Auto-Scout

• 

  Myplatform

💡 Inspiration

In competitive esports (like Dota 2, LoL, or CS2), the match is often won or lost before it even begins—in the draft. Professional teams spend hours analyzing opponent history and meta trends, but during the live draft, the pressure is immense and time is short.

I wanted to build Auto Scout to act as a real-time "AI Assistant Coach." I was inspired by the idea of combining hard data (win rates, gold graphs) with strategic reasoning (team composition synergy). The goal was to create a tool that doesn't just show stats, but actually understands the game state.

🤖 What it does

Auto Scout is an intelligent drafting assistant that:

1. Monitors the Draft: Listens to live pick/ban phases using the GRID API.
2. Analyzes Strategy: Feeds the current team composition and opponent data into the Google Gemini API.
3. Suggests Moves: Provides real-time recommendations for the next best pick or ban, complete with reasoning (e.g., "Pick Hero X to counter their physical damage output").
4. Predicts Threats: analyzing the opponent's historical data to flag high-threat players or comfort picks.

⚙️ How we built it

The core of the application is written in Python.

• Data Ingestion: We utilized the GRID API to fetch granular, real-time match data and tournament statistics.
• The Brain: The processed game state is sent to Google Gemini. We used advanced prompt engineering to teach Gemini the nuances of MOBA mechanics so it could think like a coach rather than just a calculator.
• Development: The entire project was built and debugged inside IntelliJ IDEA, utilizing its robust tools for Python development.

🚧 Challenges we ran into

• Real-time Data Parsing: Mapping the complex JSON structures from live esports matches into a format Gemini could digest quickly was a significant hurdle.
• Hallucinations vs. Strategy: Tweaking the temperature and system prompts for Gemini to ensure it gave strategic advice based on actual game mechanics rather than generic gaming terms.

🏆 Accomplishments that we're proud of

We successfully connected a live data pipe (GRID) to a generative model (Gemini) effectively. Seeing the AI suggest a ban that actually made sense for the specific team composition was a huge "It works!" moment.

🧠 What we learned

I learned that Generative AI isn't just for text generation—it's a powerful reasoning engine for structured data. Integrating the GRID API also gave me a deeper appreciation for the depth of data available in professional esports.

Built With

• git
• google-gemini-api
• grid-esports-api
• intellij-idea
• juni
• python