The science behind investment algorithms

ASIA'S capital markets over the past couple of years have been nothing short of volatile, but Covid-19 gave the global markets one of the biggest shocks in recent memory. While we have had a year to adapt to the new normal - and we are seeing markets rallying - it will take some time before we see long-term stability. Until then, the rebound environment remains fraught with risks.

Volatility is undeniably part of our new normal, and the current environment has pushed investors to be more careful with preserving the assets they have on hand. For those who seek to actively navigate the still tumultuous landscape, digitally powered asset preservation algorithms can help them manage their portfolios with efficiency and scale unmatched by humans, especially since many are now powered by artificial intelligence (AI).

However, there is only so much that AI-powered algorithms can do. Even when built with best-in-class inputs by experienced investment managers, these robo-advisors remain limited in their predictive capabilities, which is especially vital to help investors make sense of markets when they are still in flux. It is here where the human element remains vital.

Essentially, capital preservation is where investors focus on protecting the absolute monetary value of an asset as measured in nominal currency. It is a typical strategy adopted during uncertain times as the main objective is to reduce the risk of loss.

However, smart investors do not want to just guard themselves against risk, even in this current environment. They do not want to leave their money idle but expect a certain amount of monetary increase or return - all while keeping their underlying assets intact.

But "return" is not something as objectively well-defined as "risk". It is highly subjective; for instance, one person may correlate a good "return" with getting the maximum profits out of their investments, while another may think of "return" as being equivalent to meeting a certain personal financial milestone (like repaying a home loan in five years). Basically, investors want to optimise their portfolios without losing what they currently have. And in all cases, an asset preservation model would work best.

The rise of algorithms

When it comes to portfolio optimisation, most methods extend from the mean-variance approach proposed via the Modern Portfolio Theory (MPT) pioneered by Harry Markowitz. The theory posits that assets can be allocated based on a computation of risk and return that does not view either characteristic in isolation. Instead, they should be assessed by how the investment impacts the overall risk and return of a portfolio.

For example, an investor may build a portfolio of multiple assets that will maximise returns for a specified risk level. In this same vein, with a specific level of expected return, an investor can also build a portfolio with the lowest possible risk. Due to the use of statistical instruments such as correlation and variance, the performance of one individual's investment has less impact on the entire portfolio. The caveat is that the MPT assumes that investors are risk-averse, but it is a particularly salient assumption in the pandemic-struck environment.

While the MPT is particularly useful for these times, doing the risk-return computation of algorithms at the speed and scale needed to exhaustively search for the best portfolio can no longer be feasibly done by humans alone. Also, given today's highly complex state of capital markets, human emotions and cognitive biases will have more impact on objective portfolio assessments.

Hence, the popularity of robo-advisors has exploded in recent years to help meet the growing demands for automated, algorithm-based services. In fact, industry reports suggest the global algorithmic trading market size is expected to grow from US$11.1 billion in 2019 to US$18.8 billion by 2024, expanding at a compound annual growth rate of 11.1 per cent.

Still, the caveat of digitally powered algorithms themselves is that they all differ from one another, so their effectiveness rests heavily on the investment experts building them. This also means that the best algorithms should be based on the experts' understanding of the current markets, while also being able to reduce human bias.

Making asset preservation smarter with both AI and humans

The effectiveness of asset preservation algorithms depends on the human minds behind them. This means investment experts must ensure that portfolio identification is conducted with asset preservation as the underlying investor objective.

This is where the catalytic role of AI comes in. When used for asset preservation algorithms, the technology can be developed to apply the best methodologies to select the best portfolio design for every individual's needs - often from a pool of thousands of permutations. To exemplify: at each stage, curated strategies are tested against user-defined objectives. Then, only the "fittest" strategies should be chosen to be evolved into a risk-limiting model which optimises asset preservation, while maximising user-expected returns.

But by which criteria do the algorithm vet the strategies? This is where the role of the experts comes in. As AI feeds on data to make itself more intelligent, experts must use historical data and simulations of real and improvised stock market scenarios so that the algorithm can understand market movements and patterns.

Allowing AI to learn from simulated scenarios enables it to generate risk-return profiles that are stable over a range of actual stock movement. To do this, investment managers must keep the algorithm robust and on-point with market trends; they must actively track recommendations and the algorithm's output so that expert-based modifications can be made based on expected market fluctuations. This then helps make the algorithm unbiased to short-term fluctuations.

Algorithms must go beyond just AI

While AI-based digital algorithms have brought about a much-welcome transformation in how investors manage their portfolios, we must still be cognisant of their current limits. As algorithms learn from the past and use repeated patterns to suggest the best course in the present tense, they cannot predict the future. This means they cannot pre-empt a trade war, geopolitical tensions, or a global pandemic.

Also, asset preservation strategies typically have investment horizons longer than two years, so it is important to have some inkling of the future. Therefore, the role of experienced investment experts remains as important as ever - only humans can step in to rate assets on an ongoing basis, change their weightage on client portfolios, as well as modify and update an asset's expected returns. These ensure that the algorithm's recommendations are always real-time and give investors a more comprehensive, balanced view of their investments.

• The writer is head of private wealth at Kristal.AI