I recently made some league-wide shot probability maps based upon the 2009-10 NHL regular seasons. Each map is of a particular shot type and strength of the shooting team. The strength of the shooting team is whether they were at even strength, on the power play or were shorthanded. The maps are limited to shots taken in the offensive zone. Most of the methodology is the same as what is found in the DIGR paper. I did some non-parametric smoothing using the LOESS function in R. This creates locally smooth maps of the ice. In these maps, I used a slightly smaller span in the LOESS which probably makes them slightly more affected by outliers. As with any spatial mapping, there are ‘edge effects’ primarily due to a some possibly outstanding observations. This can be seen best in the even strength backhand shots map (page 5). I would have a hard time believing that the probability of a backhand along the goal line increases as you move away from the crease.
It is clear that the probability at each location increases from Even Strength to the Power Play. That’s not a surprise. The maps do allow you to get a sense of where the high probability areas are for each type of shot.
Here they are: Shot Probability Maps.
I would certainly like to know more about your data in these maps. It is not clear to me what you are actually recording. Are these simply the probability that shot from a particular area is likely to be one type of shot over another? Are these any types of shots (in that a shot was attempted, and it may not necessarily have been a SOG because it was blocked or missed the net)? Is success (a goal) in the data anywhere?
As a coach, I am very intrigued by this study, and frankly would love to know more about what types of shots had what kind of success from what areas.
The shot probability maps are the probability that a particular type of shot is a goal from a given location.
Just came across this post as I have been thinking about shot probability for some time now. Where did you get your data? I have been able to collect the JSON data from NHL.com, but prior to this season, the shot type was only associated with goals. Thanks in advance and keep up the great stuff!
Where did you get the data for this shot map? CBS or NHL or ESPN?
Or was it a combination of several sources? In my experience the CBS shot locations are very wrong.
Never mind I see from the paper you got it from ESPN.com‟s GameCast.
I’m not convinced any of this is ‘great data’ relative to ‘ground truth’ but that matters less for THoR than for other analyses since we need things to be relatively good rather than absolutely good.