Imran Lakha, Senior Advisor, Vanguard Capital AG & CEO and Founder of Options Insight, Financial Markets Training
What is factor investing?
Factor investing has been around for over a decade and has been especially popular among quantitative equity investors. Essentially, it is a way of filtering stocks based on specific types of factors such as Value, Momentum (both price and earnings), Quality, Riskiness and Size to mention a few. This enables one to explore companies from different industry sectors and regions through an alternative lens and identify relative value trading opportunities. Also known as smart beta, these quantitative styles are supposed to allow investors to extract some risk premium from the market which over time should generate excess returns. Analysing a portfolio in terms of its quant factor exposure can help investors to identify biases or risks that they might have otherwise missed and enables better risk management of their assets.
However, as I mentioned earlier, these are not ground-breaking or new approaches and given the sheer amount of AUM that goes into these strategies nowadays I would argue that their value proposition is becoming less clear. It is often now the case that broad market indices don’t move much but the factors (often referred to as the market internals) are having extremely large moves relative to each other, creating all kinds of pain for quant investors. The defined nature of what they should or shouldn’t be investing in (especially around MSCI index rebalances) leads to position crowding which almost always creates pain and sub-optimal returns.
Why use machine learning?
The leading practitioners in this space, often from hedge funds or maybe more surprisingly pension funds, have realised that to maintain an edge, they need to innovate and find factors that are not so commonly followed. This is where machine learning and artificial intelligence comes in. In this context, ML is basically where a massive amount of past stock return and company data is given to an algorithm, and its job is to use statistical analysis in a looping trial and error process to spot patterns in the data that can be used to make money. With the explosion in data availability and processing power, computers are now able to perform a vast number of trails and finds patterns and dependencies in data relatively quickly, something a human would never be able to do efficiently.
The hunt for alpha has also led some firms to use alternative data sources and NLP (natural language processing) with the most sophisticated deep learning techniques to perform analytics and extract meaningful insights. Deep learning refers to the number of layers in a neural network which makes it possible for an algorithm to learn in an unsupervised manner as opposed to supervised ML where it is given targets to help train and optimise itself. It is the unstructured nature of the data being used which requires the use of ML and extreme computational power.
Combining man and machine
Despite this seeming dependence on artificial intelligence, it is also well understood that market domain knowledge is still crucial and human oversight from experienced professionals must be used to ensure sensible investment strategies are being employed. A machine learning algorithm could find some obscure non-linear relationship in the data, but if it doesn’t make any sense from a rational economic perspective then it shouldn’t get implemented. A common sense approach is key and handing over all the controls to the AI and trusting it blindly is obviously not the way forward.
Also worth considering is the quality of the data source, which is often the determining factor in the success of this type of investment strategy. This is another reason why human intervention makes sense for logically pre-processing the data and spotting any biases that may exist before allowing the model to start training. There is no escaping the fact that the quality of a model is dependent on the quality of the data used to build it.
All the institutions with genuine experience using these technologies, see ML as an addition to the investment toolkit and something that assists them in making their trading decisions rather than making the decisions for them. Whilst the large scale adoption of ML and AI in finance is inevitable, it is clear to me that the winners will be the firms that utilise the technology to augment the human capital they already have which has successfully been generating alpha for many years.