Inspiration
As an avid league player, I wanted to create some useful statistics based on machine learning algorithms which can help players improve.
What it does
My project provides the user with information on their winrate and performance (grade score from S to F).
How we built it
I used XGBoosting to determine which features were most important based of games collected from challenger elo.
Challenges we ran into
Although I did manage to calculate performance metrics, the machine learning algorithm kept returning wildly different performance metrics from what I expected. I think a reason for this was that we did not clean the data before training our model, due to natural time restrictions. The ideal approach would have been iterating after calculating validation scores for a variety of models,s not just XGBoost. My fix for this was just adding more data, which seemed to improve the criteria, but it is still far from perfect. If I were to continue the project to a full-fledged web-app, I think this would be one of the main challenges. How to aggregate different ML models to create an accurate model.
Accomplishments that we're proud of
I'm proud that I correctly extracted challenger data, separated them per role and trained a model on each role, since every role will have different features that are important to lead the to the path of victory.
Reflections
This was a great opportunity for me to learn about the Riot API and AWS architecture. I didn't know that so much data on matches were publicly available! Also, I really learned a lot on using lambda and bedrock. Although I didn't have enough time to fully host the project on the web without running into importing issues, I still learned a ton on how using AWS can streamline shipping production code.
Log in or sign up for Devpost to join the conversation.