NPP builds a Emini system

Discussion in 'Journals' started by no_pm_please, Dec 8, 2006.

  1. bidask

    bidask

    what do you mean when you say "the profit dropoff isn't too bad?"

    don't you want the narrowest parameter fit instead of the widest?



     
    #21     Dec 10, 2006
  2. Have any of you read "Smarter Trading" by Perry Kaufman? It's an old book but I just happen to be reading it now....it has quite a bit about system development that seems relevant here, especially stuff about using small periods of time to develop systems. As an amateur at this stuff I find it very interesting....but it seems similar to the problems you're encountering. Definitely check it out (it's pretty cheap on amazon) and worth reading, even if some of the stuff is kind of dated....
     
    #22     Dec 10, 2006
  3. Optimize from 10 - 99. Then view the optimization. Move over to the average trade column and sort it in descending value. Then go down until the average trade drops below $80. Then move to the start of the row and see what number it is ex. 1 - 90. Then use that number to divide by 90 (number of tests) to come up with the %.
     
    #23     Dec 11, 2006
  4. When you increase the length of time in bars you normally have fewer trades and hence lower profits. I was looking at the sum of the profits to see how much of a drop there was from 15 min. - 30 min. and from 30 min. - 60 min., etc. The 60 min. looked like a good compromise. At 120 min. the total profit available dropped off at a faster rate.

    The parameter I'm using is affected by the length of time in a trade. Ideally all periods would be somewhat profitable with the best happening at some peak with very little dropoff. That's what I meant. A narrow parameter fit would only work for a very specific time period (classic curve fit). All other periods would fail.
     
    #24     Dec 11, 2006
  5. Thanks for the book. I'm not really into systems trading. I use it mainly for trying to understand market concepts. Then I use lots of screen time to come up with my trading approaches. In the end most everything I use can't be converted into code, so I don't try to do automating.
     
    #25     Dec 11, 2006
  6. Ok, I did lots of tests to figure out what I can use to trend trade the emini. In the end I found a combination of using pullbacks in a trend and the addition of locking in larger than average profits worked well enough to say "I'm done". Also, it seems my original profit taking point should have been symmetrical between the longs and shorts.

    Here's the TS summary. I used 84 for the parameter. It's not optimal but left of it. Closer to a performance valley. The optimal parameter optimizing on net profit used 92 (which is consistent with the higher the number the better the profit factor). The % of average trade above $80 went up to 33% so it's not great, but it gives some hope that it will continue to do well (especially since the losing parameters were all on the low end). The valley we found in net profit previously still exists to a small extent, however even in the lowest point it's still profitable.
     
    #26     Dec 11, 2006
  7. amg

    amg Guest

    Hi there-- joining the chorus of "great to see you posting". I too like RSI and once made an Ensign template of one of your RSI ideas (BBands on RSI).

    Curiously, because price has been in chop, I just posted an RSI template that is uses RSI's ranging tendency: when price is between 45 and 55, the probability of chop is the highest. This is from long time observation rather than anything rigorous.

    I suppose this is a way of saying that my tests of 50 crosses have generally been abysmal and a two-fold test of 45 and 55 breakouts might yield more interesting results.

    best, amg
     
    #27     Dec 11, 2006
  8. Here's the code that generated the TS report. I forgot to mention I included $30 for slippage and comm. per round-turn as well. If it ends up being less, then thats great.
     
    #28     Dec 11, 2006
  9. And the last I'll post on this is the annual performance. You can see it's pretty consistent.
     
    #29     Dec 11, 2006
  10. So, I'd go short on a cross below 45 and go long on a cross above 55? Stops on a cross back to avoid a big loss? And use some trailing stop to treat it as a breakout?
     
    #30     Dec 11, 2006