How this package it built?
A total of 27 different datasets using basic, advanced and micro stats from NaturalStatTrick, NHL.com's API, Dobber, MoneyPuck, Elite Prospects, Evolving Hockey and Corey Sznajder are calculated on in Python with over 3900 lines of code. With the package model written in Python instead of macros in Excel, it allows for quick iteration to make the package more accurate without redoing hours of calculations with each change.
Using this large swath of data, I feel I've created some of the most accurate ratings for contentious categories such as SK and particularly DF.
This package includes a complete rating set for any player that played a single game in the AHL or NHL this past season.
The ratings only reflect the most recent season of play and doesn't come in a multi-year weighted flavour.
This package doesn't include a PO rating.
For this year, it also has all forwards set to "C". This will be changed for next year's package.
Why this package?
The ADS ratings package is a 100% stats based approach. Yes, that includes the SK category. I've been able to find a combination of basic, advanced and micro stats that accurately act as a proxy to calculate SK for all NHL players. This means it is 100% objective and has taken human bias out of the equation.
Additionally, since the ratings model is written in Python. It allows for superior consistency year-over-year as the same model is used yearly with minor tweaks instead of the ratings being built from the ground-up at the end of every season.
I am a proponent of transparency and open source, so included in the package is slider settings for each category as well as a breakdown of the formulas used for each category and which stats were used. Additionally, for those interested, I can link my GitHub repo which holds the Jupyter Notebook of the ratings model as well as any scripts used to pull data from NHL.com and Elite Prospects.
Can I see a sample?
Here is a link to a Google Sheet that has the current rosters (as of July 16th, 2022) of the Canucks and Avalanche based on the ratings to give you a feel of what they look like. https://docs.google.com/spreadsheets/d/ ... sp=sharing
What is the cost?
The package costs $50CAD. If interested you can email me at [email protected].
Since I'm the new kid on the block, I asked a couple of commissioners their thoughts on the ratings package after using them for the most recently finished season. Here's what they had to say:
- Eric, Commissioner of the Virtual Hockey LeagueI have ran the Virtual Hockey League since 2006, and in my 15+ years as a commissioner, I have never been more impressed with an NHL ratings package before. Adam demonstrates insane commitment and accuracy in generating authentic ratings. His first package was developed for the 2021-2022 season which we used, and they went above and beyond expectation. If you want authentic ratings that are pulled from raw NHL datasets with no subjective or personal opinion factored in, Adam's ratings are bar none the most accurate set you'll ever find in the sim hockey world. There is even a detailed glossary index demonstrating how each individual category was determined, and if you have any questions, Adam is always happy to help answer your concerns thoroughly. You won't be disappointed!
- Tom,The PHL is a league that has been going for 15 years now and having used Adam's for the most recent season, I can say they are some of the most accurate we've used. Every rating has a clear and concise reason for being what it is...and Adam is very transparent about the formula. He is able to blend traditional stats with advanced metrics in a way that makes sense, giving each player a true representation of their real life self. If you are a real life re-rate league like we are, I would highly recommend Adam's rating package.
Commissioner of the Perennial Hockey League
Additional Bits
If you have any questions, comments or concerns, please let me know and I'd be more than happy to chat. Thanks!