LoL Predict, match prediction via ML
Real-time win probability during a League of Legends match, inspired by esport broadcast overlays.
The project
LoL Predict is a continuation of BestPick, with a slightly different ambition. The idea is to get real-time win probabilities during an ongoing game, kind of like what you see in esport broadcasts. Knowing exactly when a condition flipped, when we just lost something important, that kind of insight. I ended up pausing the project over time, but the concept is still cool, and I may come back to it later with more perspective.
My contributions
Data collector via the Riot Games API (riot_api_client, storage_client), exploratory analysis in Jupyter, and implementation of prediction models.
Takeaways
Applying machine learning to video-game data is fascinating but complex. Real-time data quality and the required volume are real challenges, especially if you don’t have big storage at hand. It’s also a project that taught me to recognize when to pause something, because grinding isn’t always the right strategy.
Context
Personal project, following BestPick. Inspired by the live probability overlays you see in pro esport broadcasts, which are typically the kind of thing that makes you very jealous once you’ve seen them in a stream.
Tech stack
-
Python
-
UV (package manager)
-
Riot Games API
-
Jupyter Notebooks
Cover image generated by Nano Banana (Google), since I no longer have the mockup template I used before. So it’s not a screenshot of the actual tool.