Rift Analyzer

Match analysis Page
Year Recap Page
Performance Analytics Page

Rift Analyzer is an AI-powered League of Legends coaching platform that transforms a full year of match data into personalized, actionable retrospectives. By combining the Riot Games Developer API with Amazon Bedrock's Claude 3 Sonnet, the application delivers insights that go far beyond traditional stat trackers—offering narrative storytelling, conversational coaching, and shareable social moments that help players reflect, learn, and improve.

Key Innovation: A multi-step reasoning AI agent that doesn't just report statistics but interprets trends, diagnoses persistent habits, and prescribes tailored practice plans through natural conversation.

1. Problem Statement & Opportunity

League of Legends players generate massive amounts of match data throughout a season, but existing tools (e.g., op.gg) focus on raw statistics without context or actionable guidance. Players struggle to answer questions like:

"Why am I stuck at this rank?"
"What patterns are holding me back?"
"How do I improve my late-game decision-making?"

Rift Analyzer addresses this gap by using generative AI to interpret data, surface hidden patterns, and deliver personalized coaching that feels like a conversation with an expert analyst.

2. How It Works

The platform follows a five-stage pipeline:

Match Ingestion: Fetches a full year of match data from the Riot API and caches it locally to ensure fast subsequent loads.
Feature Engineering: Computes advanced metrics like objective conversion rates, damage share trends, vision control patterns, and gold swing analysis across 20+ dimensions.
Pattern Detection: Applies statistical analysis to identify persistent strengths, weaknesses, and standout performances using z-score outlier detection and rolling window aggregations.
AI Synthesis: Feeds structured insights to Claude 3 Sonnet through Amazon Bedrock, using context-aware prompts and validation guardrails to generate grounded coaching responses.
User Experience: Delivers instant headline stats from cache while AI narratives stream in progressively, keeping the interface responsive throughout.

3. Core Features & Innovation

3.1 AI Coaching Agent (Primary Innovation)

What makes it special:

Multi-Step Reasoning: Uses Claude 3 Sonnet's tool-calling capabilities to fetch analytics mid-conversation, compare performance to target ranks, and assemble practice roadmaps.
Conversational Context: Maintains dialogue history so players can ask follow-ups like "What changed after I swapped to jungle?"
Grounded Recommendations: Every suggestion is validated against actual match data using context hashes and validation checkpoints.

Example Interaction:

Player: "How can I improve to reach the next rank?"

Agent executes multiple tools in sequence:
1. analyze_recent_performance() → Identifies win rate trends and game length patterns
2. detect_patterns() → Finds vision control drops in late game
3. compare_to_rank() → Shows gaps vs target rank benchmarks
4. generate_practice_plan() → Creates 3-week structured improvement plan

Final response synthesizes findings with specific metrics and actionable next steps.

3.2 Year-End Recap & Analytics

AI-generated narrative summarizing season journey with milestone achievements
Interactive charts for win-rate trends, KDA, and objective control
Champion mastery breakdown and role proficiency analysis
Shareable summary cards for social platforms

3.3 Performance Optimization

LocalStorage caching reduces page loads from 16s to <0.5s (97% improvement)
Lazy loading and progressive rendering keep UI responsive
Per-player, per-endpoint caching with manual refresh controls

4. Going Beyond op.gg

Feature	op.gg	Rift Analyzer
Raw Statistics	✓	✓
Win Rate Tracking	✓	✓
AI Narrative Storytelling	✗	✓
Conversational Coaching	✗	✓
Pattern Diagnosis	✗	✓
Practice Plan Generation	✗	✓
Context-Aware Insights	✗	✓
Shareable Social Moments	Limited	✓
Highlight Match Selection	✗	✓

Key Differentiator: Rift Analyzer doesn't just show what happened—it explains why, predicts what's next, and prescribes how to improve.

5. Technical Approach

Data Processing: Fetch match data via Riot API, compute advanced metrics (objective conversion, damage share, vision trends), apply z-score analysis for pattern detection
AI Orchestration: Claude 3 Sonnet with structured JSON context, tool-calling functions, temperature tuning (0.3 for stats, 0.7 for narratives), validation pipeline to prevent hallucination

7. Key Learnings

Caching Breakthrough: LocalStorage + lazy loading reduced page loads from 16s to <0.5s (97% improvement).

AI Grounding: Context hashes and validation checkpoints eliminated hallucination issues in early prototypes.

Insight Discovery: Late-game vision score emerged as strongest rank predictor (r=0.67), now anchors practice recommendations.

8. Impact & Production

Value: AI-generated narratives for reflection, pattern analysis for learning, personalized coaching for improvement, shareable moments for social engagement

Performance: <200ms cached responses, <2s AI generation, 85% cache hit rate, deployable on AWS Lambda + API Gateway

9. Hackathon Alignment

✓ All requirements met: AI agent, full-year data, insights, visualizations, sharing, AWS integration

✓ Key Innovations: First conversational coaching agent, context-aware AI, 97% performance improvement, validated responses, playstyle-driven recommendations

10. Conclusion

Rift Analyzer transforms match data into personalized coaching through AWS Bedrock's Claude 3 Sonnet—interpreting trends, diagnosing habits, and prescribing improvement plans. Beyond a dashboard, it's a conversation partner for player development.

Future: Multi-player recaps, real-time coaching, support for other Riot titles.

AWS AI Services & Data Sources

AWS AI Services:

Amazon Bedrock (Claude 3 Sonnet): Primary AI service for generating personalized coaching responses and year-end narratives. Uses tool-calling capabilities to execute multiple analytics functions in sequence with temperature tuning (0.3 for stats, 0.7 for storytelling)
AWS IAM: Provides secure, scoped credentials for Bedrock API access with least-privilege patterns
Boto3 SDK: Handles authentication, request batching, and retry logic with exponential backoff for Bedrock calls

Data Sources:

Riot Games Developer API: Focused on player's complete season using two endpoints:
- GET /lol/match/v5/matches/{matchId} - Match summaries (KDA, damage, gold, vision, objectives)
- GET /lol/match/v5/matches/{matchId}/timeline - Minute-by-minute timeline data (positions, events, vision control)
- GET /tft/summoner/v1/summoners/by-puuid/{encryptedPUUID} - Details of the summoner by their ID
- GET /riot/account/v1/accounts/by-puuid/{puuid} - Account Details by ID

Data Processing:

Fetch full-year match list and individual match details
Process timeline data for early/mid/late game performance patterns
Compute advanced metrics: objective conversion rates, vision trends, damage share deltas
Apply z-score and rolling window analysis for strength/weakness detection
Structure findings as JSON context for AI agent consumption