Description

 I built a Binary Classification Neural Network using the PyTorch library, and trained it to predict the outcomes of League of Legends games, based on each participant's stats. I achieved approximately 92% validation accuracy on post-match data, on a dataset of over 10 000 matches.

The dataset was created by me using Riot's API and the RiotWatcher wrapper library. The script that collects the data starts with a match, then, for a random summoner in that match, pulls the past match history and then collects the data of that match, and continues the process recursively.