A rat named Mr. Lehman accurately predicts the price of a
crude-oil future. (Nora Friedel/Rattraders.com)
Mr. Lehman could predict the prices of
foreign-exchange futures more accurately than he could call a coin flip.
But, being a rat, he needed the right bonus package to do so: a food
pellet for when he was right, and a small shock when he was wrong.
(Also, being a rat, he was not very good at flipping coins.)
Mr. Lehman was part of “Rat Traders,” a project
overseen by the Austrian conceptual artist Michael Marcovici, whose work
often comments on business and the economy. For the project, Marcovici
trained dozens of rats to detect patterns in the foreign-exchange
futures market. To do this, he converted price fluctuations into a
series of notes played on a piano—if a price went up, the next note was
higher—and then left it up to the rat to predict the tone of the note
that followed. With some prodding, the rats began forecasting
price changes, and Marcovici says that they were outperforming human
traders after a few months of training—a claim, though, that’d require
testing more thorough than what was done here.
As I exchanged emails with him, Marcovici never referred to his rats as part of an art project, and in a video interview he gave back in 2009,
when the project was nearing its end, he never so much as displayed a
sliver of a smile. But his satire is not hard to detect: “Rat Traders”
has its own mock-corporate website, where it says that the company is headquartered in the Cayman Islands.
Marcovici got the
idea of training rats to make investment decisions after thinking about
the highly-paid jobs that might not need humans in the future. The
default assumption is that these jobs will be taken over by robots, but
Marcovici wondered if rats might be able recognize patterns in the data
that humans, with their messy biases and status concerns, overlook.
To test this, he set up a semi-scientific
training program. Rats spent as much as five hours a day for three
months making predictions in the temperature-controlled boxes Marcovici
built for them. Correct picks were rewarded with food, and incorrect
picks were punished with minor shocks. “The good rats became fat very
fast,” Marcovici wrote on his website.
Over the course of months, he began weeding out
the rats that traded at less than 52 percent accuracy. After trials
with about 1,000 piano tracks, Marcovici was left with four “really
gifted traders,” which he then cross-bred to produce a generation that
outperformed their progenitors. One rat in this second generation, Mr.
Kleinworth Morgan Jr., had a 57 percent accuracy rate. “I managed to
outperform some of the world’s leading human fund managers,” Marcovici
wrote. (His findings did not undergo tests for statistical significance, and the rigor of his experimental design is limited at best.)
Accuracy Rates for the Offspring of Mr. Morgan and Ms. Kleinworth
In an interview five years ago,
he said that multiple hedge funds were interested in testing his rats,
but that interest didn't pan out. Even if it were verified that a rat
could predict prices, Marcovici says now, one bottleneck is that a rat
can only make about 20 trades before getting tired—so hedge funds would
need a lot of rats to accumulate any useful amount of data. That said,
he’s still in “loose contact” with some hedge funds, should they change
their minds. Marcovici himself retired the project years ago. “With
about 100 rats at home I had to stop at some point with the experiment,”
he says.
http://www.theatlantic.com/business/archive/2014/10/the-artist-who-trained-rats-to-trade-in-foreign-exchange-markets/381456/
<a href="http://www.backlinktool.biz/">Backlink Tool</a>
<a href="http://www.backlinktool.biz/">Backlink Tool</a>
Post a Comment