At Saturday night’s all-star game, we got our first glimpse of how coming puck- and player-tracking data can be used to enhance a broadcast for the audience at home.
#HiTechHockey pic.twitter.com/PlrOrabErt
— NHL on NBC (@NHLonNBCSports) January 26, 2019
Using transmitters in players’ shoulder pads and inside the puck, the system captures the location and identity for each entity in each moment. The all-star game used this data to show us overlays for who each player is, how long their shifts have lasted, and how fast their shots are, but that is just the tip of the iceberg. This information will totally transform how coaching and management decisions are made.
Here are some important, practical questions that can be answered using the new data. These are questions that might be somewhat answerable now, but the new data will introduce a previously impossible economy of scale while reducing data-entry errors, human bias, and rink-to-rink variations. There’s no guarantee that we will ever see this data (and it’s not entirely clear who, the players or the league, will own the data), but this is what I’d want to know if I were an analyst or coach.
Pretty simply, we’ll know how far John Carlson skated per shift, per game, and per season. We’ll know if he skated 20 percent more than usual last night and therefore should get today off. We’ll also know who is chasing icings more than his linemates and who chooses to conserve their energy instead. We’ll find out if there’s a correlation between skating distance and point production on the macro scale or in the seconds before goals are scored.
We’ll know every player’s speed, for example, in the 15 meters between the blue lines. We’ll know average speeds per age and position, if one player’s speed has fallen off by 20 percent since last season, or how much speed a big-minute player has lost from October to April — or between the start and end of a too-long shift. We’ll know if we should choose a fast player like Brianna Decker to get big stretch passes. We’ll know the difference between Tom Wilson’s top speed and his acceleration, who’s got the most powerful first stride in the league, and who might be nursing a groin injury.
(There are major health and labor implications here, and I wonder if they’ve been considered fully.)
If you take one shot that has a ten-percent chance of going in and another shot that has a four-percent chance of going in, then you’ve created 0.14 expected goals (i.e. 10 plus 4). The math to figure out those percentages now are based on data we already have: shot location, shot type (e.g. wrist vs slapshot), etc. With new puck-tracking data, we’ll know a lot more: for example, if the puck crossed the slot (“royal road”) just before the shot was taken.
One practical application: we’ll be able to differentiate the danger of an Ovi Shot from the Ovi Spot vs other shots taken from that same location because we’ll know it’s a one-timer following a pass from the far side of the ice, which forces the goalie and defense to adjust as it comes in.
We’ll also be able to improve our understanding of shot quality by considering the position of players in front of the goalie and net. Was Rask able to see Giroux’s shot, or was Simmonds blocking his view? If so, maybe Simmonds should get some credit, even if he doesn’t get a point. And did Rask have to move from left to right to accommodate the screen? And was the shot coming off a three-on-one rush or were all the defenders back?
And by extension, we can compare how certain formulations of power plays work against certain formulations of penalty kills. For example, does shadowing Ovechkin make the Capitals’ umbrella setup less effective, or does it just make the slot more vulnerable? What should the Caps do tactically when Ovechkin has a shadow?
Right now we know which player takes more shots on his line, but soon we’ll know which player is carrying the puck and for how long — either in total seconds or as a percentage of ice time. We’ll know if Andre Burakovsky defers his puck-carrying duties when he’s with Lars Eller but carries it more in other circumstances. We’ll know which zone of the ice each player carries the puck in, and we’ll know which defenders are puck-movers, which are puck-carriers, or to what extent of each they are.
We’ll know if Evgeny Kuznetsov’s drop passes high in the offensive zone are actually succeeding. We’ll know who makes a good stretch pass and who is handcuffing his teammates. We’ll know which two players exchange passes successfully and productively, and which two players have no chemistry at all. We’ll have a history of all their passes and what happens immediately after the pass was either made or blown. We’ll know which passes are high-risk/high-reward and which are just plain bad based on their success rates and what happens after similar passes in the seconds following.
Right now we know when and where and by whom shots are taken, but the action between the blue lines is almost nonexistent. New tracking data will tell us about zone exits and zone entries without a massive (yet noble) manual data-collection project. We will know which players carry the puck out of their zone and into the opponent’s zone a lot and which ones prefer to pass it. We’ll know that Marcus Johansson succeeds in his entries, that Nate Schmidt succeeds at defending against them, and what happens when they’re head-to-head. We’ll be able to identify players’ traits (e.g. this one is strong with entries, this one needs support in neutral) and create combinations accordingly. And we’ll have all of the passing information I discussed above in the context of blue lines.
We know when a shot is a rebound because it takes place soon after another shot, but now we’ll know the direction and velocity of a blocked or saved shot. We’ll know when Braden Holtby is making a safe block towards the walls and when he is serving up the puck in the slot for a second chance. The same deal for blocked shots: did Brent Seabrook’s block really reduce the risk of giving up a goal or just postpone it?
Right now it’s very hard to know much about the effectiveness of defensive defenseman, but player tracking will help. We’ll know how far away Matt Niskanen or Marc-Edouard Vlasic stay from puck carriers when backchecking, and how effectively one or the other limits a pass or shot. We’ll know when a defender gets beat and blown past by the attacker because, all of a sudden, the defender is no longer located between the puck and the net.
A faceoff win means the player’s team has possession of the puck after the faceoff. That has implied that the faceoff taker needs his teammates to get the win, but now we’ll know who exactly is the one retrieving those pucks, and the extent to which a faceoff specialist like Jay Beagle is reliant on a winger to secure possession for him.
The NHL currently tracks giveaways and takeaways, but it’s notoriously unreliable. With player- and puck-tracking data, we’ll know clearly that Patrice Bergeron was carrying the puck, then Brooks Orpik was carrying it. We’ll know the ratio of a player’s giveaways to puck-carrying ice time — who is stronger on the puck and who is more vulnerable, as well as who is a sneaky puck-stealer.
If we know how fast two players are moving, we can also measure the impact of them hitting one another. We’ll know if Tom Wilson hits harder than Milan Lucic and how much harder. We’ll be able to learn how concussions are related to the change in velocity of the hits that caused them. Plus, if we know where the puck is when the hit occurs, and exactly how late the late hit is — and if Antoine Roussel makes more hits 0.25 seconds after a player loses the puck than anyone else in the league.
Most of these questions can be answered now, but they require lots of manual tracking to do so. And even then, that analysis is often just for one player or game or team — without establishing a baseline of the whole league to compare against. There are many new opportunities here, and my questions just scrape the surface. There’s a lot we won’t know until a community of data scientists are sharing and reviewing one another’s research. Either way, there’s a revolution coming, finally combining anecdotal analysis with information that will help us (or, um, help authorized data licensees) understand a chaotic game in a more coherent way.
Russian Machine Never Breaks is not associated with the Washington Capitals; Monumental Sports, the NHL, or its properties. Not even a little bit.
All original content on russianmachineneverbreaks.com is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)– unless otherwise stated or superseded by another license. You are free to share, copy, and remix this content so long as it is attributed, done for noncommercial purposes, and done so under a license similar to this one.
Share On