If 79% of Experts Can’t Beat Index-Based Trading, Why Would AI Be Any Different?

The news that AI agents will soon be able to match human stock traders shouldn’t surprise anyone. Vlad Tenev, CEO of financial services platform Robinhood, recently spelled out how any capacity available to a human will soon be available to an AI agent, and that much of institutional trading has long been automated. The novelty is not that the machines enter the markets, it’s that the small investor can access a packaged version of tools that for decades were reserved only for those who could afford computers, models, data and impossible latencies.

Let’s dismantle a few myths first. The typical image of the stock trader at their desk, surrounded by screens, adrenaline pumping, making calls, using complicated jargon owes much to Hollywood. Traders supposedly offered something mere mortals couldn’t see. The problem is that, when that promise is subjected to empirical verification, it quickly unravels.

Instead, a better metaphor is that of the blindfolded monkey picking stocks by throwing darts at a wallchart. The Wall Street Journal’s Investment Dartboard contest, which ran from 1988 to 2002, demonstrated that random stock selections often performed as well as portfolios chosen by professional analysts who charged so much to produce so little differential value.

Eugene Fama, who won a Nobel Prize in Economics in 2013 along with Lars Peter Hansen and Robert Shiller, for the empirical analysis of asset prices, formulated it from the hypothesis of efficient markets: if prices reflect the available information, no one can consistently beat a simple index. Naturally, the markets are not perfect. They have bubbles, panics, absurd narratives, asymmetric information, biases, manipulation, fashions and collective errors. That is precisely why they are fascinating. But it does not follow that there is a caste of “good” traders capable of exploiting them all the time. It follows that the market is a complex system in which distinguishing talent from luck is extraordinarily difficult.

The accumulated evidence is devastating. The S&P Dow Jones SPIVA scorecard shows time and time again that most active managers fall short of their benchmarks, even before talking about the problem of identifying in advance the few who will do well. In 2025, 79% of active U.S. large-cap funds did not outperform the S&P 500. If an entire industry, with thousands of analysts, terminals, models, and investment committees, fails like this on a recurring basis, perhaps the problem is not the lack of individual talent, but the premise itself. When your bank calls you to offer you some shares or funds, what it is really saying is “let me get you a few commissions so I can keep this beach bar open and continue paying my employees’ salaries at the end of the month”.

AI, therefore, is not going to kill the trader because it is a useless profession, but because much of what it did was automatable, repetitive, emotionally contaminated and statistically indefensible. AI will read more documents, react earlier, detect correlations faster, and not fall in love with a thesis because it was defended in a meeting. But that doesn’t mean it’s going to make the markets perfect. On the contrary: as the IMF warned in its Global Financial Stability Report, AI can increase opacity, concentration, speed of reactions, and the risk of algorithmic herd mentality. Replacing unpredictable humans with machines trained on similar data doesn’t eliminate systemic risk: it can make it less visible and faster.

That’s why my conclusion as an investor is deeply boring, and that’s precisely why it works. Many years ago I decided to put my money in Indexa Capital, a Spanish company that invests in diversified portfolios of index funds and that has built its proposal on something as apparently unglamorous as reducing commissions to the bare minimum. It doesn’t try to sell me clairvoyance, privileged calls or the fantasy that someone knows which stock will go up tomorrow. It offers me diversified exposure, discipline, low costs and a fundamental idea: in investing, many times, winning is simply about not paying someone to lose for you.

The trader’s replacement by AI will be presented as a revolution. In reality, it will be a ceremony of sincerity. For decades we called talent what was often random, we confused movement with value, and we accepted that some intermediaries appropriated a substantial part of the profitability in exchange for a statistically flimsy promise. The algorithm won’t do magic. It will simply reveal that, in the markets, the most profitable intelligence was almost never to trade more often, but to humbly assume that as one does not know more than the market, the only real option is to diversify, reduce costs and play the long game.

(En español, aquí)

—

This post was previously published on MEDIUM.COM.

—