Want to beat your benchmarks? There’s a machine for that.


When there’s blood on the streets, buy. It’s a market of stocks, not a stock market. The trend is your friend, until it’s not. Sell the rumour, buy the fact. Be a long-term investor, and stay the course.

We’ve been hearing these little investment truisms for years. In fact, we hear them to this day. That’s because investors have traditionally built their investment strategies around rules. They find patterns, and stick to them, hoping to be the next South African Warren Buffett. Sadly, not everyone can be Buffett – and the reality is that most who try to follow him underperform the benchmarks.

There are a few reasons for this. One is that the market is crowded with highly gifted active managers who have adopted the long term, value-biased approach to investing. However, many managers simply do not have the budgets to attract the quality and breadth of research required to beat their peers at this particular style of investing.

Another reason is that many investment houses simply are too rigid in their approach, and confuse sticking to their investment style with stubbornly holding views that may not be supported by empirical data trends. In being faithful to a particular investment style, many investors are simply not looking for returns from other sources – a good example being passive investors.

But perhaps the biggest reason for underperforming benchmarks is that not enough active managers are using modern technologies, like artificial intelligence (AI) and machine learning, to find the opportunities amidst the mayhem of the global markets.

Let’s make one thing clear upfront: AI and machine learning techniques are not a substitute for an investment strategy or a portfolio manager. They’re investment tools that unpack more data than humans can possibly digest to provide an information edge, and ultimately deliver a material advantage for investors.

There’s a perception in the industry that AI models are only useful in identifying technical, short-term trading opportunities. And make no mistake, finding numerous smaller opportunities at a lower risk offers an alternative – and lucrative – way of beating average returns. There’s a lot more certainty in smaller gains than waiting for the unicorn opportunities. To use a cricketing analogy, many fund managers want to hit sixes, and forget that you can score well in other ways, with far less risk.

In addition to the patient, long term investor, there are also large global banks that use high frequency trading to exploit momentary mispricing (for example, when the price of BHP is 5c higher on the JSE than on the LSE). This requires a huge amount of capital and technology dedicated to being faster than everyone else. Therefore, in our view, the market is like a barbell: a significant amount of effort at the very short investment timeframe, and the same at the longer-term horizon.

For us, there’s a huge untapped source of alpha in between: lots of small wins. Fund managers are often so focused on the barbell, or looking for the next big thing, that they overlook the amazing space in between. Many investors see a blip. We see an opportunity to make money.

There’s far more to AI than quick wins, though. It’s superb at picking up the patterns that investors would probably pick up themselves if they had the time. AI could allow them to find a broader range of investment ideas, and identify mispriced assets more easily. But the real value of machine learning and statistical methods is that these are incredibly powerful risk mitigation tools. It is difficult for human beings to process how portfolios will act in various scenarios, an in our opinion machines are superior for this task.

AI also brings to the table an uncanny ability to use alternative sources of data. It’s hard to compete when everyone’s managing the financials. But throw in the ability to go beyond share prices to find differentiated data sources, and build an investment model around it, and suddenly you have an edge few others do.

That’s why where typical fund managers employ CFAs, we have machine learning experts who have honed their skills in real-world industrial applications other than investments. They not only have advanced machine learning skills, but also the ability and nous to pull that data together in a manner that supports traditional fundamental analysis.

So why isn’t everyone doing it? Well, increasingly they are. There’s been a rapid uptake of AI and machine learning in the investment industry over the past three years. Since 2016, the CFA Institute has been surveying institutions to gauge the adoption of data science techniques across the investment value chain. In 2019, 10% of portfolio managers were using AI or machine learning techniques. By 2020, this number had almost doubled, with a further 7% of portfolio managers that were in the process of implementing AI in their investment strategies.

We think there’s immense potential for local investment firms to generate strong alpha by adopting AI technologies. What they need is a sound understanding of the nature and drivers of the mispricing that they wish to exploit, and to use this knowledge to build effective machine learning models.

What they’ll find is a world of opportunities they never imagined possible; a level of automation that cuts out a staggering amount of manual admin work; and the ability to free their time up to focus on the things that really matter, like ESG and delivering real value. That’s the future of investing right there.