Far from replacing human expertise, machine learning in the investment world will allow professionals to better differentiate themselves based on the questions they ask. However, the nature and quality of the questions asked can have a huge impact on the relevance of the results.
Machine learning can help investment managers gain an edge over their peers in the execution of their investment strategies. Now they have a way to test their philosophies on real data, get a better understanding of the stocks they are investing in and explore the drivers of their performance.
One particularly important role of machine learning in the investment space is the optimisation of research time. Algorithms are able to do much of the heavy lifting required to narrow down the number of candidate stocks to be analysed.
However, while machine learning algorithms 'learn', this process does require some guidance in the form of questions.
Asking the right questions of machine learning in the investment space
For instance, asking a model to forecast the future price performance of a stock will lead to much worse results than if a model was asked to forecast whether a stock’s price will increase or decrease. The latter question is much easier to solve, and has two possible outcomes as opposed to the infinite number of outcomes possible in the former question.
Moreover, imagine if an investor was able to predict the future price of a stock with 100% accuracy. To act on it, they would have to measure how much of an increase or decrease this would represent relative to the present price. Only then could they decide whether it would be a good investment or not.
As a result, asking the right question in this case would not necessarily mean to forecast a rise or decrease, but to ask whether the increase or decrease would exceed the investor’s required threshold for to invest (e.g. +10%).
The potential of machine learning in investment is enormous, so long as investors are able to approach questions with this logic in mind.
The devil is in the detail
More broadly, the questions’ spirit should aim to assess a potential alignment with a thesis as opposed to its exact realisation. In order to achieve this, it is important to break big questions down into several smaller ones. Not only will they be easier to answer, but the insights uncovered will be more relevant and useful since the models will be more accurate.
Machine learning has created an opportunity for investors to test theories on real data, gain a better understanding of stocks and explore their performance drivers.
Moreover, the efficacy of data and machine learning in the investment world can be quantified much more easily than other forms of research. As a result, individuals now have the visibility and accountability to refine their questions and improve their own ways of working.