How to Run

1. Install dependencies: pip install -r requirements.txt

2. Start backend and live ingestion: python app.py

3. Start dashboard: streamlit run ui/dashboard.py

The backend runs FastAPI for analytics and alerting while ingesting real-time data from Binance WebSocket. The Streamlit dashboard consumes backend APIs for visualization.

Project Structure

quant_analytics/ │ ├── ingestion/ │ └── websocket_client.py # Binance WebSocket client │ ├── state/ │ └── market_state.py
│ ├── analytics/ │ ├── hedge_ratio.py
│ ├── spread.py
│ ├── zscore.py
│ ├── correlation.py
│ └── adf_test.py
│ ├── alerts/ │ └── rules.py
│ ├── backend/ │ └── api.py
│ ├── ui/ │ └── dashboard.py
│ ├── app.py
├── requirements.txt └── README.md

Project Explanation

1. Data Ingestion Live tick data is streamed from Binance Futures WebSocket Data is normalized into {timestamp, symbol, price, quantity} Stored in: 1. SQLite (persistent) 2. In-memory rolling buffers (fast analytics)

2. Resampling Tick data is resampled into configurable intervals: 1s, 1m, 5m OHLC bars are generated backend-side

3. Analytics Engine (Backend) All analytics are computed in Python backend:

   Metric Description Hedge Ratio - OLS regression for pair hedging Spread - Price difference adjusted by hedge ratio Z-Score - Standardized deviation of spread Correlation - Rolling correlation between assets ADF Test - Stationarity check for mean reversion Signal Quality Score - Composite confidence metric (0–100) Half-Life - Speed of mean reversion (bars)

   Analytics only activate after sufficient data is collected, ensuring statistical validity.

4. Alerting Logic Alerts are triggered only when: Spread is stationary (ADF p-value < 0.05) Z-score exceeds threshold Correlation is sufficiently high This prevents false or unstable signals.

5. Dashboard (Frontend) The Streamlit dashboard provides: Live prices for both assets Spread with mean-reversion bands Z-score with entry thresholds Key metrics (hedge ratio, confidence score, half-life) System warm-up status The frontend does not perform analytics — it only consumes backend APIs.

6. Key Design Choices Backend-first analytics for correctness In-memory + persistent storage for performance & durability Warm-up gating to avoid unstable statistics