© 2013 Michael Parkatti Shots5

Dissecting the Effectiveness of Shot Types

As I tumble further down the rabbit hole on shot quality analysis, I’ve come to a place where I need to establish whether shot types differ in efficacy by distance, and if so, does it matter enough to break them out in detail when assigning expected goal probabilities.  I’ve already looked at the concept of rebounds, and found that that shot type is incredibly more likely to end in a goal.  The pertinent chart was the following:


That’s a sexy blue line towering over the measly non-rebound line.  But what about that red line — do the shot types that constitute it differ in any meaningful way, or is it just an ugly lump of randomness around the mean?

First, here’s a look at what percent of non-rebound shot attempts are made up by each shot type by distance from the net — ie, how frequent is each shot type used as shot location gets further away from the net?


Now isn’t this an interesting picture.  Probably not all that surprising, but it’s fun to see charted out like this to test against your own assumptions about shot selection.  You’ve got flashes in the pan like wrap-arounds that give way to tip-ins.  Then it’s the long age of the wrist shot all the way until you reach point-shot distance, after which slap-shots evolve lungs and start to walk on land.

That area chart breaks it out by distance, but here’s each shot type (including rebounds) as a percent of all shot attempts:


Wrist shots are predominant as a whole, followed by slap-shots and snap-shots.  There’s some smaller minority shot types in there that seem insignificant, but what if we were to only concentrate on goals?


Here we see rebounds and tip-ins punching well above their weight compared to their shot attempt numbers.  This suggests that they are more dangerous than average.  On the other end of the scale, we see wrist shots and especially slap-shots give up many points in terms of goal percentage compared with their shot attempt percentage.

I’d say it’s worth breaking these shots up separately by unblocked shooting percentage (USH%) over each distance to see how their probabilities of goal scoring differ.


To be included in this chart for a particular distance, I needed to see at least 500 shot attempts taken at a specific distance in feet for each shot type.  Where a line disappears indicates that it did not reach this threshold for sample size.  At first I’d only looked at tip-ins, and thought that a brief high departure from the average non-rebound USH% wasn’t enough to consider it as “different”.  But when I started plotting out each shot type, I realized that there is a massive difference in the effectiveness of shot attempts at each distance by shot type.

Snap-shots, in particular, seem to show a broad-based increase in shooting percentage compared to the average — this is something also seen in slap-shots when there is enough sample size.  This would make sense simply as a function of assumed velocity, as slap shots and snap shots are likely the two hardest shots on average, giving the goalie less time to react to make a save.

Also recall that wrap-arounds are the majority of shots from certain close-in distances, and then see how low their shooting percentages are here.  I had wondered why there was a dip in the average non-rebound shooting percentage from 6-10 feet.  Players stuffing low-percentage wrap-arounds at bad angles into goalie pads would be why.  Those wrap-arounds are a quarter or a third as likely to go in as other shot types at those close-in distances, and it’s likely very important to account for that when assigning expected goal probabilities.  Ryan Smyth may be stuffing a lot of close shots at the net, but many of them are likely these low-percentage wrap-arounds.

And with that, I’ve largely exhausted all you can glean from NHL.com’s play by play data feeds to characterize a shot.  We know its distance, we know if it’s a rebound or not, and now we know what shot type it is.  We’re very close to being able to tie this all together into a much more accurate expected goal measure that combines shot differentials with shot quality.


  1. Sebastian
    Posted November 7, 2013 at 10:41 am | #

    Amazing and fantastic post and research, can’t wait to see you put it all together! One thing about your visuals though, you should really keep your colors consistent per shot type across all your chart types, it’s better formatting and helps with cross comparisons.

    Keep up the good work!

  2. Frag
    Posted November 7, 2013 at 7:43 pm | #

    Very good. Pretty much what I’ve been hoping for.

One Trackback

  1. […] question becomes how good are the shots that the Oilers are taking. Michael Parkatti of Boys on the Bus has been tracking shot quality. His work is well worth reading as shot quality is a critical factor […]

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>