Beating the market is difficult, even with AI's help

Artificial intelligence continues to struggle in picking market-beating stocks.

This surprises many on Wall Street, since AI models supposedly are getting more sophisticated and less prone to hallucinations. But that doesn't necessarily help traders beat the market, since they're competing with others who also are using equally sophisticated AI models to pick stocks. Since not everyone can be above average, the net result is that beating the market remains as difficult and elusive as ever.

Consider the performance of an AI-powered ETF that has a longer track record than the more recently launched ETFs that also use AI to pick stocks. I'm referring to the Amplify AI Powered Equity ETF AIEQ, which was launched in 2017 and "uses IBM Watson [AI] for machine learning, sentiment analysis and natural language processing to select securities." The ETF "utilizes AI to analyze data points across news, social media, industry and analyst reports, and financial statements on thousands of U.S. companies, technical, macro, market data and more."

As you can see from the chart above, the ETF's recent performance has been worse, relative to the S&P 500 SPX, than it was many years ago.

Additional anecdotal evidence comes from a firm that each business day produces a list of stocks that AI rates as most likely to outperform the market over the subsequent three months. Take the 10 stocks rated most highly three months ago: Since then, according to LSEG data, six of the 10 have lost money - including one that has lost 46.3%. The average three-month return of all 10 stocks is a 4% loss, versus a gain of 2.7% for the S&P 500.

To be sure, a few of the newer ETF entrants in the AI-powered stock-picking game have beaten the market over more recent periods. But, as you can see from the above chart, most have not. Going forward, I see no reason to believe that AI-powered strategies, on average, will beat the market more frequently than more traditional forms of research and analysis.

'The Arithmetic of Active Management'

Furthermore, this situation is unlikely to change, regardless of how sophisticated Wall Street's AI models become in the future. Only if you alone are using an AI model that is significantly better than everyone else's will your odds of beating the market become potentially better than a coin flip. But even then, your better-than-even odds won't last long, since those other AI models will undoubtedly catch up to yours in short order.

One reason to be confident that AI's market-beating odds won't improve in any lasting way is a famous article written nearly forty years ago by William Sharpe, the 1990 Nobel laureate in economics. In the article, entitled "The Arithmetic of Active Management," Sharpe showed that active managers, on average, are guaranteed to lag the stock market. Though Sharpe wrote his article before the advent of AI, he stressed that his conclusion depends "only on the laws of addition, subtraction, multiplication and division. Nothing else is required."

Sharpe's argument boils down to the simple insight that beating the market is a zero-sum game before transaction costs - but after transaction costs are taken into account, it becomes a negative-sum game. And developing an AI-powered stock-picking model is not cheap: The average expense ratio of the ETFs in the chart above is 0.66% - versus just 0.03% for the Vanguard Total Stock Market ETF VTI.

Mark Hulbert is a regular contributor to MarketWatch. His Hulbert Ratings tracks investment newsletters that pay a flat fee to be audited. He can be reached at mark@hulbertratings.com.

-Mark Hulbert