In the last few days, I’ve been thinking about the next steps for Hockey-Statistics.com. I’ve decided to write an article about my thoughts. This puts some pressure on me to get things done, and hopefully I will get most of the following ready for the upcoming season.
Model Building:
As soon as the post-season finishes I will start working on the models. The models are the basis of all future work and visualizations, so this is by far the most important task.
xG model: The first step will be to build an xG model. I don’t expect my xG model to change drastically, and I expect the model to be similar to most other public xG models out there.
Player models: I’m going to build two player models – A descriptive “Game Score” model and a predictive “Player projection” model. Think of it as a GAR model and a projected GAR model.
Team models: Again, there will be two different models – A descriptive model and a predictive model used to calculate Game projections. This season the team models were simply the sum of the player models, but I might add some team metrics as well (Evolving-Hockey does this) – I haven’t decided yet and it will depend on the testing.
Projection Models: The projected team strength will obviously be the main input into the Game projections and Season projections. I will be testing a different way to translate team strength into projection – I’m quite excited about this.
As a new feature I plan to project goal score instead of just win probability, and depending on the models I might try out player prob projections as well.
I will do write-ups on the models as I finish them!
Visualizations/Cards:
For the most part I plan to keep the current Power BI visualizations and maybe further develop them.
When it comes to “Player Cards”, I want to make historic descriptive cards, so you can compare the performance of players from different seasons. I will also make player projection cards for the 2023/2024 season. Here you can see a player’s projected value right this instance and how the player has performed compared to the projection.
Similarly, I will make historic descriptive team cards as well as team projection cards.
eBook: Guide – NHL 2023/2024:
I plan to make a comprehensive guide to the upcoming season. This will include projected standings and analysis of every team including player projection cards of every player. I think this will be pretty cool.
The plan is to sell this guide as an eBook… Or give access to the guide via a subscription on the website (more information later).
The website:
First and foremost, I need to clean up the website. It’s quite messy at the moment because I just add things when I make them… I hate cleaning up.
When we get closer to the new season, I will probably make a video on how to use the website. I think it has a lot of cool features that no-one really knows about.
Lastly, I plan to make a subscriber section. A subscription will cost 10 dollars per year. I will still put all the visualizations and data up for free, but a subscription will give access to the eBooks. You should view it as a way to support. I might put more things in the subscriber section, but generally I want my work to be available for free.
Social Media:
I will continue to post Game Projections and more on Twitter, but I might eventually move on to other platforms as well.
I would like to increase my following, so I will probably focus more on player and team specific content. This drives traffic much more than model content and video tutorials.
Automation:
I’ve already automated a lot of processes, but I want to get even sharper in this area. An update that takes 5-10 minutes may not sound like a lot, but if you have to do it every day it quickly adds up. So, for next season I want to automate all model tracking. I haven’t been motivated to do the daily model tracking this season, so I haven’t posted model evaluations nearly as regularly as I wanted.
I plan to automate other things as well, but that’s more behind the scenes, so you probably won’t notice. The more things I can automate, the more time I can spend on the fun parts.
Design:
I want to streamline the design, so that you can instantly spot a @HockeySkytte visualization. In Power BI I’m able to make very beautiful designs. I plan to go with a light grey background and red/blue colors. I think this design is distinctive from the other cards in the public sphere, while it’s also soothing to look at.
Here’s an example of what I’m going for:
That’s some of the things you can look forward to in the coming months.
Hockey-Statistics 