There are certain things that as longtime fans of hockey, we innately know must be true. One of those things is that an inordinately high amount of goals are scored on rebounds. In my recent work to incorporate shot distances into a shot differential statistic, I wanted to explore whether I could tease out the effect of rebounds on expected goal scoring and shot making.
Again, the gist of the initial story was told by this chart:
There’s a distinct non-linear trend in this line as it gets closer to the net. Shots that are taken closer to the net tend to go in more. But why is this? There’s only so much we can isolate from the historical record, but one that I figured would have a big influence is rebound chances. Perhaps the effect of rebounds are inflating closer in shots.
To test this I expanded my dataset of shots to include 5 seasons of data between 2007-08 and 2011-12. This amounts to about 420,000 unblocked, non-empty net, even strength shots at the net. I then queried this dataset to pull out all shots taken within 0 seconds of the last shot attempt by the same team, 1 second, etc. Notice I said shot attempts: this would include a shot attempt taken after a miss or even a block. Those are in the small minority, but still afford an interesting advantage.
The following graph does a great job of illustrating the effect rebounds have by shot distance:
This shows the percentage of shots *on* net that resulted in goals by rebound timing and shot distance. Generally we see that rebounds taken in the same second have the highest chance of going in, those taken after 1 and 2 seconds result in very similar rates of goal scoring, while those after 3 seconds have a much lower but still higher than average shot of going in. Shots taken after 4 and 5 seconds clearly show an adherence to the overall average across all times and distances.
Here’s a chart that summarizes the key information:
SH% is shooting percentage, USH% is the shooting percentage of unblocked shots, HIT% is the percentage of shots taken that hit the net. We see that all three measures tend to sag as the time since the last shot attempt increases. Rebounds taken quicker tend to hit the net more and go in the net more, which will have a compounding effect on goal scoring. By 4 seconds, the advantage of a rebound very closely approximates the overall average.
So, we can reaffirm the traditional definition of a “rebound” being a shot taken within 3 seconds of the last shot attempt. That gives us two distinct categories of shots: rebounds and non-rebounds. Let’s now review the major shooting metrics by these two shot types. First, traditional shooting percentage:
Shots that hit the net are basically twice as likely to go in if they are rebounds compared to non-rebounds — this relationship holds surprisingly well across all distances. This suggests that the element of surprise and catching a goalie out of his stance is worth an incredible amount to a shooter.
Now, let’s have a look at the percentage of shot attempts that hit the net:
This one probably mystifies me the most. Why is a shot attempt so much more likely to hit the net when it’s a rebound compared to non-rebounds? In many areas of this graph, a rebound shot attempt is around 10% more likely to hit the net. Is this because the shooter recognizes that he has a good scoring chance and he bears down on his shot more? Is it because a rebound chance likely takes defenders off-guard as much as the goalie, and the shooter has less obstacles to avoid? Is it because when he takes the rebound shot he likely has a lot of net to look at, and is just more interested in getting the shot on net as opposed to trying to pick a corner? Maybe it’s all three. In any case, rebound chances are a lot more likely to hit the net.
Now let’s see the compound effect of these two forces — quicker shots hitting the net more often and beating goalies more often:
In some places of this graph, a rebound shot attempt is 2.5 times as likely to result in a goal than a non-rebound chance. And to me, that’s astonishing — imagine a player setting up to shoot from just 12 feet out. That’s a prime scoring chance, no matter how you frame it. But if he takes that shot within 3 seconds since his team’s last shot attempt, he’s 2.5 times more likely to score. More than 1 out of every 4 rebound shot attempts within 15 feet result in goals. Remember that Bachman sequence where he stopped 3 shots in a row on his doorstep in Los Angeles? The last two shots almost had a one in three chance of going in the second they left the player’s stick.
Last night I tweeted some figures that were a tad high about the percentage of shots and goals that are rebounds (working with 500,000 rows in Excel is fraught with danger!). Here are the actual figures, or so Excel leads me to believe: 7.2% of all shots on net are rebounds. 22.1% of all goals are rebounds. In any case, the disparity is massive, and enforces the value of creating rebounds for your team and denying them to the other team.
EDIT: added section below
Point of curiousity — where are most rebounds? They are overwhelmingly close to the net:
Rebounds are an in-close phenomenon. If you consider Rebounds as a class of shot attempts that are of higher quality, and that this class is much more likely to be taken close to the net (and be more effective close to the net), you can see why the age-old hockey adage of “going hard to the net” has such resonance. Teams that have players all on the periphery during an offensive attack are missing out on all these tasty points of goal-scoring probability.
The Case for Including Rebounds off Misses and Blocks
Earlier I mentioned that I’ve included “rebounds” off misses and blocks, as well as those coming from traditional rebounds off saves. I’d like to include a bit of rationale as to why. They are obviously different kinds of plays — a rebound off a save is more likely to be close in and with the goalie already prone, etc. But how many times have we seen (smart) teams try the intentional shot miss play off the end boards that actually works. And I’m guessing we’ve all seen more than a few goals scored off of previously blocked shots. Here are the stats:
We can see that shot attempts coming within 3 seconds of saves, misses, and blocks are all heightened threat plays. What do rebounds coming off of misses or blocks look more like — a rebound chance, or a regular shot? They have almost triple the probability of going in compared to a regular shot (USH%). I think it’s appropriate to include them in the classification of “Rebound” chances because of it.
To me this shows a fascinating aspect of game play — catching a goalie off guard is worth so much in hockey. Whether you’ve got him prone after making a previous save, have messed up his angles by capitalizing on a previous miss, or relied on a previous blocked shot to provide you with an unpredictable play situation, you’ve essentially just created an element of surprise. NHL goalies are very good, and beating them clean on first shots is a very difficult thing, even from in close (unless they play for the Oilers). Creating the chaos seen in rebounds is worth a lot — it seems to mess up the goalie’s initial calculations.
What percentage of rebounds come from shots off saves, misses, and blocks?
- Saves: 68.2%
- Misses: 12.5%
- Blocks: 19.3%
So rebounds off saves are still the dominant breed here, but you can see that plays off misses and blocks are not a minuscule trifle.