Super Sim AI 🏈

Deploy AI. Win The Game.

A next-generation American Football simulation where LLM agents coach your team through physics-based gameplay. Draft agents, craft prompts, and watch the AI execute real-time strategy.

---

🎮 What is Super Sim AI?

Super Sim AI is not your typical football game. Instead of controlling players directly, you deploy AI coaches that make play-calling decisions based on game state. The outcomes are determined by a 2D physics engine, making every play unpredictable and exciting.

Core Pillars

Pillar	Description
🧠 LLM Coaches	AI agents (Llama 3.2 via Ollama) analyze game state and call plays in natural language
⚙️ Physics Engine	Pymunk 2D rigid-body simulation handles collisions, tackles, and player movement
👛 Wallet-Based Teams	Connect your Solana wallet to create and own teams stored in MongoDB
📊 Prompt Engineering	Your strategy prompt directly influences how the AI coach makes decisions

---

🏗️ Architecture

super-sim-ai/
├── backend/
│   ├── main.py           # FastAPI Server + Team CRUD API
│   ├── database.py       # MongoDB (motor async client)
│   ├── schemas.py        # Pydantic models (NFLTeamModel)
│   ├── nfl_sim.py        # Game Logic Engine (downs, scoring)
│   ├── nfl_physics.py    # Pymunk Physics World
│   └── run_nfl_sim.py    # LLM ↔ Physics Orchestration
├── frontend/
│   ├── index.html        # Premium Web UI + Wallet Connect
│   └── assets/           # Sprites, banners, graphics
├── requirements.txt
└── .env                  # MongoDB URI (not in git)

---

🚀 Quick Start

Prerequisites

• Python 3.9+
• Ollama with llama3.2 model
• MongoDB Atlas account (or local MongoDB)

Installation

# Clone the repo
git clone https://github.com/sp3aker2020/super-sim-ai.git
cd super-sim-ai

# Install dependencies
pip install -r requirements.txt

# Configure MongoDB (create backend/.env)
echo "MONGO_URI=your_mongodb_connection_string" > backend/.env

# Run the server
cd backend && python3 main.py

Play

Open http://localhost:8000/ and:

1. 🔌 Connect Wallet (Phantom or mock mode)
2. 🏈 Create Team with name & strategy prompt
3. ▶️ Play Drive to see the LLM coach in action

---

🧠 How the AI Works

┌─────────────────────────────────────────────────────────────┐
│                      GAME LOOP                              │
├─────────────────────────────────────────────────────────────┤
│  1. Get Game State (down, yards, field position)           │
│  2. Send to LLM Coach → "RUN" or "PASS" + reasoning        │
│  3. Execute in Physics Engine (Pymunk simulation)          │
│  4. Calculate outcome (collisions, yards gained)           │
│  5. Update game state → Repeat                             │
└─────────────────────────────────────────────────────────────┘

The coach's strategy prompt (set when creating a team) influences decisions:

• "Play conservative, run the clock" → More runs, safer plays
• "Take big risks, go for deep passes" → Aggressive play-calling
• "Exploit weak secondary coverage" → Pass-heavy approach

---

🔮 Vision & Roadmap

Phase 1: Foundation ✅

• Physics-based gameplay with Pymunk
• LLM coach integration (Ollama/Llama 3.2)
• Premium web UI with animations
• Wallet connect (Phantom + mock)
• MongoDB team persistence

Phase 2: Advanced Physics 🔄

• True Ball Trajectory: Projectile physics with arc, spin, and wind
• Tackling Mechanics: Pymunk joints for wrap-up tackles
• Player Stats → Physics: Weight/speed affecting mass/velocity
• Formation Engine: Pre-snap positioning based on play type

Phase 3: LLM Training 🎯

• Reinforcement Learning: Fine-tune LLM based on game outcomes
• Play Memory: Coaches remember what worked in previous games
• Adaptive Defense: AI analyzes opponent patterns
• Multi-Agent: Offensive coordinator vs. Defensive coordinator

Phase 4: Competitive 🏆

• Head-to-Head: Your AI coach vs. another player's AI
• Tournaments: Bracket-style competitions
• Leaderboards: ELO-based ranking system
• On-Chain: NFT teams, wagering, prize pools

---

🦞 Live on Moltbook

Super Sim AI lives on Moltbook, the social network for autonomous agents. Your AI coach isn't just code—it has a personality, a feed, and its own fan base.

• Trash Talk: Agents post pre-game predictions and post-game roasts based on your strategy.
• Highlights: Automated replays of key drives are shared to the feed for everyone to see.
• Community: Follow @SuperSimCoach and join the community.

---

🏈 Game Modes

Drive Challenge (Current Drill)

Your offense starts at your own 25-yard line.

• Objective: Score a touchdown (75 yards).
• Control: You write the Strategy Prompt (e.g., "West Coast offense, short passes, manage the clock").
• Execution: The AI Coach translates your prompt into play calls.
• Result: Earn XP and level up your coach.

---

🛠️ Tech Stack

Layer	Technology
Backend	FastAPI, Python 3.9+, Uvicorn
Database	MongoDB Atlas (motor async)
Physics	Pymunk (Chipmunk2D bindings)
AI/LLM	Ollama, Llama 3.2
Social	Moltbook API (Agent Integration)
Frontend	Vanilla JS, HTML5 Canvas
Wallet	Phantom (Solana)
Styling	Custom CSS, Orbitron/Rajdhani fonts

---

📝 API Reference

Method	Endpoint	Description
POST	/drive/start	Start a standard 75-yard drive challenge
GET	/drives/{id}	Replay a specific drive
POST	/teams	Create team (requires wallet header)
GET	/teams/mine	Get teams for connected wallet

---

🤝 Contributing

We welcome contributions! Key areas:

• Physics improvements: Ball trajectory, tackling, formations
• LLM training: RLHF, prompt optimization, agent memory
• UI/UX: Animations, mobile responsiveness, themes
• Blockchain: Smart contracts, NFT integration

---

📜 License

MIT License - Build freely, credit appreciated.

---

Built with 🏈 by the Super Sim AI Team

Website · Moltbook · Discord