The Artist Who Trained Rats to Trade in Foreign-Exchange Markets

Mr. Lehman’s predictions of currency futures prices were more accurate than his alleged coin toss. But, as a rat, he needs the right rewards to do this: when he’s right, he gets a food pellet; when he’s wrong, he gets a small shock. (And, being a mouse, he’s not very good at flipping coins.)

Lehman is part of the “Rat Traders” project, which is overseen by Michael Marcovici, an Austrian conceptual artist whose work often comments on business and economics. In the project, Markovich trained dozens of rats to detect patterns in the currency futures market. To do this, he translated price fluctuations into a series of notes played on a piano—if the price went up, the next note would be higher—and asked rats to predict the pitch of the following notes. Given some prodding, the mice began to predict price changes, and after several months of training, Markovich said, they outperformed human traders—though that would require more thorough testing than the one done here.

When I exchanged emails with him, Markovich never referred to his mice as part of an art project, and Video interview he gave in 2009As the project drew to a close, he didn’t smile. But his irony is not hard to detect: “The Rat Merchant” Your own mock company websitewhich indicates that the company is Headquartered in the Cayman Islands.

Markovich came up with the idea of ​​training rats to make investment decisions after thinking that humans might not be needed for high-paying jobs in the future. The default assumption was that these jobs would be taken over by robots, but Markovich wondered whether mice would be able to recognize patterns in the data that humans miss because of their messy biases and status issues.

To test this, he developed a semi-scientific training program. For three months, the rats spent five hours a day making predictions in a temperature-controlled box Markovich built for them. Correct choices were rewarded with food, and incorrect choices were punished with a mild electric shock. “Good mice get fat quickly,” Marcovici wrote on his website.

Over the course of several months, he began weeding out the rats whose trading accuracy was less than 52 percent. After experimenting with about 1,000 piano pieces, Markovich was left with four “really gifted traders,” whom he then interbred, producing a generation that performed better than its ancestors. One of the second-generation mice, Mr. Kleinworth Morgan Jr., achieved an accuracy of 57 percent. “I outperformed some of the world’s leading human fund managers,” Markovich wrote. (His findings were not tested for statistical significance, and the rigor of his experimental design was limited at best.)

Accuracy of descendants of Mr Morgan and Ms Kleinworth

exist Interview five years agohe said multiple hedge funds were interested in testing his mice, but that interest was unsuccessful. Markovich now says that even if rats are proven to predict prices, one bottleneck is that rats can only make about 20 trades before they tire, so hedge funds need a lot of rats to accumulate useful data. Still, he maintains “loose ties” with some hedge funds in case they change their minds. Markovich himself retired from the program several years ago. “With about 100 mice at home, I had to stop the experiment at some point,” he said.

“Rat Trader” is based on the assumption, perhaps the universal desire, that historical market data can be used to predict prices. Markovich insisted in an interview that “because people are the ones who influence prices,” there are patterns formed by human biases that can be found in the numbers. There is little literature to support this, but A 2011 study Patterns have indeed been discovered that can be used to predict stock prices for up to a minute. Another recent study Found that stock prices can be predicted by activity on Twitter.

But when it can be mixed and matched Talk to one of the authors of this Twitter studywho said he wasn’t sure why the correlation was found. That’s the point here: It’s clear that the price of any given stock is impossible to predict, and anyone who tries to predict it, without enough luck, is doomed to fail. one far-reaching investigation A survey of 60,000 households conducted by researchers at the University of California, Berkeley, in the early 1990s confirmed this: “Our central message is that trading is bad for your wealth,” they concluded.

It turns out that most forecasting strategies are empty correlations that don’t last. Still, people want to believe in the predictability of the stock market, perhaps as a way to cope with the stress of investing or even the discomfort of a truly chaotic system. So when someone like Markovich comes along, many people are willing to believe that his artwork is a shortcut to personal gain: business insiderno doubt tone Covering “The Rat Merchant” A few weeks ago, prompting the site to post a disclaimer at the top of the post (“Note: This is a joke.”).

Fund manager and author David Leinweber once tried to prove the coincidence of predictions by finding historical data that miraculously aligned with the stock market at that time. Finally, he found a promising indicator: Bangladesh’s cream production could be used to predict changes in the S&P 500 over more than a decade with 75 percent accuracy. When he took into account U.S. cheese production and the number of sheep in the U.S. and Bangladesh, his accuracy shot up to 99% over the same period. After writing about Bengali butter relevance, his book Leinweber Tell wall street journal He received requests from overzealous traders to distribute his material. “The painful thing is that a lot of people don’t understand that this is a joke,” he said.