MicroMentor

AI-powered esports coaching platform that transforms raw player stats into data-backed insights, automated match reviews, and what-if scenario analysis for pro players.

Comment

Inspiration

Every week, esports coaches spend 15+ hours manually reviewing VODs, scrubbing through spreadsheets, and trying to correlate raw statistics with actual in-game decisions. We've watched Cloud9's coaching staff struggle to answer seemingly simple questions like:

  • "Why did we lose 78% of rounds when OXY died without a trade?"
  • "Should we have contested that Baron, or was the 4-1 split the right call?"
  • "Is our mid-laner actually improving, or are the numbers just noise?"

The brutal truth? Current esports analytics tools are data-rich but insight-poor. They tell you your CS@10 is 78, but not what that means for your role, how you compare to top performers, or what concrete drills will help you improve.

We built MicroMentor because pro players deserve AI-powered coaching intelligence, not just dashboards full of numbers.

What it does

MicroMentor is a comprehensive AI-powered assistant coach that transforms raw esports data into actionable coaching insights:

Core Features

  • 20+ Micro-Skill Metrics across laning, vision, combat, and objectives
  • Role-Specific Benchmarking with percentile rankings
  • Data-Backed Insights with confidence scores and correlations
  • Automated Macro Review Agendas with timestamps
  • "What-If" Scenario Engine with probability analysis

Premium Features (New!)

  • Dark Mode - Toggle between light and dark themes
  • AI Coach Chat - Ask questions about gameplay in natural language
  • Voice Input - "Hey MicroMentor, what if I contested Drake?"
  • Player Comparison - Side-by-side radar charts with head-to-head stats
  • PDF Export - Download professional coach reports
  • Animated Charts - Smooth transitions when switching players
  • Loading Skeletons - Beautiful placeholders during data fetching
  • Goal Tracker - Visual progress bars with milestone celebrations

How we built it

Architecture

GRID API -> ETL Pipeline -> SQLite -> Flask API -> React Dashboard
              |                          |
         ML Models              AI Insight Engine
    (RandomForest, XGBoost)   (Correlation Analysis)

Tech Stack

Component Technology Why We Chose It
Backend Flask 3.1 + SQLAlchemy Lightweight, fast API development
ML Engine Scikit-learn + XGBoost Proven models for performance prediction
Frontend React 18 + Plotly + Recharts Beautiful, interactive visualizations
Database SQLite Portable, zero-config for demos
Data Source GRID API Official esports data with GraphQL
Voice Input Web Speech API Native browser voice recognition
PDF Export jsPDF Client-side report generation
AI Assistant JetBrains Junie End-to-end development acceleration

Key Technical Decisions

  1. Role-Specific Benchmarks: Separate percentile curves for each position
  2. Confidence Scoring: Every insight includes sample size and correlation
  3. Dark Mode: CSS variables with localStorage persistence
  4. Voice Recognition: 20 lines of Web Speech API integration
  5. Animated Transitions: Plotly's built-in easing functions

Challenges we ran into

1. The Correlation vs. Causation Problem

Raw stats often show spurious correlations. We built multi-factor analysis that considers role, matchup, and team composition.

2. Making Voice Input Natural

Mapping spoken questions to API calls required custom keyword extraction and prompt engineering.

3. PDF Generation Without Server

Used jsPDF for client-side generation, carefully formatting multi-page reports with proper styling.

4. Dark Mode Across Components

Required comprehensive CSS variable system and careful attention to Plotly chart backgrounds.

5. Real-time Feel on Static Data

Implemented loading skeletons and smooth animations to give the app a polished, responsive feel.

Accomplishments that we're proud of

Feature-Complete Platform

  • 9 major features implemented in one hackathon
  • 10+ API endpoints for comprehensive data access
  • Dark/Light modes with smooth transitions
  • Voice input working across browsers

Production Quality

  • Proper logging and error handling
  • Clean component architecture
  • Responsive design for all screen sizes
  • PDF export with professional formatting

Beautiful UX

  • Animated radar chart transitions
  • Loading skeleton placeholders
  • Real-time indicator animations
  • Intuitive tab navigation

AI-Powered Intelligence

  • Natural language query processing
  • Contextual response generation
  • Probabilistic scenario analysis
  • Voice-enabled interactions

What we learned

Technical Lessons

  1. CSS variables are powerful: One theme system, infinite possibilities
  2. Web Speech API is underrated: Voice input in 20 lines of code
  3. jsPDF works great client-side: No server needed for reports
  4. Plotly animations are smooth: Built-in easing functions

Product Lessons

  1. Dark mode is expected: Users immediately look for the toggle
  2. Voice input delights: "Wow, I can just ask it!" reaction
  3. Export is essential: Coaches need offline materials
  4. Comparison drives engagement: Users want to see themselves vs. others

Process Lessons

  1. Junie AI accelerates everything: 3x faster development
  2. Start with the hardest feature: AI chat first, polish last
  3. Demo mode is crucial: Mock data lets you build without API keys

What's next for MicroMentor

Near-Term (Next 3 Months)

  • Video Clip Extraction: Automatically pull 15-second clips for each agenda item
  • Mobile App: Push notifications for performance alerts
  • Real-Time Tracking: Live monitoring during scrims

Medium-Term (6 Months)

  • Team Synergy Analysis: Cross-player correlation and communication timing
  • Multi-Game Support: VALORANT, CS2, Dota 2, Rocket League
  • Preference Learning: AI learns individual player goals over time

Long-Term Vision

  • Pro Team Integration: Direct integration with scrim tools
  • Streaming Integration: Automated highlight detection
  • Community Benchmarks: Anonymous aggregate rankings

MicroMentor transforms raw esports data into coaching intelligence - because pro players deserve AI-powered insights, not just spreadsheets.

Built With

Share this project:

Updates