How AI can transform the stock market

Algorithmic trading is nothing new, but artificial intelligence is taking it to a whole new level.

Financial trading has always suffered from a certain problem – people. Humans are inevitably fallible. They can be slow, get tired and make mistakes and, what’s worse, can be influenced by the fluctuation of their emotions, which is why algorithmic trading is growing in popularity. Now, with the arrival of AI technology, it’s becoming even more important.

Rise of the machines

Of course, algorithmic trading is nothing new. Financial institutions have been using it for years. Ever since the 1600s, the aim has been to use technology to increase the frequency of trades, get information before anyone else and to remove the guesswork from trading decisions.

That last point is crucial because the human element can be a major weakness. Studies have shown that stock markets are driven by the vagaries of human emotions. Greed, disappointment, fear, anxiety, and overconfidence can all cause mistakes.

When things have been going well, traders can become too aggressive. Likewise, when they haven’t they can become cowed and overly cautious. The aim of algorithmic trading is to take the human element out of it and make trading a more rational fact-based process. The hope is that funds will perform better.

A team of researchers used AI algorithms to show that it can out-perform the market. In a test using old market data and real-time trading, their system smashed the market average. One system scored annual returns of 73% between the trading years 1992 and 2015, compared with the real market average of just 9%. The computers performed particularly well during times of market turmoil, such as the financial crisis of 2008, when the task of picking good trades becomes particularly difficult.

This is why so many hedge funds and pension funds are taking advantage. As early as 2015, most of the best-paid Hedge managers were relying on machines to do their trading for them. The top managers took home more than $13bn, out-performing the revenue of many nations, but they could not have done it without machines.

phone displaying stockmarket performances
Global algorithmic trading market will rise at a 10%+ CAGR by 2020 (source: Research and Markets)

One of the biggest Hedge Funds, Man Group PLC is now investing heavily in AI after several years spent experimenting with the system. The system evolved autonomously to identify money-making strategies which human traders had missed. What was groundbreaking about this system was that the engineers themselves couldn’t explain why it was working. The system was so autonomous it was trading in ways its own designers couldn’t understand.

That autonomous nature is also one of its problems. Man Group PLC delayed implementing the system because the explanation ‘I don’t know why’ wasn’t going to work well with their clients. The technology was so complex and self-evolving, that humans couldn’t keep track of it. That lack of knowledge and the fear of the new is one thing which holds systems back.

Trouble ahead?

AI is not, it should be added, without its problems. It’s a new technology, far from maturity, and systems can vary in effectiveness depending on who designed it. A bad algorithm can be costly. Take the example of Knight Capital which had to shoulder $440million in losses after a faulty software test.

Some insist machines can’t fully emulate human intuition. Even the most successful investors in the world who have openly shared their techniques rely on qualitative judgments about the strength of a company and the people who lead it before making a decision. Will an AI platform be able to make the same assessments?

The technology is also far from mature. For all the hype surrounding AI much of its promise lies in the future. The time may come when no human investor can beat a machine. Some of the results being declared by AI trading companies are indeed eye-catching, but it still has some way to go before it reaches its goal.