Imagine having the ability to predict the next day's market trends and trade ahead of the rest of the market. Sounds enticing, but a new study by Elm Wealth shows just how hard it is to avoid ruin and turn a profit.
The study, led by Jerry Bell, Victor Haghani, and James White, simulated a scenario where both humans and leading AI models had access to the next day's news in advance and could place their bets before it broke. Despite this advantage, it was difficult for both humans and machines to avoid ruin and turn a profit.
The study found that both humans and AI models were bad at sizing bets, with the AI models taking too much risk and the humans taking even more. The AI models correctly predicted the direction of stocks and bonds only about 60% of the time, yet applied an average leverage of 7-12 times to their bets.
The humans were even worse, with players in aggregate betting no more heavily when the news made price moves easy to predict. They also took too much risk overall, with 30% of days featuring leverage above 20 times.
However, the study did find that expert macro traders were able to perform better, with an average return of 130%. They varied their position sizes a lot, betting more when they felt confident and not at all when they did not.
