regholl2023 / trading-engine-alpaca

Trading Engine on Alpaca Broker

Introduction

This project presents a trading engine tailored for medium-frequency trading environments on Alpaca Broker(Why Alpaca? It is comission free). Built with Python, it leverages asynchronous programming and object-oriented principles for efficient market data processing and trade order management. The engine is versatile, supporting sophisticated trading strategies like one demonstrated here, which uses a Hidden Markov Model for market regime identification.

Key Components

• DataClient: Fetches and processes real-time market data, including 1-minute bars, trades, and quotes, alongside order updates via WebSocket.
• PositionManager: Manages and updates position information for different instruments within the trading engine.
• OrderManager: Handles the execution and management of diverse trade orders.
• RegimeTrader (Strategy File): An example strategy that utilizes DataClient to gather minute bar data, calculate features (log returns, fractions of the bars), and determine market regime with a Hidden Markov Model, guiding long or short positions based on identified regimes.

For more details, please refer to the core and strategy files in the repository.

Thoughts on Latency

• Broker Response Time: Approximately 200ms for equity orders with Alpaca Broker.
• Network Latency: Due to data and account server locations, there's an inherent network latency (estimated at around 20ms). https://forum.alpaca.markets/t/alpaca-data-center-locations/5640
• Deployment Suggestion: Using AWS in North Virginia or Google Cloud in us-east4 can potentially reduce application-server latency to around 10ms.
• Internal Latency: Depends on strategy complexity (e.g., < 20ms for the HMM Strategy).
• Round Trip Time: Minimum expected is around 250ms from market data receipt to order response.

Improvement in the Future

• Modify Credentials class for more secure key management(e.g. load credentials from .env)
• Implement limits on stored order IDs in OrderManager.
• Add order amendment and cancellation functionalities.
• Develop RiskManager and SystemMonitoring classes for risk oversight and latency monitoring.
• Transition to an event-driven architecture with a Strategy Abstract Base Class (ABC) and EventHandler class.
• Introduce a user interface.
• Expand capabilities to include options and cryptocurrency trading.