Gemscap Quant Analytics

A complete end-to-end real-time quantitative analytics dashboard for MFT style trading.

Features

• Real-time Data Ingestion: Connects to Binance Futures WebSocket for live tick data (BTCUSDT, ETHUSDT).
• Efficient Storage: Stores normalized trade data in a local SQLite database with indexing for fast retrieval.
• Advanced Analytics:
  • Real-time Spread Calculation (BTC vs ETH).
  • OLS Hedge Ratio estimation.
  • Rolling Z-Score with customizable windows.
  • Augmented Dickey-Fuller (ADF) test for stationarity.
• Interactive Dashboard: Streamlit-based UI with Dark Mode, interactive Plotly charts, and auto-refresh.

Quick Start (Single Command)

1. Install Dependencies:

   pip install -r requirements.txt

   streamlit>=1.30.0 plotly>=5.18.0 pandas>=2.0.0 statsmodels>=0.14.0 websockets>=12.0 aiohttp>=3.9.0 watchdog>=3.0.0

2. Run the Application:

   python run.py

   This will start both the background ingestion service and the frontend dashboard.

Architecture

Architecture Diagram

Components

1. Ingestion Service (backend/ingestion.py): Asynchronous WebSocket client that subscribes to trade streams, buffers data, and batch writes to the database.
2. Storage (backend/database.py): SQLite interface handling efficient insertions and time-window queries.
3. Analytics Engine (backend/analytics.py): Pandas/Statsmodels module for rescheduling tick data to OHLC, computing spreads, and running statistical tests.
4. Frontend (frontend/app.py): Streamlit application that queries the DB, runs analytics on the fly, and renders visualizations.

Usage

• Configuration Sidebar: Adjust the "Lookback Window" to analyze different time horizons. Change the Z-Score window to tune sensitivity.
• Alerts: The Z-Score metric will turn RED and show an ALERT label if the value exceeds the configured threshold (default 2.0).
• Data Export: Use the "Download CSV" button in the Data tab to export processed OHLC data.

Methodology

• Resampling: Tick data is aggregated into OHLC candles (1s, 1m, 5m) to handle high-frequency noise.
• Pair Trading Logic: We model the relationship between BTC and ETH. The spread is calculated as $Price_{BTC} - \beta \times Price_{ETH}$, where $\beta$ is derived from OLS regression on the selected lookback window.
• Stationarity: We use the ADF test to confirm if the spread is mean-reverting, which validates the Z-Score signal.