Machine Leagueing

If you're in it to win it, use machine learning to know whether or not to play it

Comment

Inspiration

After attending the Microsoft Azure workshop at QHacks, I felt inspired enough to explore its potential. I remembered the creators of a popular eSport called League of Legends created an API for their user data, so I decided to hack around with it for a while and ended up with this, a tool that helps people see which way the winds blow for active players with a similar playstyle.

What it does

The tool uses the Riot Developer API to build up a database of matches by first getting the matches of a specified seed user, and then recursively building matches using the respective player ids from those matches. Once a sizable enough dataset is collected and formatted, I send it over to Azure to train a simple model using a Two Class Boosted Decision Binary Tree algorithm.

How I built it

I used a python backend for the webpages and all the processing, Microsoft Azure for the machine learning/prediction service, and little old HTML, CSS and Javascript for the website.

Challenges I ran into

Both of the technologies leveraged had relevant limitations. The Riot API was a little more convoluted than I thought and it took a while to figure out the best way to get the data I wanted.

Deciding on what data to use was also a problem. Throughout the hackathon I was hitting the limits of what I could figure out how to do on Azure in a timely fashion, so this idea went through many iterations.

Accomplishments that I'm proud of

Everything I did over this weekend I did for the first time!

What I learned

I learned:

  • different types of machine learning algorithms, what each are good for
  • how to allow scripts to execute on a backend based on input through the frontend

What's next for machine leagueing

Machine Leagueing could use a lot of refactoring in order to increase speed and bandwidth efficiency. Furthermore, in its current state it only tells you how likely you are to win in general, but this could be expanded into focusing on the success of your playstyle, focuses and champion choice in the current season.

Share this project:

Updates