How analytical models can inform player development:

A case study using WHIMS V2

At the professional level, every hockey team has access to expected goal models (among a number of other models available for player and team evaluation). The current use cases for these models are great; they allow for teams to evaluate their performance at individual levels, and at team levels, and the access to very granular data allows for better player and team evaluation. It also means in today's day and age, another expected goal model isn’t very exciting, unless you don’t have one to begin with. I started the Women’s Hockey Insight Model for Scoring (WHIMS) project a little over two years ago, working in what I call a “data poor” sporting environment (Canadian Collegiate Women’s Hockey, formally known as USport). While representing a pretty high standard of hockey, USport doesn’t have any centralized play by play data, and there’s no third party data provider options. So that lead to the creation of WHIMS, with over 30 thousand unblocked shots collected across three seasons, we have an expected goals model specifically tailored for women’s USport hockey. 

Expected goals models, when used as a form of player and team evaluation, really are the bare minimum in terms of getting better information on performance. It’s a really nice starting point analytically, and with little extra sprinkles of information added on, you can start to build out sequencing and possession models, ultimately showcasing how analytical models can be used to improve our understanding of performance. With this in mind (and to my knowledge) I think one of the most underexplored use cases of expected goal models (amongst other analytical models) is in athlete development. There seems to be limited work in examining how measures of good performance at higher levels (i.e. professionally), should be used to inform the skills and tactics younger athletes are working towards. One example of this is: if a play that we believe should be a high danger scoring chance, doesn’t actually act as a high danger scoring chance, should we work on improving the skills behind it, or should we spend less time on it? To examine this question, I'll be using WHIMS V2, and data from OUA women's hockey. 

How analytics should force us to think differently about development

When I was reviewing the updates to my model, I found myself fascinated by the seemingly lack of importance of one timers. In professional hockey, one timers are tracked by third party data providers as a unique variable, largely because those third party data providers have decided they matter enough to be tracked – but this tracking is inspired and derived largely from men’s hockey. A lot of my research time has been spent trying to figure out the differences between men’s and women’s hockey, and what that means for coaching, team tactics, player development, and more, and so this one timer question grabbed my attention. I pulled every one timer from the 2024-2025, and 2025-2026 seasons in my data set to look at them. A lot of them centralize to the net front/low slot, and then off the sides of the crease in a “back door” area, but it becomes pretty quickly apparent that one timers from distance don’t go in very often. This is better seen by just looking at one timer goals.

There are some interesting qualities to these one timers that I didn't expect. First, one timer goals go in at a much higher rate on the rush relative to how often one timers are taken (on 1107 total one timers, 15% are on the rush despite rush one timers making up 26% of the goals). Second, the shooting percentage for all unblocked one timer attempts (saved, missed, goal), is 11.2%, which is extremely high. However, when we filter out all one timer attempts that are more than 20 feet away, that shooting percentage drops to 4%. Again, these are unblocked attempts, which means there’s likely a solid number of blocked (or misfired) one timers not being counted here. Shooting four percent beyond this arc feels bad! I think a lot of coaches and players think about a cross seam one timer to the dot area as a high quality scoring opportunity. For reference, here’s an arc of shots at about 20 feet from the net.