direct statistical trading a "clearly" defined approach by NTW31

Discussion in 'Strategy Building' started by nukethewhales31, Dec 31, 2008.

  1. Dr Who

    Dr Who

    OK, since I made my last post, I've been trading this on the GBP/USD, Monday morning UK time.

    I've been stopped out a few times and the more I think about it the more I realise that, although the maths can work out correctly, you can lose continuously because of the timing of the market.

    For instance, the new hour opens 1.6160 and you short at 1.6151 with your TP at 1.6143. The market falls to 1.6151 bringing you in to the market. It falls to 1.6149, then rises to 1.6167, stopping you out at a loss. The market then falls to 1.6140 and closes at 1.6155.
    Everything has worked out well mathematically but the timing hasn't.

    This scenario has played out a few times this morning, so I'm standing back and taking a closer look at the logic because it seems that its not just the distance a price moves from the open but also the timing.

    I'm not the sharpest knife in the drawer, so perhaps someone can tell me what's wrong with my thinking here.
     
    #461     Sep 21, 2009
  2. in the thread I talk about the problems with this and the resolution.. if you want to just trade bars i wouldnt recommend hourly bars i would recommend something like 3 to 4 hour daily bars.. take trades indirection of IRD and also when you measure the data measure as its building price from zero moves 5 then 5 more before it retreats to 0.. so you know if you place a order at 5 itll go 5 and you can stop at 0... get it?
     
    #462     Sep 21, 2009
  3. Dr Who

    Dr Who

    Thanks for the advice. I'll give it a try that way....
     
    #463     Sep 21, 2009
  4. Nuke i am still a bit lost about how to calculate accuracy level,
    could you please be so kind and explaine one more time using
    a sample you have used earlyer in the tread

    LSD= 7 DFO
    SLD = 14 DFO
    n = 21 bars

    entrance= 8 DFO
    stop = open
    take prof = 14 DFO
    MDRR= 4:3
    %tile accurate = 19/21 or 90.48%
    POT= N trades 114
    LOT= N trades 16
    NPOT=N trades 98

    LSD= longest shortest distance
    SLD=Shortest Longest distance
    N= period of calc bars
    MDRR= market defined risk reward
    POT=Profit over term
    LOT= loss over term
    NPOT= Net POT

    next bars trade
    open +8 entrance
    open + 14 TP
    open stop
    expected success rate = 90.48%


    what is the calculation made, to arrieve at that 90,48% outcome
    for the succes rate / accuracy level ?
     
    #464     Sep 28, 2009
  5. I have re-read the first 35 pages and hope i got it now.

    am i correct to say that before every trade you will take your current LSD and SLD and virtually go over the last N bars to calculate the % of succes rate for those N bars and then if that succes rate ( which is called in this tread EXPECTED SUCCES RATE ) is higher than your chosen ACCURACY and NPOT is high enough and MDDR fits your Risk, ONLY then you will trade the next candle ?

    so in other words you will take current stats and compare them to last N (21) Bars and only take the trade if those current stats fits those old N bars stats enough to give a virtual, statistical and "historical" profit ?

    is this correct?
     
    #465     Oct 13, 2009
  6. i was wrong,
    got it now.
    Cheers
     
    #466     Oct 13, 2009
  7. how is it working out for you flying?
     
    #467     Oct 15, 2009
  8. Hey nukes, have you been able to find tradable edges using tick data with the original concepts in the thread? I've written a program that implements your original LSD/SLD algo on OHLC data and it seems to work pretty well. Later I wrote another program that crunches tick data to do the same statistical analysis. Problem was it didn't have an edge, basically just a coin toss.

    It seems strange that there's an edge with OHLC data but not ticks.
     
    #468     Oct 16, 2009
  9. Thats interesting Steve.

    It would indicate to me that the bars are being watched by enough decision makers to generate said edge.
     
    #469     Oct 16, 2009
  10. Hi Nuke:

    Long time, no hear.

    Are you back in the saddle?

    I still have that code I was working on.
     
    #470     Oct 16, 2009