in Carolina (pun not intended), and it could be because a different system is taking some getting used to.

A few more takeaways from Sam and soccer:

• Individual player measures should not be seen as indicative of their capability, but are rather just their raw output. Again, the role in which players are used is crucial, and should always be considered.

• The most useful application of the information is in player acquisition. If a team knew a role in their system requires a great skater (forechecking winger, say), that becomes a priority. If they know they play a zone defence with a patient neutral zone forecheck, maybe they instead want a smart player who can read and make switches easily.

• “Fast” teams are most typically transition teams that get turnovers and go, often sending players from the back-end up into the play. It doesn’t necessarily mean they have the fastest players.

In all, a lot of the new information is about style, from both players and teams. Yes the cream will rise to the top of the speed categories within their own teams, but a lot of what we’ll see as “top-10” lists will have been built by the coaching staffs and team styles, intentionally or otherwise.

The more info the better, and these days we’ve got piles of it.

Computers have done their job producing information and now it’s up to the humans to determine what comes next, picking up each piece one at a time, and wondering “is this something? And if so, what, exactly, is it?”