Artificial intelligence (AI)-based strategies are being increasingly applied in investing and portfolio management. Their contexts, utility, and results vary widely, as do their ethical implications. Yet for a technology that many anticipate will transform investment management, AI remains a black box for far too many investment professionals.
To bring some clarity to the subject, we zeroed in on one particular AI equity trading model and explored what it can bring in terms of benefits and risk-related costs. Using proprietary data provided by Traders’ A.I., an AI trading model run by our colleague Ashok Margam and team, we analyzed its decisions and all-around performance from 2019 to 2022.
Traders’ A.I. has few constraints on the market positions it takes: It can go both long and short and flip positions at any point in the day. By each day’s closing bell, however, it completely exits the market, so its positions are not held overnight.
So how did the strategy fare over different time periods, trading patterns, and volatility environments? And what can this tell us about how AI might be applied more broadly in investment management?
Traders’ A.I. outperformed its benchmark, the S&P 500, over the three-year analysis period. While the strategy was neutral with respect to long vs. short, its beta over the time frame was statistically zero.
Traders AI Model vs. S&P 500 Monthly Equity Curve ($10k Investment)
Traders’ A.I. leveraged moments of higher skewness to achieve these results. While the S&P 500 had negative skewness, or a strong left tail, the AI model displayed the opposite: right skewness, or a strong right tail, which means Traders’ A.I. had few days where it generated very high returns.
So, where was the model most successful? Was it better going long or short? On high or low volatility days? Does it choose the right days to sit out the market?
On the latter question, Traders’ A.I. actually avoided trading on high return days. It may anticipate high risk premium events and opt not to take a position on which direction the market will go.
Traders’ A.I. performed better on a market-adjusted basis when it went short. It made 0.13% on average on its short days while the market lost 0.52%. So the model has done better predicting down days than it has up days. This pattern is reflected in bear markets as well, where Traders’ A.I. generated excess performance relative to bull markets.
Finally, the AI model performed better on high-volatility days, beating the S&P 500 by 0.19% a day on average while underperforming on low-volatility days.
AI Model’s Return Percentage vs. VIX Percentage Change
All in all, Traders’ A.I.’s results demonstrate how one particular AI equity trading model can work. Of course, it hardly serves as a proxy for AI applications in investing in general. Nevertheless, that it was better at predicting down days than up days, succeeded when volatility was high, and avoided trading all together before big market-moving events are critical data points. Indeed, they hint at AI’s vast potential to transform investment management.
For more on this topic, don’t miss “Ethics and Artificial Intelligence in Investment Management: A Framework for Professionals,” by Rhodri Preece, CFA.
If you liked this post, don’t forget to subscribe to Enterprising Investor.
All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.
Image credit: ©Getty Images / Svetlozar Hristov
Professional Learning for CFA Institute Members
CFA Institute members are empowered to self-determine and self-report professional learning (PL) credits earned, including content on Enterprising Investor. Members can record credits easily using their online PL tracker.