Inspiration League of Legends draft phase is one of the most strategically complex parts of competitive play, yet most players rely on intuition or outdated tier lists. We wanted to create a tool that brings data-driven decision making to champion select - combining real pro esports statistics with team synergy analysis to give players the same analytical edge that professional coaching staffs have.
What it does Draft Assistant is an AI-powered companion for League of Legends champion select that provides real-time pick and ban recommendations. It analyzes:
Team synergy - How well champions work together based on archetypes and role balance Counter matchups - Win rate advantages against enemy picks Meta strength - Current competitive viability from pro play data The app displays ranked recommendations with score breakdowns and human-readable reasoning (e.g., "Excellent team synergy • Strong counter to enemy picks"). It also integrates directly with the League Client to provide live suggestions during actual ranked games.
How we built it Frontend: React 18 with Vite, styled with TailwindCSS using a custom League of Legends dark theme Backend: FastAPI (Python) with async architecture for non-blocking API calls Data Sources: Leaguepedia API for real pro match statistics (last 90 days of competitive data) Riot Data Dragon for champion information LCU (League Client) API for live game integration AI System: Weighted scoring algorithm combining synergy (35%), counter matchups (40%), and meta strength. Uses an archetype-based system mapping 150+ champions to 20+ playstyle categories for intelligent synergy detection Temporal Weighting: Exponential decay formula prioritizes recent patch data over older statistics Challenges we ran into Balancing multiple factors: Tuning the weights between synergy, counters, and meta strength required extensive iteration to produce sensible recommendations Data availability: Leaguepedia API has rate limits and inconsistent data formats, requiring robust caching and fallback systems LCU integration: The League Client uses self-signed SSL certificates and dynamic ports, requiring custom authentication handling Real-time performance: Keeping recommendations responsive while recalculating scores for 150+ champions after each draft action Accomplishments that we're proud of Explainable AI: Every recommendation comes with human-readable reasoning, not just a number Live client integration: Actually works during real ranked games, not just as a simulator Archetype system: Built a comprehensive champion classification system that enables intelligent synergy detection even for champion combinations never seen in pro play Full-stack polish: Clean League-themed UI with real-time updates and visual score breakdowns What we learned How to work with the Leaguepedia Cargo API and process large amounts of esports match data Implementing temporal weighting to keep recommendations relevant across patches Integrating with Riot's local League Client API (LCU) for real-time game state Balancing algorithmic complexity with response time in a real-time application The importance of explainability - users trust recommendations more when they understand the reasoning What's next for Draft Assistant Machine learning model: Train on historical pro drafts to learn deeper patterns beyond heuristic scoring Player-specific pools: Factor in individual player champion proficiency and comfort picks Team composition templates: Recognize and suggest proven team archetypes (poke comp, dive comp, etc.) Voice integration: Audio callouts during live draft for hands-free recommendations Expanded data: Incorporate solo queue statistics alongside pro play for broader applicability
Log in or sign up for Devpost to join the conversation.