Arthur Deco Mentors ET Noobs

Discussion in 'Psychology' started by Arthur Deco, Jan 30, 2011.

  1. I am trying to get them to pay me not to say, or at least to recommend the competition. I have had a love/hate relationship with ESignal for ten years. They are as good a service as I am a trader.
     
    #201     Feb 2, 2011
  2. Arthur Deco you may of said before but what software do you use to backtest?
     
    #202     Feb 2, 2011
  3. Thanks for the previous responses Mr Deco, I've found them insightful.
    Happy to be of service. I hope I've worked my way up from noob to boob -- but that's ok, I like boobs!!
    There it is -- that's the one step I haven't done. Recently I've run a backtest and optimized, then compared the performance against an equal period in the past, and they're never consistent. I'll go in the opposite direction -- get some old systems from last year from the archives, check them for consistency to the present, but probably not likely.

    Is there an optimal ratio of how many bars should go into the signal for how many bars away it will probably take to hit the limit? Some of my first system attempts, the signals rarely worked well going forward. Then it occurred to me, it was unreasonable to expect a signal produced in 2-3 bars from some indicator values should target a limit of an amount that would probably take 50 bars for the instrument to travel that far. Now I'm looking for a new approach.
     
    #203     Feb 2, 2011
  4. pspr

    pspr

    Hey, Art. I just thought I would stick my head up here to get out of the Politics & Religion gutter forum for a while. I was getting a headache, I need a little fresh air.

    You're really giving the noobs some good info. This is almost like an "Ask God A Question" thread! Just don't give them the Holy Grail for at least a few weeks.
     
    #204     Feb 2, 2011
  5. Sorry you didn't see the answer: EasySignal.
     
    #205     Feb 2, 2011
  6. I am sorry, I didn't understand what you meant. Happy hour started early today. Are you talking about how far back to back test?
     
    #206     Feb 2, 2011
  7. Nah, they're not from here! Non-Texans need not apply!

    I think rather than ask God it is more like ask the oracle. Oracles were always stoned.
     
    #207     Feb 2, 2011
  8. Sorry for interfering with Happy Hour, I'll give an example and with easy numbers.
    Suppose an instrument with its volatility usually moves about $1/bar. Take a Moving Average crossover signal, with fast n = 5, and slow n = 10, I suppose that signal would incorporate 10 bars. Could such a signal be expected to target a limit of $50, since the instrument would usually take about 50 bars to move $50 up or down? Or, should the target amount be closer to $10 to match the number of bars that went into the signal? Or, is this consideration not meaningful and irrelevant?
     
    #208     Feb 2, 2011
  9. Thanks for explaining. Let me restate the question this way: if a setup takes N bars to become manifest out of market noise, how many bars should we expect the trade to last? Correct me if I am wrong, as English is not my native language.

    That is an interesting way to think about setups. Rather than thinking about the profit potential, that is thinking about the duration of the validity of the signal, without reference to the signal being invalidated, as by the opposite crossover.

    It seems like a valid hypothesis for testing. Instead of testing for the optimum profit, test for the optimum time in the trade. I have tried something like this as an exit criterion where I have already optimized for profit. In other words, if my profit target has not been achieved, should I bail after a certain time span? Turns out for my systems that it didn't work.

    Bottom line is that if you aren't already doing so, you should think about learning to code and test. Programs like ESignal, much as I loathe the ugly bitch, make it relatively simple now with autocoding wizards to pick this skill up quickly. Then once you have their autocoded examples in hand, you can launch off on your own to do more unconventional things. The simple example you gave could be autocoded inEasySignal licka-da-split.
     
    #209     Feb 2, 2011
  10. Thanks for the response Mr Deco, sounds like you have the idea I was going for.
    I agree about coding, the best way to test ideas. For anyone who doesn't, I'd recommend it too. Start by spending just 2 weeks to get introduced to it, from there as you work your way through slightly larger projects it comes quickly.
     
    #210     Feb 2, 2011