Cloud9 Coach

• 

Inspiration

Esports teams generate massive amounts of match data, but most players and coaches rely on manual review or surface-level statistics. Inspired by the Moneyball philosophy of data-driven decision-making, we wanted to build an AI-powered assistant coach that could analyze match data automatically and convert it into actionable performance insights.

We were particularly inspired by how professional teams like Cloud9 use analytics to improve gameplay. Our goal was to bridge the gap between raw match data and real coaching advice by combining statistical analysis with natural language explanations.

What it Does

The Cloud9 Assistant Coach AI analyzes esports match data and provides:

Personalized player insights — identifies recurring mistakes and statistical anomalies.

Team-level macro analysis — highlights key strategic decisions and critical moments.

Predictive “what-if” scenarios — estimates alternative outcomes if different decisions were made.

In simple terms, it turns numbers into coaching advice.

How We Built It

The system is built using a full-stack architecture:

A Python + FastAPI backend handles data processing and analytics.

A React frontend displays insights in an interactive dashboard.

Match data is ingested from demo datasets and optionally from esports APIs.

We use statistical analysis combined with AI-generated explanations to produce human-readable coaching feedback.

Core performance metrics are computed using formulas such as:

𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑆𝑐𝑜𝑟𝑒 = 𝛼⋅𝐾𝐷𝐴 + 𝛽⋅𝑂𝑏𝑗𝑒𝑐𝑡𝑖𝑣𝑒𝐶𝑜𝑛𝑡𝑟𝑜𝑙 + 𝛾⋅𝑉𝑖𝑠𝑖𝑜𝑛𝑆𝑐𝑜𝑟𝑒

This allows us to quantify player performance while still providing qualitative insights through AI-generated summaries.

Challenges We Faced

Data complexity: Esports match data is high-dimensional and noisy, making meaningful feature extraction difficult.

Insight generation: Translating raw statistics into understandable coaching advice required careful prompt design and logic tuning.

Time constraints: Building both backend analytics and frontend visualization within hackathon limits was challenging.

Model alignment: Ensuring the AI outputs were helpful, consistent, and non-generic required multiple iterations.

What We Learned

How to design an end-to-end AI system combining data analytics + LLMs.

How to structure esports data for meaningful performance evaluation.

How to balance automation with interpretability.

How to rapidly prototype and deploy a multi-service application under time pressure.

Future Improvements

Integrate live esports APIs for real-time match analysis.

Add role-specific coaching (e.g., support, entry fragger, jungler).

Improve prediction accuracy using machine learning models.

Add historical player trend analysis across multiple matches.

Built With

• fastapi
• github
• grid
• intellij-idea
• javascript
• junie
• numpy
• pandas
• python
• react
• restapi