Will the algorithms beat Warren Buffett?|Lesaffaires.com

PORTFOLIO MANAGEMENT 4.0. In May 1997, the six-game chess rematch took place between Garry Kasparov and the computer Deep Blue, whose program had been improved after its defeat the previous year. For the first time in history, the world champion bowed against a machine which at the time measured 1.80 m, weighed 1.4 tonnes and required 20 people to operate! This victory, although disputed at the time, announced the beginning of a new era for machines and artificial intelligence.

Nearly a quarter of a century later, algorithms are now everywhere. In the world of portfolio management, artificial intelligence is mainly used in risk and transaction management, with more than 80% of the daily trading volume now being performed by algorithms. However, according to the Alan Turing Institute, only 9% of hedge funds say they use artificial intelligence for stock picking and investment idea generation.

Returns

What are the results? It is still early to conclude. The firm Cerulli Associates reported in May 2020 an average return over three years of 34% for funds using artificial intelligence, almost three times better than that of their benchmark indices.

The EurekaHedge AI Hedge Fund Index claims otherwise, noting an underperformance of around 1.5% annually against the benchmark. And the hedge fund considered by many to be the most successful in history, the Medallion fund from Renaissance Technologies, also strikes the imagination. Indeed, algorithm pioneer Jim Simons offered his investors, from 1988 to 2019, an average annual return of 66%, far ahead of Warren Buffett, George Soros, Peter Lynch and Ray Dalio. Has the age of machines arrived on Wall Street? First you have to define them. We are referring here to automatic learning (or machine learning) which allows computers to learn without having been previously programmed specifically for this purpose. This learning requires big data since to learn and grow, machines need massive streams of data to analyze.

Machines have no ego

The benefits of using artificial intelligence are numerous. First, the “machine” has no ego, stock aversion, or investment doctrine. She continually admits her mistakes and then corrects them and improves the algorithms the next day. The computing power has no equivalent in humans, with more than 10,500 financial instruments analyzed and millions of calculations performed daily.

For fans of fundamental analysis, the machines scan the financial statements of thousands of public companies on a daily basis, analyzing profits, debt, market shares, etc. For the disciplines of technical analysis, hundreds of factors including moving averages, volume, and support and resistance levels are continually fed into the algorithm.

The machine is however also “creative” and constantly generates new ideas to then test them using its vast database. For example, it can track social media platforms to detect investment opportunities, which it did in 2020 with GameStop (GME, US$167.52) and AMC Entertainment Holdings (AMC, US$27.47) . She can try to cross the price of bitcoin with those of American banks to find a new relationship increasing the reliability of her models on the Etherum cryptocurrency. Or it can analyze the volume of insider buying in a particular industry to anticipate a new hostile acquisition target.

The algorithm also chooses unloved titles; it detected Bombardier (BBD.B, $1.70), in March 2021, which was then trading at $0.66, long neglected by many investors. Finally, the machine excels even more on the stock market indices, with average success rates of 97.2% on its forecasts on the main American indices over one year.

The disadvantages

Even if emotion must be put aside, human intervention still seems necessary to us. The machine's stock picking should be integrated with a solid portfolio structure, rotation frequency and number of stocks, among others.

Second, the algorithms are opaque and often influenced by a stock's recent momentum, which can create false buy signals. Even with a strong recommendation, not all investors would feel able to buy a stock such as Nuvista Energy Ltd (NVA, $6.25) after rising more than 600% since the start of the year. Finally, the algorithm is not immune to human intervention, such as the new regulations of the technology sector in China, the American elections or a possible financial scandal.

GUEST EXPERT Richard Langevin, Vice President and Portfolio Manager at Nymbus Capital