Commentary by Dr. James McCann, Chief Scientist, Plotinus Asset Management
The news that the Norwegian sovereign investment fund is employing sports psychologists and crash simulations to train its investment managers shouldn’t be surprising. A pre-Covid walk through an airport bookshop is enough evidence that the world of business is overloaded with self-help seminars and get-rich-quick schemes that promise to let you in on the secrets of success. Perhaps these books offer a more honest account of how to get ahead than the business school MBA, and a great deal cheaper.
The financial benefit of having an astute, well-rested, highly-trained trader should be obvious. It is why banks offer complementary falafels and gym memberships. It is good business. The boozy lunch and the four-hour round of golf is replaced by paint-balling and fire-walking. But the ideas of teamwork and business philosophy are the key to keeping your star managers humble. War Gaming markets is one method of testing if your cadre of financial officers can act in a disciplined and efficient manner in a crisis. The idea is that a sequence of simulations of bad things happening can create muscle memory that will be useful in the future. It is the capital market version of a fire drill.
War Gaming markets is one method of testing if your cadre of financial officers can act in a disciplined and efficient manner in a crisis.
Simulations can be made very realistic. This world of financial virtual reality is a vital training ground for the staff and the financial institution. In fact, such stress tests are a now commonly required by regulators to ensure there is a safety net in place. The role of AI in this field covers many aspects. The most important of these is the most humble: book-keeping, making sure that the numbers add up correctly. In practical terms, AI allows us to click a trade and allow the machine to do all the important housework, including automatic hedging, for example. It is a short step to implementing error detection, anomalous transfers, and fraud. Our most common daily interaction with AI is the spam filter on e-mail.
However, spotting an anomaly is one thing, knowing the action to take is another. For an AI-trained trading platform, black-swan events are most often treated as data errors. Attempting to develop a model to trade in reaction to such events is problematic. If we bend the model to account for these black swans, then there is a risk that it ruins the model for the more prosaic bread-and-butter trades. It can also lead to a domino series of adjustments such as occurred in the “flash crash.” Technically, this is called overfitting. This explains why human interaction is important when an emergency arises, and the glass needs to be broken. But AI is incredibly good at keeping a watchful eye and applying the handbrake when required. So, we still need to war-game extreme events. We are not quite at the stage where the automatic pilot controls the entire financial flight, and we sit back enjoying the complimentary charcoal-grilled lamb brochette with cinnamon-scented sauce, but we are not far away. ■
February 2021
© 2021 Plotinus Asset Management. All rights reserved.
Unauthorized use and/or duplication of any material on this site without written permission is prohibited.
Image Credit: Estradaanton at 123RF.