Hi I'm new to trading, just bought my first stock a little over a month ago. I decided to create a program to tell me which stocks to purchase. I'm documenting the process on an old blog I've resurrected: http://wharfworks.wordpress.com/category/stock-market-propagation/ I'm hoping to learn what software others might be using, identify any pitfalls in my method, improve it, and hopefully just become less of a stock newb. What I've done so far is use some Fourier Transforms and matrix regressions to propagate stocks, and then used a genetic optimization to improve the propagation variables and a few buy/sell criteria variables. There are some obvious flaws (using past behavior to predict the future and over specialization to name a couple), but I'm actively improving the algorithms. Anyway I'm looking forward to learning and contributing to this forum! Byron
This is a great article, it's true that I'm potentially not conforming to his recommended criteria for an un-biased system, although I did recognize that I have an over-fitting problem. I've also not implemented some of his recommended methods for testing if a method isn't being biased (a random stock data generator is a great idea). I'll have to implement some of his ideas and see if my algorithm can be improved. I'm trying a few different methods to avoid over-fitting as his article does suggest. One of the attempts is here: http://wharfworks.wordpress.com/2014/02/22/avoid-over-specialization-attempt/ The simulation in the figure, however, does still have some over-fitting behavior that I want to get rid of. I think implementing a few of Michael Harris' ideas might help. Thanks for the help!