Over the holiday break, I was catching up on some tracking, particularly with shot assists and puck recoveries. I’ve said it before, but I find tracking hockey games to be a very effective space to think deeply about the game. One of the things that has come to my mind all season is the residual effects of shots, or what happens after you shoot the puck. When you shoot the puck, there are three outcomes:
Stoppage of play
Puck recovery
Loss of possession
Sometimes, there is a stoppage in play because you scored. Other times, there is a stoppage in play because the goaltender froze the puck. There are many different events that are covered by this umbrella. If there is not a stoppage in play, then there is a loose puck. This simple fact of the game should guide much of your thinking surrounding shot selection, offensive and defensive zone structures, and more.
Focusing on one residual outcome, and to build upon my last article, rebounds are forever an interesting topic in hockey analytics and coaching alike. Starting from a coaching perspective, “Getting bodies to the front of the net,” or, “Scoring a dirty goal,” are buzzwords I find coaches often use to describe trying to create high danger shots through chaos as opposed to “skill.” This is fine as it’s a concise, illustrative verbiage which is important in leadership, but I do think it skips over some of the inherent tactical decisions you are making.
Let’s take an example of a point shot stemming from a low-to-high pass. In a more modern offensive zone system, the structure should more or less resemble a 1-3-1. In the scenario in which the “bumper” player is above the faceoff dots, it will look something like this.
In this structure, F2 and D2 are in positions where they are free to engage in puck recovery efforts in the corners because they know they will have the support of F3. If the puck goes all of the way up the wall, there may also be an opportunity to quickly attack as F3 is in an offensive position inside the dot line. Where this structure is weak is in its ability to recover soft loose pucks in dangerous areas around the net.
When you go to put two at the net (I kept the labelling the same for illustrative purposes), you are creating a much larger defensive triangle. This puts much more pressure on each player to read, react, and cover ground in the event of a loss of possession.
When coaches and analysts talk about paying the price for going to the front of the net, they are typically referring to the physicality. The cross checks from an opponent or the potential to get hit by the puck. From a mathematical perspective, I think the price you’re paying to try to create a rebound or deflection in close is the increased likelihood your opponent can exit the zone and create a chance against you if you miss. To me, something feels right about that. Game situation and personnel will dictate how you balance conservative, possession dominance from the outside versus chaos creation in tight.
From an analytics perspective, rebounds create some of the most interesting philosophical discussions. For the average person, detailed discussions surrounding expected goals are boring. Who wants to listen or read about gradient boosting? But rebounds are an exception because they are the highest quality shot in hockey.
If you are reading this, you’ve probably seen an expected goals chart that illustrates each team’s expected goals over the course of the game. And it’s probably stuck out like a sore thumb to you when you see a spike of roughly 1 goal over the course of a few seconds and no goal was scored. That’s caused by rebounds! Let’s look at the Leafs-Kings game from last week.
As you see, there is a very large spike in the Kings expected goals around a third of the way through the third period. This all occurred within a matter of seconds from one shooter, Quinton Byfield.
The initial shot comes from point blank directly off of a pass that caused a quick change in angle. The expected goals model on Evolving Hockey had this shot as a 34.1% chance of becoming a goal. Intuitively, I think that’s a little high, but nothing that seems outlandish.
The initial shot was followed up by an initial rebound attempt, given a 28.5% chance of becoming a goal, followed by another rebound attempt given a chance of 25.2% chance of becoming a goal. Add those together and you get 0.88 expected goals in a matter of three seconds.
This should feel inflated. From a mathematical perspective, the second and third shots in this sequence - the rebounds - don’t happen if the shot before them goes in. You can dilute the expected goals of the sequence by multiplying them by the chance there isn’t a goal before them. It goes something like this.
By flurry adjusting, you bring the sequence down from 0.88 expected goals to 0.65. This is still a high danger sequence but brings the expected goals of the sequence down to a more accepted level.
There are players around the league that are notorious for boosting their expected goals with putting rebounds into a goalie’s pads like this. For example, Brady Tkachuk has 106.8 career expected goals at 5v5 but 90 actual goals. This lands him as the 12th-worst shooting forward since he joined the league in 2018-19. I reckon some of that is due to him smashing rebounds into pads. How do we think about that in the context of flurry adjustments? Should we be adjusting his individual expected goals, reducing them and making him look better, or should we let him be punished for not finishing those rebounds he is credited for?
I don’t like flurry adjustments for two reasons. First, I don’t think it fits with what an expected goals model is, evaluating the likelihood of a goal for each individual shot. Flurry adjustments revolve around sequencing certain events; therefore, the shots are no longer being evaluated independently. Secondly, hockey is an interconnected game. Each event affects all of the events that come after it. We can’t adjust for what might happen if the penalty early in the third doesn’t get called, so why should we make adjustments for rebound shots?
I find diving into topics like this to be much more of a philosophical practice as opposed to a call to action. Thinking more deeply about the game is a training exercise to become a better evaluator, coach, etc., because it helps us better understand how we see and feel the game, as well as how it’s affecting the numbers. If you are going to be a continuous learner, you need to reflect on what you’re learning.