League Insights AI

Overall season stats including KDA, CS/min, impact metrics, and role distribution.
Extreme match performances and month-to-month performance trends.
Playstyle identity visualized through radar metrics and AI-derived tags.
Timeline insights, power spike timings, comeback patterns, and objectives.
Full-season kill heatmap showing map pressure and positioning trends.
Personalized season summary powered by Amazon Bedrock, including strengths, weaknesses, playstyle analysis, and actionable improvement tips.

Inspiration

Most League stat sites are good at showing numbers, but they do not tell the full story of a player's season. They don’t explain long-term patterns, playstyle tendencies, map behavior, or how a player actually impacts games beyond isolated matches. I wanted to create a tool that analyzes an entire year of gameplay and produces insights that feel personal, meaningful, and immediately useful. That led to building Rift Rewind.

What It Does

Rift Rewind analyzes a full year of League of Legends match history and produces:

Detailed performance statistics across hundreds of matches
Timeline analytics including early game dominance, midgame swings, throws, and comebacks
Role and champion trends
Kill heatmaps across all matches played
Monthly performance patterns and season-long trends
A generated season recap using Amazon Bedrock that provides a personality profile, strengths, weaknesses, and a summary of your playstyle

The goal is to deliver an end-of-year experience that goes beyond traditional stat dashboards.

How It Was Built

Backend (Flask + PostgreSQL)

Asynchronous pipeline using aiohttp to fetch and process large volumes of match data efficiently
Riot API integration for match history, match details, and timeline data
Database schema designed to support analytics across hundreds of matches
Batch insertion, retry logic, region auto-detection, and rate-limit aware scheduling
Timeline processing for gold swings, spike detection, kill positions, roaming, power spikes, and objective presence
Amazon Bedrock used to generate personalized season recaps

Frontend (Next.js + TypeScript)

Built with Next.js App Router and TypeScript
Animated UI constructed with Tailwind CSS, shadcn/ui, Radix primitives, and Framer Motion
Recharts for data visualizations (radar, line, bar, and pie charts)
Canvas-based rendering for kill heatmaps
Fully responsive design for desktop and mobile

Cloud and Deployment

Frontend deployed on Vercel
Backend deployed on a cloud platform (Render/Railway)
PostgreSQL used for persistent storage
Amazon Bedrock for AI model inference

Challenges

Handling Riot API rate limits while fetching hundreds of matches
Designing a scalable timeline processing system that stays performant
Creating accurate heatmaps required transforming raw in-game coordinates into canvas coordinates
Ensuring SQL inserts remained consistent under concurrency
Prompt engineering for producing useful and consistent AI-generated recaps