League of Legends Esports Global Power Rankings
My team's first place submission for the AWS League of Legends Global Power Rankings Hackathon hosted by Devpost.
We collect, clean and aggregate game statistics from professional League of Legends games from 2020-2023. Then, we train various classical machine learning models to predict individual League of Legends game outcomes using these stats. Our best gradient boosted tree models predict outcomes with 65% accuracy over all games from Worlds and MSI qualifying regions from 2022-2023. On Worlds and MSI games alone they make predictions with 70% accuracy. These models are then used to create League of Legends esports power rankings for different tournaments and sets of teams by simulating our own round robin tournaments. Link to how our ranking method works.
Different power rankings are displayed on a web application.