It just to find them argues Arnaud Vincent

Find a martingale market through collective intelligence on the Internet. It is the dream of Arnaud Vincent. This PhD at the school of mines has created with a computer friend, Christophe Méheut, virtual robots of a new genus. His "Krabott" are born, reproduce and die online, not without having bequeathed their precious information on the Exchange to their descendants.

They are distant cousins of the algorithms used for quantitative hedge funds in the market halls. But, while the latter automatically spend stock exchange orders, it is not yet the case of young people Krabott, with only a few weeks of experience. Other size difference: while current robots are parameterized by human geniuses, they have been programmed to become brilliant... by themselves.

"Each Krabott is unique because it has its own genetic heritage," explains Arnaud Vincent. "DNA sequences" are in fact rules of stock market behavior. They are established by the two creators of the game, which manufacture of tanks of Krabott and then deliver these babies traders to the members of the Community () online. They are set to react for example with variations of moving averages, a collapse in the prices, but with what is said on the Web. "We have set up alerts by which such robot will for example stop positions for ten minutes," details Arnaud Vincent.

Full-scale test

Several scientific studies have shown that the intensity of the exchange of short messages on this communication platform could predict changes in stock market prices. According to researchers at the Indiana University Bloomington in the United States, Twitter makes foreseeable 87 the evolution of the Dow Jones Industrial Average three days later. "It will for example able to monitor the 50 most popular keywords on Twitter, all minutes, and change the position if terms that were never created buzz before arise," explains Arnaud Vincent.

The fact to speculate blind poses no problem for the young PhD, instead. "We could even create an astrological gene," quips that young wise man, who the day worked as Director of innovation in a large French Bank. And for cause, it is a specialist of heuristics, the science of empirical strategies. "It's 12 years I have been working on algorithms of this type." The principle is to put all the rules in bulk, even zany, and allow the algorithm to sort. "In the past, this method allowed him to improve the composition of a shampoo, optimize traffic in a city cadençant red lights and resizing channels, to improve the journeys of elevators in hospitals...

"We live in a chaotic world, but there are paths." "It just to find them", argues Arnaud Vincent. Place players for this full-scale testing. They are now over 120, including 15 real traders. In deciding to match their Krabott with the best trader of the community, they give birth to an improved race. As ineffective robots, on "dismissed them" with a click. "The life of a Krabott is limited, because the stock strategies faltering." "But the population is intended to survive," said Arnaud Vincent. That is why it hopes to create a "méta-algorithme" who will survive the crises, no mathematical model is today understand.