Machine designed strategies. Do they work?

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

  1. Well, luck follows the brave as they say. And it's pretty brave to use tech like TSL. I think milewski is probably well aware of the amount of previous "hyped up scams" that have existed on the markets and fooled people into wasting money for the past 100 years. Maybe if TSL isn't like that, and if it's truly innovative, well then he's way ahead of the rest of you. Considering he already profited from it, i'd say there's a pretty good chance he is! With all my studies on the subject, by now we all know my opinion. It's great to hear from a TSL user so that i can finalize that opinion.
     
    #81     Mar 30, 2012
  2. tim888

    tim888

    Here is another example for XLF, which is an ETF that due to its price behavior I had trouble finding a good strategy for it. The PAL results are for 2% profit target and stop-loss, win rate > 63%, pf> 1.3 and trades > 27 in the in sample from 1998 to end of 2008:

    [​IMG]


    In this case there are 5 long and 8 short patterns for those that complained before that the mix was not good. Naturally, they are all profitable in the in sample.
     
    #82     Mar 30, 2012
  3. tim888

    tim888

    Here are the OOS results:

    [​IMG]

    Only 3 patterns have pf < 1 and the overall pf I calculated comes out to 1.74 on a sample of 167 trades (OOS).

    I recommend to users of other programs to post results that are verifiable with a demo version. Talk is cheap. Anyone can verify the PAL results I posted with a demo version only, he just won't be able to generate code. Search, backtesting, etc. are fully operational in the demo version.

    I went with PAL because I could see what results I should expect while trying the demo version and I thought that was very honest from the part of the company.
     
    #83     Mar 30, 2012
  4. here is a example chart for you guys

    [​IMG]

    its on BO:

    PF 1.72
    % profitable 54
    avg month return 1200
    std deviation of monthly return 1500
    largest DD 5300$ (10/25/10)

    a total of 7010 trades with 10 outlying

    all market orders. only trading during market hours so will be some slippage but BO is tight spread so will be very minimum
     
    #84     Mar 30, 2012
  5. kauflaune

    kauflaune

    Just to clear up, BO is soybean oil? You're obviously using the pit traded contract starting 1972. Is the pit contract identical to Globex? I mean do you get the same backtest in recent years? What's your average trade in ticks?
     
    #85     Mar 30, 2012
  6. Yes Bean Oil

    use the pit until the early 2000's then stitch together globex for last 10 years or so.

     
    #86     Mar 30, 2012
  7. kauflaune

    kauflaune

    Yeah I remember that's a TS feature, the composite symbol BO.C which is pit and electronic copied together. With an average trade of around $60, your system seems tradable, considering the small BO tick size of $6. Not bad, I never thought of trading the more exotic markets.
     
    #87     Mar 30, 2012
  8. tim888

    tim888

    The problem is I cannot verify your results. They may be actual but science says that results that cannot be verified mean little.

    If TSL provided a demo that everyone here - not just those who have the program - could use it to verify the equity curve then your results would make sense.

    Why don't you use as an example a more popular contract like ES?
     
    #88     Mar 31, 2012
  9. IMO the OOS is too short compared to the IS. If this system is curve-fitted that amy not show in such a short in proportion OOS. You have only 6 years for the OOS in a total of 40 years. Most texts recommend 70%:30% proportion. That should bring you back to maybe 2001 for the start of the OOS.
     
    #89     Mar 31, 2012
  10. IMO i would never trade a system that's already 30% OOS but just to prove a point here. ill do 60% OOS 40% IS.

    [​IMG]

     
    • bo2.jpg
      File size:
      68.9 KB
      Views:
      684
    #90     Mar 31, 2012