I am not sure how seriously to take the above paper found on arxiv.org. This is its conclusion:
By analogy with the way motor enzymes trap favourable brownian fluctuations, we have built an algorithm which is able to make the best from out of equilibrium price fluctuations and to play the market. Testing its efficiency with genuine historical data, positive cumulative returns have been measured even in presence of a 0:1% transaction cost. Especially stupendous are the results dealing with the application of the algorithm to EMS currencies or with the Cac40 components.The results of using the algorithm with Cac40:
Fig.(14) displays the stupendous results of the application of the MD3 algorithm to the components of the Cac40 between 2000/01/01 and 2006/05/12 ( 6:5 years). First, the optimal value of m does not depend on the time interval (out of sample). Second this optimal value is found to be close to 25 days, so to say one month since only workdays are taken into account. Finally but not the least, average yearly return up to 60% are obtained!There is a catch, however:
The money which is captured by our algorithms comes from the irrational behavior of uninformed noisy traders. Therefore we really expect the present algorithms will become unprofitable as soon as our paper will be published, either because irrational traders will be taught a lesson or because the profitability of the algorithms will vanish with the number of users.The thing is, this paper has only just been published, and maybe there aren't that many people who sit at home on a Saturday night reading arxiv.org papers. If I have alerted any market player to a way to make a killing in a short time, please sent me a cut of your profit!