Machine designed strategies. Do they work?

Discussion in 'Strategy Building' started by v75z52, Mar 20, 2012.

  1. EZQuote does forex in addition to stocks. I will be testing that next week.
     
    #121     Apr 8, 2012
  2. tim888

    tim888

    Pls let us know your findings.
     
    #122     Apr 9, 2012
  3. I liked it very much. It has many conversion options listed you can easily pick and a nice scheduler.

    By the way PAL has published a step-by-step procedure and results of a machine generated system for SPY in their blog. It includes in-sample and out-of- sample tests.
     
    #123     Apr 11, 2012
  4. The only out of sample test that counts, is the one where the vendor stops marketing, and starts trading.

    No one is so stupid as to sell software for a living, when they could be trading instead.

    Economic reality is a bitch.
     
    #124     Apr 12, 2012
  5. There are as many economic realities as possible parallel words. You maybe sound too dogmatic after adopting just one of them.

    For example, if the income received from software sales is comparable or exceeds income from trading due to capitalization and risk aversion constraints and at the same time if the vendor can use its software with different functions no publicly disclosed, then it is economically rational to sell it. Selling the software is like buying a put option and trading it is like buying a call option, or whatever...

    If you are a young man you must abandon dogmatic views and move ahead by broadening your horizons...there is not a single rational economic reality...
     
    #125     Apr 12, 2012
  6. Time is finite.

    Allocating time to the most productive endeavour is rational.

    If you're failing as a trader, then sure, go sell software..

    Ask an entrepreneur about how they allocate their time.
     
    #126     Apr 12, 2012
  7. You can't speed up trading above certain point. Unless you believe in 10000% a year...
     
    #127     Apr 12, 2012
  8. I think the reasoning behind selling these machine learning kits is that they are able to create a very wide array of results, since they originate from a random and narrow down to something meaningful with the search process. This means that even if the seller of the software is using it for trading, it would take a very long time for his buyers to hit upon the same exacty combinations that are making him money.
     
    #128     Apr 12, 2012
  9. I would agree that some failed traders sell software but logic dictates that not all software is sold by failed traders. As I said, if the rate of growth of profits due to software sales exceeds the rate of growth of profits due to trading because of risk aversion and capitalization constraints then the decision to sell software is rational. Besides, no such software is likely togenerate a complete system for you and no system will impose on you prudent risk management and as a result, most customers of those software will blow up and blame the developers rather than themselves.

    Anyway, this is interesting discussion. Here is another system PAL posted for QQQ. The equity curve for the out-of-sample is a bit volatile but that may have to do with the followed position sizing rule, which basically equates risk percent with the stop-loss percent value according to my calculations.
     
    #129     Apr 13, 2012
  10. My contention is that no software is sold by successful traders.
     
    #130     Apr 13, 2012