Which Teams are Relying on the Rush?
I tracked every 5v5 goal so far this season to find out
At this point, most analytically-minded hockey fans recognize that there is a distinction between the NHL data that is collected for public consumption (known as “play-by-play data”) and the far more extensive information collected by private companies like SportLogiq and Stathletes as well as NHL teams themselves. While tremendous value can come from intelligent analysis of what is available to the public, private data is expensive for a reason, and I am always interested when we get limited glimpses of what’s behind the curtain.
As anyone who’s read a good amount of my work knows, certain members of the public have heroically filled that gap a bit by manually tracking NHL games themselves and making the data available to fans. Corey Sznajder’s Tableau and Patreon display mountains of information that we would otherwise be without, and they have been an invaluable resource for anybody trying to work out the “why” behind player results. Inspired by this, I decided to take on a relatively limited but hopefully revealing project: tracking every 5v5 goal of the 2020-21 season so far based on whether or not it came off the rush.
My hope with this was to gain some understanding of how teams and players were generating and giving up goals, and possibly even learn why certain teams were significantly under- or over-performing their expected goals so far. While public expected goals models have rush chances baked into them (those shots and goals are part of the data that goes into building them), because of data limitations they cannot identify whether a shot came off an odd-man rush or breakaway or what have you.
I defined a rush goal as being a goal where a controlled zone entry occurred and a goal was scored within five seconds without being given away to the opposing team. Non-rush goals typically came off the forecheck or the cycle.
To make the most of this tracking, which would have been basically impossible (or at least have taken up many more hours of my life) without InStatHockey, I decided to link my team-level tracking to the NHL’s play-by-play goal data as assembled by EvolvingHockey.com. This gave me the ability to look not only at which teams scored but who allowed them, who was on the ice for both teams, and individual player and goalie stats.
In this piece, I’ll be focusing on team-level metrics. In a few days, I’ll follow up with player stats.
Team Stats
What I was most interested to see was which teams scored most of their goals off the rush, and which teams gave up the most goals off the rush. Let’s start out with the stat I was most focused on: which teams scored the largest share of their total 5v5 goals off of rush chances?
It’s the Capitals. Running away. The Caps lead the league in goals above expected by a wide margin right now, and this points to a possible reason why: they are the only team that has scored more than half of the goals off of high-percentage rush opportunities. Transition-oriented teams like the Avalanche and Leafs also rank unsurprisingly highly here, as do the Montréal Canadiens. Would it surprise you to know that 43% of their rush goals came against the Canucks?
At the other end, the Coyotes are by far the lowest on this list, relying almost wholly on the forecheck and cycle to generate goals. The Sabres have scored few goals this year, and only a quarter of them have come off of rush chances - Taylor Hall and Jack Eichel’s inability to get the puck in the net this year has something to do with that. And the terminally cycle-reliant Canes are near the bottom of the list here too.
Let’s reverse this and look at how teams are allowing goals against?
While I was expecting to see a number of teams that were playing high-flying, chance-trading hockey, that didn’t necessarily pan out. The Capitals, Leafs, and Avalanche, three of the most rush-heavy teams, also rank among the best at limiting rush goals relative to non-rush ones. But it should be no surprise to see the stingy Dallas Stars shutting down opponents’ transition play. At the other end… Vancouver’s struggles preventing odd-man rush chances have been well-documented but I was surprised to see Tampa Bay, Calgary, and even Nashville so far down the list.
Let’s reframe this a little bit and focus on the number of rush goals instead of the proportion of them. Here are the top rush scoring teams (per 60 minutes of 5v5 play):
Buffalo drops to the bottom because no other team is allowed to rank worst in the league in a goal-related stat. And the Oilers jump all the way to 3rd. How about goals against?
The Calgary Flames are allowing three times the number of rush goals as the Dallas Stars are, and the Flyers and Canucks are not far behind. Meanwhile near the top there’s a nice combination of low-event teams like the Coyotes, Stars, and Hurricanes and rush-dominant ones.
Let’s combine those two stats and see who’s winning the rush goal battle. This stat will function like Goals For Percentage does - it shows the team’s share of rush goals during their games (Rush Goals For divided by [Rush Goals For + Rush Goals Against]):
The Avalanche have a 59% goals for percentage at 5v5 and the difference-maker appears to be their dominance on the rush. They have scored almost 70% of the rush goals in their games, as have the Capitals and Maple Leafs. And at the bottom… oh god… the miserable Sabres have been completely destroyed by the extent to which they’re getting clobbered off the rush and unable to generate any goals in return.
Let’s say you love firewagon hockey - lots of rush goals happening overall, regardless of who gets ‘em. Here are the teams who have the most rush goals happen in their games per 60 minutes:
If you like chaos, watch the Flyers and Canucks. If you like forechecking, definitely check out the Arizona Coyotes.
What about non-rush goals?
There were some surprises here - the Coyotes, Flames, Hurricanes, and Blues are all teams that score very little off the rush but they compensate for it by scoring more frequently off the cycle. The Red Wings, Predators, and Sabres just can’t score period.
How about non-rush goals against?
I’m not the only person tracking rush-related stats this season. The aforementioned Corey Sznajder has begun to track whether shots came off the rush, forecheck, or cycle this season and is putting together similar data at a larger scale. There is a difference between the two data sets. Mine is “complete” because I had far fewer things to track - I tracked every 5v5 goal whereas Corey, being one man, has not tracked every shot from every game so far this season. As with all of his stats, he is progressively tracking more and more to the extent that it will become a stronger and stronger approximation of the whole as the season goes along, but my data and his aren’t perfectly compatible. That being said, I did think it might be interesting to compare his “Rush Shot %” data with my “Rush Goals %” data.
Conclusion
While I undertook this project this weekend, some interesting research was presented at the Ottawa Hockey Analytics Conference by Meghan Chayka of Stathletes:
This confirmed that score effects play a significant role in how teams generate scoring chances - a team that’s winning will enter a shell and give up more chances off the cycle (up to 15% more compared to tied), but the increased aggressiveness of their opponent will allow them to generate more off the rush (up to ~17% more). With this in mind, how much of a role does being a team like the Sabres who’s always losing have to do with them losing the rush battle so acutely?
Using the play-by-play data, I checked this out using the small sample of 1920 5v5 goals I had on hand - much smaller, sufficeth to say, than the war-chest of 5v5 shots Stathletes has at their disposal. And the relationship is definitely visible, although not dramatic enough to lead me to think that a massive score-effects adjustment is necessarily needed for these findings to have any descriptive value:
Something else I was curious about was the xGoal of these goals compared to non-rush goals. My hypothesis going in was that a good chunk of the non-luck portion of outperforming expected goals at either end of the ice was generating off the rush. It makes intuitive sense - if the model can’t identify an odd-man situation, a cross-crease onetimer, a breakaway, etc. then it probably is going to underrate those chances. I found a decent relationship between Rush Goals Percentage and Goals Scored Above Expected at the team level in this small sample - a 0.51 R and 0.26 R^2. The relationship between Rush Goals Against Percentage and Goals Saved Above Expected was much much weaker. This would tentatively suggest that while generating more off the rush could be a reason that a team is outscoring expectations, the goalies might not be able to use more rush chances against as an excuse.
Next up, I’ll be looking at some player and goalie stats to see who’s generating (and giving up) goals off the rush.