An interview with Stanley Speel:Q. Stanley, your upcoming webinar is titled: Systematic Portfolio Analysis and Construction with Time Series Data. Is this strictly for Quantitative Investment professionals or would people with a general interest in the topic of investing also find it useful?
A. We’re going to be covering a number of topics ranging from quantitative methods for forecasting, to interpreting predictive uncertainty. Anyone who is interested in extracting insights and information from financial time-series data should find it useful!Q. There was a recent article published in Bloomberg that talked about the “Winter” that systematic funds experienced in the long bull run prior to the Covid crash. Can you shed any light on the challenges that systematic trading has experienced, and how the tools you’re building might play a role in addressing those challenges?
A. Understanding both historical and forecasted risk is vital in constructing models that are robust to such volatile times. Some of the recent challenges faced by systematic strategies include shifts in underlying sentiment and signal behaviour that have never been observed before, resulting in models that are either not constructed or used correctly.
If any systematic investment methods are to thrive in such times, they must be able to adapt to such changes and learn to forget their prior assumptions of normality. This can be addressed in multiple ways, from statistically robust and time-invariant signal selection, to adaptive forecasting models that are able to adapt to varying event horizons. The tools we are building couple well-formulated predictive signals together with principled forecasting algorithms. By doing so, we can address many of these associated risks though surfacing and acting on issues as they arise.
Q. I’m going to ask you to speculate on a scenario. Machine learning tools have been part of professional systematic trading for some time. If retail investors were able to access analogous technologies, how do you think their outcomes might change?
The most immediate change that machine-learning tools will bring to many retail investors is a new level of data-driven support to discretionary decisions, allowing for influencing factors, human bias, and exposure to risk to be better understood and managed. With fundamental research often underpinning decisions for many retail investors, it is unlikely that machine-learning software will make discretionary approaches to trading irrelevant. However, with data sources becoming more ubiquitous by the day, everyone should feel comfortable using quantitative, ML-driven analysis to aid their investment research and decisions.
Q. What do people need to know in order to sign up?
This webinar is for anyone who works with time-series financial data. It does not require any knowledge in ML-driven time-series modelling or forecasting.
Date: Thursday May 14th
Time: 5 PM BST, Noon EST, 9:00 AM PST