MetaCoach

Player Hub
Draft analysis

Inspiration

Esports teams sit on massive amounts of data, but most of it is locked behind dashboards that only analysts can truly interpret. Coaches still rely heavily on intuition, fragmented stats, or manual prep before matches.

MetaCoach was inspired by a simple question: What if every team even without a full analyst staff had access to an AI coach that could scout opponents, understand playstyles, and translate raw data into clear strategic guidance?

We wanted to bridge the gap between data availability and decision-making, especially in high-pressure moments like drafts and match preparation.

What it does

MetaCoach is an AI-powered esports coaching assistant that transforms historical match and roster data into actionable insights. It helps teams:

Scout opponents automatically using historical match data
Analyze roster compositions and tendencies
Generate strategic profiles (early-game pressure, scaling potential, volatility)
Differentiate between pre-match scouting and draft-time analysis
Present insights in a coach-friendly dashboard, not raw spreadsheets
Rather than overwhelming users with stats, MetaCoach focuses on interpretation: what this team is likely to do, why it works, and how to respond.

How we built it

We built MetaCoach using a data + AI pipeline(Gemini API):

GRID Esports GraphQL API for authoritative esports data (teams, players, matches)
Supabase Edge Functions as a backend layer for:
Fetching and validating GRID data Normalizing rosters and match history
Caching stable data (teams, players) while keeping analysis dynamic
AI reasoning layer that: Interprets roster composition and historical patterns
Generates human-readable scouting summaries and recommendations
Frontend dashboard designed around coaching workflows rather than raw analytics

Special care was taken to separate:

Live / draft-time analysis (when available)
Pre-match auto scouting using historical data (hackathon-safe and reliable)

Challenges we ran into

GRID GraphQL schema constraints Not all intuitive fields (like matches, region, or shortName) exist where we expected, requiring careful schema exploration and query restructuring.
No full historical match dump GRID does not provide a single endpoint for “all matches,” so we had to design incremental, team-based historical fetching strategies.
Live draft limitations Hackathon APIs don’t fully support real-time draft data, forcing us to clearly separate draft analysis from pre-match scouting.
Player media inconsistencies Player images are not consistently available across official APIs, requiring fallback strategies and UI-safe defaults.
Caching vs freshness trade-offs Deciding what to cache (rosters, teams) versus what to fetch live (recent matches, analysis) was critical for performance and reliability.

Accomplishments that we're proud of

Successfully integrating GRID’s GraphQL API into a real product workflow
Building a clear scouting vs draft analysis mental model
Turning noisy esports data into coaching-language insights
Designing a system that works within hackathon API limits while still feeling powerful
Creating a foundation that can scale into live coaching, scouting reports, and market intelligence