16 October 2019
Systematic strategies have a long history in the field of investment area, encompassing the high-frequency ones as well as low-frequency strategies. Over the last decade, the rise of ETF, Robo-allocator made them a popular choice compared to discretionary strategies. More recently, progresses in machine learning renew the theoretical development in that field as well as highlight new perspectives.
Here, we focus on low-frequency strategies and first recall briefly the history of such strategies through a common statistical framework (dynamic basket allocation): Markowitz, CPPI, Buy-Write, Vol. Control, Risk Budgeting, Factor-based, Arbitrage based,... We illustrate those strategies through actual use cases and highlight the importance of underlying risk framework.
In the second part, we focus on the various machine learning methods available to develop or optimize systematic strategies. Especially, we underline the paradigm difference with traditional statistical/stochastic methods by deepening on the fundamental concept of learning vs calibration, as well as the role of prior knowledge.
In the final part, we will evoke some potential future research to go beyond the paradigm of covariance matrix: neural control, graph representation learning.