Machine learning has the potential to transform our world, just as electricity transformed almost everything around a century ago. However, it's hard to imagine how your profession will be affected. So, let's consider the potential of machine learning for investment managers.
Implementing an investment philosophy
Investing requires a philosophy to act as a framework for finding, evaluating and selecting new investment opportunities. These philosophies enable investment managers to differentiate themselves and thereby attract a particular client base.
The implementation of any investment philosophy has two components. The first involves introducing the thesis to the real world, testing, validating and refining it by asking the right questions. The second is its execution.
Both components are vital to the success of an investment, but the execution - what you buy, how much you invest, when you make a deal - will cause a good investment to be a failure or success. Moreover, this needs to be scaled to a wide enough universe of stocks in order for a fund to allocate sufficient capital to generate a positive return for investors.
The benefits of machine learning for investment managers
Investing is therefore extremely complex. A deep and thorough analysis of many internal and external factors is required to find stocks that qualify for the philosophy and bear the early signs of significant upside potential.
However, machine learning for investment management could provide a competitive edge in the time-constrained and resource-heavy execution phase of any chosen philosophy.
By learning the relationships between key factors of a philosophy and future price performance, machine learning can identify at scale candidate stocks, and help investors optimise their research analysis by narrowing down viable options.
What are the limits of machine learning for investment banking?
Machine learning is not as abstract as it may appear, and in many cases it is simply an extension of the widely-used linear regression. However, machine learning algorithms are much better than humans and linear regression when it comes to handling edge cases, where the relationships between the input and target variables are non-linear, or where there are many dimensions to the problem.
Nevertheless, the success of machine learning for investment banking is tied to the quality of questions that an investment manager tasks it to answer.
For example, machine learning can identify and anticipate early signs of sharp price drops or rises, as well as forecast the accuracy IBES analyst estimates.
Moreover, because machine learning models are highly flexible, possible questions that can be asked are only limited by the creativity of investment managers.
In conclusion, machine learning will not replace investment managers, because it cannot think on its own. However, in time, individuals who use machine learning for investment banking will replace those who don’t.