How AI Can Outpace Uninformed Social Traders

A phenomenon that’s become increasingly common in the financial world is so-called social trading. The term was popularised by the trading site eToro, which now boasts around 3.8 million customers in the UK, and it refers to a strategy where traders leverage social networks and community-driven platforms to make financial decisions. 

Social trading allows you to follow and copy the trades of more experienced traders, share trading insights, strategies and market updates, and learn through observation and interaction with other traders. 

Advocates of social trading (principally the platforms that make money from it) claim that it’s “democratising investing” by enabling novice traders to benefit from the experience and expertise of others. Sadly, however, the reality is very different. Social trading is more akin to gambling than investing, and study after study has shown that those who engage in it end up underperforming the stock market by some distance.

New Study Confirms Social Traders Fare Badly

The latest such study, by researchers from Purdue University and the City University of New York, highlights just how badly social traders are performing. (1) It also suggests that one of the key reasons why they struggle is that they’re being systematically outfoxed by professional investors using strategies powered by artificial intelligence. 

The researchers analysed 77 million messages posted by more than 800,000 users on Stocktwits, a social media platform for investors, from 2012 to 2022. They examined how both positive and negative sentiment about specific stocks influenced trading on platforms like Robinhood, and then tracked the performance of those stocks over time. Finally, they compared retail investors’ strategies, inferred from social media messages to AI-powered trading signals developed using machine learning techniques.

Here’s a summary of what they found:

• Social media had a direct impact on the trading decisions that people made. To quote from the paper: “Stocktwits message activities and sentiments are not just a sideshow; they are significantly correlated with the aggregate retail trading activity of the corresponding stocks.”
• Retail investors who acted on social media sentiment generally performed very poorly. The stocks that were tipped to outperform often incurred the heaviest losses. Conversely, the stocks that were tipped to fall in value often outperformed. Over the ten-year period, retail traders saw almost a 40 per cent decline in the value of the stocks they bought, while they would have seen a 30 per cent increase by holding on to the stocks they sold.
• Retail investors who used predominantly fundamental analysis, grounded in financial data like earnings and valuation, tended to outperform the market on average over the course of the following week. However, the positive information value of fundamental analysis sentiment quickly disappeared. 
• Messages based on technical analysis, which focuses on past price patterns, were linked to lower future returns, including very short-term returns. 
• Even those retail investors who self-identified on Stocktwits as fundamental investors in fact hardly used fundamental analysis at all. About 80 per cent of them published almost exclusively technical analysis, based on charts.
• While day traders used predominantly technical analysis, even people who identified as long-term investors were predominantly posting exclusively technical analysis.
• Stocktwits posts expressing strong “buy” sentiment were associated with episodes of herding on platforms like Robinhood, leading to overcrowding. Such behaviour tends to inflate stock prices in the short term, leading to sharp corrections later.
• AI-powered trading strategies, grounded in advanced technical signals, often deliberately trade against retail sentiment. Over the period analysed, these strategies were significantly more profitable when their predictions opposed the bullish or bearish outlook shared by retail traders in social media posts. 
• Overall, AI-powered trading strategies were far more profitable than retail traders in aggregate. A long-short portfolio based on AI-generated signals yielded an annualised return of 10.2 per cent. AI strategies that traded against retail sentiment achieved annualized returns of 13.4 per cent.

Questions for regulators

Interestingly, the researchers warn that their findings raise questions about market fairness and integrity which regulators need to consider. 

They write: “Our documented disparity in investors’ ability to benefit from technical analysis, and social media’s potential role in amplifying this disparity, highlights critical concerns about the fairness and integrity of financial markets in the era of rapid AI adoption by sophisticated players. Striking the right balance between innovation and regulation will be key to harnessing the benefits of AI while mitigating its potential downsides.”

Takeaways for Investors

This latest paper adds to a mountain of evidence that it’s a bad idea for ordinary investors to trade individual stocks. It also exposes the myth that traders can learn to make better decisions by observing how other traders are thinking and acting on social media. The reality is the very opposite: basing trades on what you read on channels such as Stocktwits is likely to lead to inferior results. It also leaves you highly vulnerable to being outmanoeuvred by the AI-driven strategies professional investors are increasingly adopting.

The study also shows how particularly unhelpful technical analysis can be. If you really do insist on trading individual stocks, fundamental analysis may be more useful than relying on charts. Even then, however, you’re still going to find it very hard to beat a low-cost equity index fund in the long run.

This article is produced by us for Financial Advisers who may choose to share it with their clients. Timeline Planning and Timeline Portfolios do not offer direct-to-consumer products.

ROBIN POWELL is a freelance journalist and the editor of The Evidence-Based Investor.

References:

1. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4867401