LoL Esports Analytics & Prediction
Blockchain Development

LoL Esports Analytics & Prediction

Real-time LoL Esports Data Pipeline, ML Prediction & Automated Betting System

GitHub
TypeScript Python React ML Polymarket

Problem

Polymarket lists live markets on LoL esports matches, but pricing often lags the in-game state because most traders cannot watch every league simultaneously. A pipeline that ingests real-time match telemetry, produces calibrated win probabilities, and acts on the gap between model probability and market price captures that inefficiency systematically.

Approach

  • PandaScore + Riot API as the two complementary data sources: PandaScore for cross-league coverage (LCK, LEC, LPL, LCS), Riot for high-fidelity in-game state like gold difference.
  • ML win-probability model trained on historical in-game features, exported to ONNX for low-latency inference from the TypeScript runtime.
  • Tier-based league organization (S/A/B) so the model and betting thresholds can be tuned per-league data quality.
  • Gold-lead threshold strategy on Polymarket: enter when the live Riot feed shows a gold delta the market hasn’t priced in; loss-cut and reversal logic on state swings.
  • Separation of concerns: data/ML pipeline, analytics dashboard, and betting engine as distinct services sharing the same feature store.

Implementation

Data Collection & ML Pipeline

PandaScore-based match data collection across LCK, LEC, LPL, LCS. Preprocessing with tier-based (S/A/B) league organization. ML training on in-game features to produce live win-probability estimates.

Real-time Analytics Dashboard

Live match tracking with auto-refresh and displayed ML win probability. Player of the Game (POG) forecasting from performance metrics. Objective tracking for Dragon soul, Baron, and Void Grub. Built on React 19, TanStack Router/Query, Tailwind CSS, and shadcn/ui.

Automated Betting System

Real-time gold-difference monitoring via Riot API. Automated Polymarket order placement when the gold lead exceeds a configured threshold. Loss-cutting and reversal betting on game state changes. Coverage for LCK, LEC, and LPL.

Esports Dashboard

Real-time match scores and upcoming schedules. Tournament information organized by league > series > tournament. Team search and details via PandaScore.

Outcome

  • End-to-end pipeline spanning data ingestion, ML inference, dashboards, and on-market execution.
  • Automated Polymarket trading tied to live in-game state rather than pre-match odds.
  • Unified analytics surface across the major professional LoL leagues.

Technologies

  • Data Pipeline: TypeScript (Bun), Python, PandaScore API, Riot API
  • ML/AI: ONNX Runtime, win probability models
  • Frontend: React 19, Vite, TanStack, shadcn/ui, Tailwind CSS
  • Trading: Polymarket API, automated order execution
← Back to Projects