GoQuant Real-Time Trade Simulator with Live OKX Order Book Data

⚡ Overview

The GoQuant Real-Time Trade Simulator is a Python-based platform that simulates live trading scenarios using OKX's Level 2 (L2) order book data. Designed for quant traders and developers, it predicts execution costs (slippage, fees, market impact), analyzes maker/taker behavior, and measures internal latency—all integrated into an interactive Streamlit UI for actionable insights.

📈 Core Features

✅ Live Order Book Processing

• WebSocket Integration: Connects to OKX's L2 order book via:
  wss://ws.gomarket-cpp.goquant.io/ws/l2-orderbook/okx/BTC-USDT-SWAP
• Real-Time Updates: Processes tick-by-tick data to reflect current market conditions.

🔍 Execution Cost Predictions

1. Slippage Estimation

   • Uses Quantile Regression (50th percentile) with inputs:

     • Spread
     • USD Amount
     • Order Book Depth
   • Output: Median slippage in basis points (bps).

2. Fee Calculation

   • Rule-based model for maker/taker fees (e.g., Tier 1: 0.02%/0.06%).

3. Market Impact Modeling

   • Almgren-Chriss Optimal Execution with dynamic programming:

     • Temporary Impact: ( \eta \cdot v^\alpha )
     • Permanent Impact: ( \gamma \cdot v^\beta )
     • Execution Risk: Volatility + inventory/time horizon.
4. Net Execution Cost
   [
   \text{Net Cost} = \text{Slippage} + \text{Fees} + \text{Market Impact}
   ]

5. Maker/Taker Classification

   • Logistic Regression trained on real-time data (100% accuracy).

6. Latency Profiling

   • Tracks processing time per tick via time.perf_counter().

🧠 Machine Learning Models

📌 Quantile Regression (Slippage)

• Library: statsmodels

• Performance:

  • MSE: 5.39e-05
  • R²: 0.798

📌 Logistic Regression (Maker/Taker)

• Library: scikit-learn
• Features: One-hot encoded order_type, side, and market depth.

🖥️ Streamlit UI

• Left Panel: User inputs (asset, order type, quantity).
• Right Panel: Real-time outputs (costs, probabilities, latency).

👉 Explore the live demo

⚙️ Optimizations

• NumPy-driven Almgren-Chriss solver.
• Low-latency JSON processing.
• Lazy evaluation to reduce idle computations.

📦 Setup

pip install -r requirements.txt  
streamlit run ui_app.py  

❓ FAQ

Q: How accurate are the slippage predictions?
A: The quantile regression model achieves an R² of 0.798, reflecting realistic median execution costs.

Q: Can I test different fee tiers?
A: Yes! The UI supports customizable fee structures.

Q: Is historical data available?
A: Currently, the tool processes live data only.

👉 Learn more about OKX WebSocket APIs