The hottest headlines in technology, whether it be autonomous driving, government surveillance, targeted ads on Facebook or quantitative investment strategies have been surrounding the ubiquitous idea of “machine learning” and “big data”. Many of the largest quantitative firms are spending significant amounts of their budgets and many years researching artificial intelligence and how it can be applied to financial markets. That said, there seems to be diversity of opinion regarding effectiveness and none seem ready to completely handover their processes to the artificial intelligence robots – the quintessential black box. When doing so a firm that was likely hypothesis-driven would move to being more data-driven, which can certainly have its challenges.
One of the great benefits of having simpler, hypothesis-driven strategies is the ability to understand what exactly is driving a portfolio’s profit and loss. Many times the difference between a successful strategy and an unsuccessful one isn’t the ability to generate outsized returns, but the ability to stick with it when it is under-performing. A strategy with a large edge, but one that is opaque is much harder to hold during the inevitable rough times. On the other hand, investors in a simpler strategy, even one with lower overall expected returns, may be able to weather the difficult periods because those periods can be anticipated and quantified ahead of time. A strategy with positive returns that can actually be followed will always beat the higher expected return strategy that gets jettisoned at the first signs of distress. Something we need to keep in mind as newer (and many times much cooler) quantitative methodologies continue to proliferate.