We wanted to give every League player the kind of real-time feedback only pros get from analysts. Post-game stats are too late — we wanted live insights while you play.
How We Built It
Predict.gg connects to the Riot Live Client API to track live metrics like kills, gold, and objectives. We trained a logistic regression model
Every few minutes, a lightweight Tkinter overlay shows the current win rate and three concise strategy tips generated by an LLM (Claude 3.5 via OpenRouter) — turning data into actionable advice.
What We Learned
We learned how to merge ML predictions with LLM reasoning, balance live data pipelines, and build low-latency overlays that don’t interfere with gameplay. Challenges
Handling inconsistent API data, syncing UI updates without freezing, and making sure AI tips stayed accurate and short were our biggest hurdles — but solving them taught us a lot about building reliable, intelligent systems in real time.
Log in or sign up for Devpost to join the conversation.