AlphaGo could mean the end for active managers

Artificial intelligence victory offers a glimpse into the future for investment management

FacebooktwitterlinkedinmailFacebooktwitterlinkedinmail   by John Spiers, 15th March 2016

News that AlphaGo, a computer program developed by Google, has beaten the world champion at Go – a game with more combinations than atoms in the universe – has got me thinking about whether there is still a role for humans in managing money.
Humans are poorly designed for investment management due to our emotional issues. We overreact to danger – a good thing when we needed to escape from sabre toothed tigers but not so helpful when dealing with a drop in share prices. We are overconfident in our predictions and then we deal badly with mistakes, which leads us to hold on to bad investments too long.

the market will become even more efficient at pricing in all known information

It’s precisely because of these failings that some humans (e.g. Warren Buffett) have been able to achieve much better returns than others. It’s also allowed some computer based strategies (e.g. MAN’s AHL Diversified) to perform well. If, or perhaps when, computer managed funds become the norm we can still expect to see variations in performance. Think of a coin tossing game: although the odds are equal for everyone, some will be luckier than others. But over time the influence of luck will diminish and everyone will gravitate to 50% success. The stock market works in a similar way.

Proponents of active managers will argue that their skill involves as much art as science and that there are still pockets of the market where dedicated research can provide an informational advantage, such as smallcaps. I agree with the latter but one of the scary aspects about AlphaGo is that it doesn’t just process millions of combinations looking for the best move. It learns from its mistakes and looks for patterns in a similar way to our brains.
If the investing universe becomes dominated by fast reacting, unemotional machines then the market will become even more efficient at pricing in all known information. That would lead to less scope for outperformance, which will increase the attractions of low cost index funds.

There will still be scope for some computer investment managers to be smarter or faster than others. It’s intriguing that ‘Alpha’ is the technical investment term given to adding value through good stock selection. Perhaps AlphaGo will start managing money when it isn’t playing games? If so I’ll be tempted to give it a try.

Search the EQ Library

View articles by topic: