Durenard, Professional Automated Trading

I’m in over my head with Eugene A. Durenard’s Professional Automated Trading: Theory and Practice (Wiley, 2013), so consider this post more of a notice than a review.

Durenard describes how to set up a framework to research and select trading models and to implement them in a real-time low-latency environment. The book “requires readers to have some knowledge of certain mathematical techniques (calculus, statistics, optimization, transition graphs, and basic operations research), certain functional and object-oriented programming techniques (mostly LISP and Java), and certain programming design patterns (mostly dealing with concurrency and multithreading).”

Durenard focuses on the design of trading strategies as trading agents, the goal being to build “robust trading systems that can gracefully withstand changes of regime.” He introduces swarm systems, which are “aggregate agents that embed various types of switching mechanisms.”

The assumption underlying Durenard’s framework is that markets are complex adaptive systems best understood, and exploited, by an aggregate adaptive agent. This agent has a collection of nonadaptive strategies at his disposal. The agent “is endowed with criteria to choose a subset of behaviors that is expected to produce a positive performance over the next foreseeable future. This is the behavior that the agent implements in real trading. As time unfolds, the agent learns from experience to choose its behavior more effectively. Effectiveness means that as the market goes through various cycles of regime changes, the performance during those change periods does not degrade.”

Durenard draws on concepts from evolutionary theory and learning to endow trading systems with opportunism, robustness, and flexibility. Learning is important because a swarm system needs not only behaviors that have proved effective in the past but also a degree of innovation. The innovation problem is “an active area” of Durenard’s current research.

This book is divided into four parts: strategy design and testing, evolving strategies, optimizing execution, and practical implementation. It offers its fair share of code to help the reader along—unfortunately, not this grossly under-qualified reader.