Skip to the content

Predicting total NHL Team Wins with Machine Learning - 2022/23 Season

With a new NHL season underway, we wanted to determine if machine learning could predict the total wins for teams in the NHL 22/23 season using machine learning.

Building a model the in the Microsoft Azure Machine Learning environment and combining NHL historical data sites allowed us to leverage data exploration, preliminary data handling, and linear regression to create predictive models. These predictions can then be used for fantasy sports (season long and individual games), as well as prediction season long and individual game results.

There are a few options for Cloud based machine learning, we selected the Microsoft Azure AI Platform as it met all of our needs and provided a web-based low code environment to quickly configure machine learning operations and pipelines.

 

Data Collection & Data Points:

NHL Historical data was obtained from Rolling Insights Historical NHL data. Key data points used for this analysis were:

 

  • Player
  • Year
  • Team
  • Position
  • GP
  • TOI
  • Goals
  • Total Assists
  • Total Points
  • IPP
  • Shots
  • SH%
  • ixG
  • iCF
  • iFF
  • iSCF
  • iHDCF
  • Rush Attempts
  • Rebounds Created
  • PIM
  • Giveaways
  • Takeaways

 

If you’re looking for historical data for you own predictive models we provide NHL, NFL, MLB and NBA data for free. Try SportWise for personal use or DataFeeds for commercial API access.

 

Read Part 2 here