Predicting a Stanley Cup Winner
Learnings from our Data Story Project
The Stanley Cup Final is now here, and there's tons of speculation online as to whether Florida or Edmonton will take the Cup. I'm personally not a huge Panthers fan, so with this re-match in sight am hoping the boys from Edmonton will hoist the Cup this year.
What are the odds? Well, in our month-long data story entitled Predicting Who Will Win the Cup, this is where we landed as of yesterday.
What is Data Punk Media?
As I mentioned earlier in the season, the Data Punk team is also building an experience in data storytelling. We're focused on sports, entertainment and business, and this data story allowed us to explore some simple ways to build predictive modeling for the Stanley Cup playoffs.
As we've written about before, the Pythagorean Win Probability (PWP) isn't necessarily the supreme model of the land, but it's pretty good as a directional model and as a teaching instrument.
At any rate, I wanted to keep it short today and highlight a couple of learnings from the time spent on this data story project.
- The first was that it reminded me that predictive modeling is part science and art. With hockey, it's important to try different combinations of features before you settle on a model – even if it's relatively simple.
- The second was that the project motivated me to dust off my Bayesian Probability textbooks to starting thinking about how we can build a more robust predictive model for hockey. Probability is at once complex and powerful, and I'd forgotten just how powerful a model can be when it can be updated at the close of a shift.
- And the third is that even after doing this for two decades, you shouldn't fight it when you end up a good distance from where you started. We had envisioned the start of the project to be more focused on one team and point-in-time, but found that this wasn't interesting – and the predictive information was there anyway so why not use it.
The project culminated in four posts on our Data Punk Media property, which I hope you will enjoy and find useful:
- The Data Story: "Predicting Who Will Win the Cup"
- Exploratory Data Analysis: "Predicting Who Will Win the Cup"
- Forecasting a Winner: "Predicting Who Will Win the Cup"
- The Final Content: "Predicting Who Will Win the Cup"
We've also created a short video in case you're feeling lazy or are short on time.
For those that are interested in the data storytelling theme, we'll try and cross-post with some regularity yet still keep to our core theme of Data, AI and Hockey here on Data Punk Hockey.
Join today to learn more about data storytelling for sports, entertainment and business!
We'll be starting our next series soon, which will focus on the incoming NHL prospects. But until then, enjoy the playoffs and go Oilers!
Subscribe to our newsletter to get the latest and greatest content on all Data, AI and Hockey.