© 2013 Michael Parkatti ShotDiff2

Season-to-Date Shot Distances per Oiler

One thing I’m playing around with recently is shot distances, one of the items captured in a game’s play by play stats but pretty rarely used or analyzed.  It’s believe it’s been shown that shots taken closer to the net have a higher chance of going in (can’t find this on the interwebs), so to me it would make sense if players who seem to be able to create ‘closer’ shots to the net would show an enhanced ability to maintain a higher on-ice shooting percentage.  Keep in mind that earlier this year I found that there was nonrandomness shown for on-ice shooting percentages at the individual level — meaning that a player can maintain either high or low levels in that metric.  I’d like to explore whether maintaining high or low shot distances for (or against) are also a skill that can be maintained.  Perhaps consistently getting shots off close to the net helps to explain why you have a high on ice shooting percentage?

To start this research I’m going to have to get some data together, which will take a bit of time and effort due to the fact that it’s not readily available on any of the excellent stat sites we have access to.  I’ll have to query the data myself.  But to start, I’ve got all the data for the Oilers this season, so I thought I’d start querying that small amount of data to see what I can put together.  These numbers include all even strength situations. Here’s the money shot:


The categories labeled “DistSF” and “DistSA” are the average distances of fenwick events (unblocked shot attempts) from the net for the Oilers and against the Oilers in feet while that player is on the ice.  I’ve separated the table into 2 groups vertically, one with more (albeit still small) sample size than the other.  The DistDiff category just takes the difference between the two, implying that a higher number is better (having further away shots against is better than further away shots for).

We can see that the regular Oilers with the lowest shots for distances are Boyd Gordon, Ryan Smyth, and the Ladi Smid/Jeff Petry defensive pairing.  Smyth showing up as a leader here probably jives with a lot of observers’ impressions of how he creates his effectiveness offensively, namely, getting close to the net and taking much higher than average percentage shots.  Guys with further out distances include Nick Schultz, Ryan Nugent Hopkins, Taylor Hall, and Andrew Ference.  Schultz and Ference have been playing a lot of third pairing minutes recently, so aren’t really expected to generate offense the same way Justin Schultz and Belov are (who have shorter shot distances here).  RNH and Hall being poor at this is interesting.

In terms of Oilers with the lowest shot distances against (meaning bad), are Ryan Smyth, Ales Hemsky, and again Petry/Smid.  I’d guess that the top Oiler defensive pairing is good at getting close shots for but also lets close shots against jives with my earlier work that suggests defencemen can’t really control their own on-ice percentages much from season to season — these guys are probably playing with forwards who are decent at getting close shots but also playing against players who have the same ability.

The players with the best differentials are Boyd Gordon, Justin Schultz, Ladi Smid, and Ryan Smyth.

Let’s move on to the next section.  Here I wanted to capture the rate at which players create and allow shot attempts taken within 25 feet of the net both for and against.  Maybe a player’s average shot distance is scuttled by a dman he plays with who consistently lobs long shots from the point — these numbers try to neutralize that by finding the players who create close chances (and give them up).

Way out in front in terms of <25 foot shot attempts created per 20 minutes is Ryan Smyth, with 6.9.  Again, this seems to dovetail nicely with the narrative on this player, and why the hell he’s been able to score so many points for not being that puck-skilled of a player — it’s easier to create offence from close-in.  Other players good at this are Mark Arcobello and Jordan Eberle, who’ve played a lot together this year.  I’m guessing their numbers were also influenced by playing with Smyth, but those numbers are clearly above the pack.  The lowest rate of close shot attempts belong to Nick Schultz, Boyd Gordon, and David Perron.

In terms of the highest amount of close shot attempts allowed against the Oilers (so, bad), Ryan Smyth is again way out in front, followed by David Perron, and Justin Schultz.  If you had to guess 3 guys who would have been bad at this metric, I’m guessing you’d have gotten at least 2 of those guys.  Justin Schultz has a bad rap for playing shoddy defence, and this would suggest he is allowing more close-in shot attempts than his peers (though, it’s mitigated somewhat by him also creating the most close shot attempts for of all defencemen).  The players who allow the least close shot attempts per 20 minutes?  Ryan Nugent-Hopkins, Nick Schultz, Jeff Petry, and Boyd Gordon.  And wouldn’t that jive with some of your perceptions of who the best defensive players are?

I’ve created a percentage metric for this “shot attempts within 25 feet” thing to compare to traditional Fenwick (which includes all distances, obviously).  For most players they are pretty close to one another, but there are notable increases for players like Ryan Smyth, Boyd Gordon, Justin Schultz, and Ryan Nugent-Hopkins when looking at the <25-feet percentage.  The only Oiler who actually declines when looking at close shot attempts is David Perron, meaning the team as a whole has been better at the close-shot game than the overall-shot game.

But do these numbers tell us anything about how distances affect on-ice percentages?  If I compare these players’ average shot distances for with their on-ice shooting percentages so far, I find a r-square correlation of 0.204, meaning that over 20% of the variability in on-ice shooting percentage can be explained by the variability in average shot distances.  The P-value is just beyond 0.1 (0.105), which is at least suggestive of a significant relationship between the two.

There’s an even stronger relationship between average shot distances allowed and on-ice save percentages.  The two variables have a correlation of 0.364, with a p-value of 0.022, meaning I can accept a hypothesis that average shot distance allowed has explanatory power in understanding on-ice save percentages.

Next up I’ll try to get my hands on a lot more data to test this at a league-wide level.  I’d also like to test whether any of these shot distance variables are skills which can be sustained year-over-year.  We’ll see where this goes…



  1. Megan
    Posted October 22, 2013 at 1:40 pm | #

    Thanks for posting this! I’d been thinking a lot recently about the various things statisticians miss in things like shooting percentage that affect the quality of scoring chances, and this is definitely relevant.

    I’ve also been contemplating the idea of quality of saves for goalies, the idea being that goalies who work behind a good defense have inflated numbers. I wonder if, similar to some speculative scoring prediction functions, it would be possible to (somewhat) predict how well goalies might perform for other teams based on things like the percentage of that team’s shots against that come from odd-man rushes, on the penalty kill, etc. But that’s just my wistful notion of where these stats might one day lead.

    I only recently found this site, but thanks for doing all this thoughtful work. It’s nice to see someone who combines economics and hockey in a meaningful way.

  2. Posted October 22, 2013 at 3:05 pm | #

    Great stuff here!

    Like Megan, I’m also interested in seeing this stuff employed at a team level. For example, maybe shot distance can be combined with corsi/fenwick data to come up with a metric that has a better correlation with team wins than corsi/fenwick does on its own.

Post a Comment

Your email is never published nor shared. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>