2020-2021 VHL Ratings Package
Jul 28, 2021 14:58:42 GMT -5
Philadelphia Flyers, Vancouver Canucks, and 11 more like this
Post by Utah Hockey Club on Jul 28, 2021 14:58:42 GMT -5
I'm happy to release the 2020-2021 ratings package. This has been quite the project to work on, but I think all the grunt work now will set-up the league for years of standardized ratings to come. I'm happy to say that every rating category has been created using real-world data so is strictly objective and not subjective.
I started this project back in April with the goal to build a model to be used over and over again. I used 2019-2020 data to build last year's ratings from scratch to test and compare against the ratings we used this year. While there were some things that were great about this past year's ratings, I found some other ratings not nearly as accurate as I'd like. Proof that the model works is that it took me a grand total of 30min to download the necessary datasets for goalies and execute the model to spit out the 2020-2021 goalie ratings.
The model consists of ~1400 lines of code for players and ~450 lines for goalies. For those interested, the model is built with the Python programming language using the Pandas, NumPy and SciKit-Learn libraries.
Please let me know if you have any questions. I am more than happy to explain how certain things were calculated. If you have an issue with why ratings are a certain way, let me know and how you'd like them calculated. I can run the model with the updated calculations to see how the ratings would turn out differently. Without further adieu:
docs.google.com/spreadsheets/d/1m28C3PuS3NL5gqp9qJx76jLxOUK9KOuu77U0EFQhBt4/edit?usp=sharing
I started this project back in April with the goal to build a model to be used over and over again. I used 2019-2020 data to build last year's ratings from scratch to test and compare against the ratings we used this year. While there were some things that were great about this past year's ratings, I found some other ratings not nearly as accurate as I'd like. Proof that the model works is that it took me a grand total of 30min to download the necessary datasets for goalies and execute the model to spit out the 2020-2021 goalie ratings.
The model consists of ~1400 lines of code for players and ~450 lines for goalies. For those interested, the model is built with the Python programming language using the Pandas, NumPy and SciKit-Learn libraries.
Please let me know if you have any questions. I am more than happy to explain how certain things were calculated. If you have an issue with why ratings are a certain way, let me know and how you'd like them calculated. I can run the model with the updated calculations to see how the ratings would turn out differently. Without further adieu:
docs.google.com/spreadsheets/d/1m28C3PuS3NL5gqp9qJx76jLxOUK9KOuu77U0EFQhBt4/edit?usp=sharing
