Trading in Python

In this talk we recount our experiences building an open architecture, largely open source, event-based trading platform in Python.

Our business objective is to profitably trade short-term price dislocations in liquid, exchange-traded markets in the Asia-Pacific region. To achieve this we are building a modern, real-time (event-time) infrastructure purpose built for intraday trading. It is fully straight-through processing, reducing operational risk and enabling scaling. We use Python for (almost) everything.

The major components are:

• trading algorithm prototyping using ipython notebooks
• market data management using Python wrappers around a market (tick) database called OneTick
• trading signal generation algorithms written in Python
• various libraries including numpy, pandas, cvxopt
• MongoDB to manage position and trade information and Redis for in-memory data management
• messaging using RabbitMQ
• controlling background tasks using Supervisor
• using Django framework for profit and risk reporting and backtesting control