Strategy MAEX – experiment with MFE & MAE

Forums › ProRealTime English forum › ProOrder support › Strategy MAEX – experiment with MFE & MAE

Tagged: ,

Viewing 15 posts - 31 through 45 (of 83 total)
1 2 3 4 5 6
  • 11/17/2020 at 10:54 AM #150764
    zilliq

    Hi @Paul

    Thanks for your work !

    Instead of MFE and MAE on which I work a lot (And don’t use them actually), why don’t you simply use 2 variable a and b with set targt pprofit a and set stop ploss b or with trailing stop ?

    Because the problem with MAE and MFE is that some winning trades with MAE can be cut too soon and some other winning trades can be cut too soon too ? Hope I’m clear

    With these 2  “simple” variables the system calculate the best compromise

    Have a nice day Paul

     

    11/17/2020 at 11:31 AM #150766
    nonetheless

    That is splendido, MM works perfick now, new version attached.

    I revised the stops and targets as it seemed they might be over-optimized for such a short backtest, still does well with more generalized values.

    Could be a winner! 😁

    6 users thanked author for this post.
    avatar ayrton78, avatar GraHal, avatar eckaw, avatar Kovit, avatar Paul, avatar josef1604
    11/17/2020 at 11:45 AM #150768
    snucke

    well done!

    how much is this optimized nonetheless? 70/30?

    11/17/2020 at 11:45 AM #150769
    nonetheless

    @ullle73, this also fixes the .2 problem

    1 user thanked author for this post.
    avatar ullle73
    11/17/2020 at 11:49 AM #150770
    nonetheless
    snucke wrote:

    how much is this optimized

    The indicators I added are optimized at 70/30, stops and targets were done with 100% of data

    11/17/2020 at 4:53 PM #150801
    zilliq

    Five minutes to explain what I said before, on a picture, and why for me it’s better to use variables a and b, or sometimes only a

    When we use MAE (same thing with MFE), as you see on the picture if we put a cutoff at -1500 USD for example, we lose all winning trades in the blue rectangle and we still have losing trades in the yellow rectangle

    And if we use a cutoff at -3000 USD you have effectively all winning trades, but all losing trades too

    That’s why, for me, it’s better to use a variable, to do a backtest, and choose the best compromise

    MAE/MFE works ONLY when it can separate quite all winning/losing trades who is very very rare

    And finally, the more complex is the code, the more curvefitting we will have

    Have a nice day @Paul

    11/17/2020 at 5:55 PM #150806
    Paul

    Hi @Zilliq,  So you also work the other way around? Doing a backtest without or large stoploss then determine the cutoff /compromise point visually and you use that as stoploss?  This with limited variables.

    It’s need to sink in a bit more & I will do some experiments with less variables.

    Thnx for your explanation & have a nice evening!

     

     

     

    11/17/2020 at 6:51 PM #150814
    zilliq

    No @Paul I use one variable to determine the better compromise in the backtest as you can see on the picture (Stop at 26, 28, 30, pip and so on..)

    The backtest determine the better stop not me and not visual

    Never more than 3 variables in all the algo and the more simple it is the less curvefit we will have

    We need always remember that it is like a polynomial regression on gain (More variable=more curvefit=bad OOS results on mid/long term and more often on small term)

    Have a nice evening Paul

    1 user thanked author for this post.
    avatar Paul
    01/16/2021 at 11:39 AM #158043
    nonetheless

    Hi Paul, here’s a v11 update – optimized 75/25 so about 8 months OOS. I tried both ATR and % TS (4MM pc), not a huge difference but I think the ATR gives a better curve. VRT is uncommonly good, so that’s encouraging.

    Best regards and thanks again for sharing this.

    4 users thanked author for this post.
    avatar Midlanddave, avatar Paul, avatar Monobrow, avatar GraHal
    01/16/2021 at 3:19 PM #158077
    Paul

    Hi nonetheless, great work, it looks really good with lots of trades!

    Only thing I see, don’t know what difference it would make yet, you defined n (5) as parameter for the supertrend, but n is also used a bit below and assigned to value 4 in the mfe&mae section.

    It’s no big deal, the performance is as it is but maybe there’s a bit more room to optimise.

    Thnx for putting it online & investing time in this concept!

     

    01/16/2021 at 5:11 PM #158088
    nonetheless
    Paul wrote:

    you defined n (5) as parameter for the supertrend, but n is also used a bit below

    Hey, well spotted – explains why I got a different result when I tried to clear the opt box and insert the values in the code. It was bending my brain trying to find the discrepancy so I just left it. I don’t normally post the itf like that but if I hadn’t I’d have never found out.

    So here’s the correct version — Supertrend (2.5,6) — gives a nice 5% bump to overall performance.

    5 users thanked author for this post.
    avatar Ryugin, avatar Roger, avatar Paul, avatar Monobrow, avatar GraHal
    01/17/2021 at 4:49 PM #158177
    Tanou

    I just arrived here I don’t know how bu I’d like to say congrtas to you guys! This strat looks very promising 🙂

    01/17/2021 at 5:12 PM #158178
    OboeOpt

    I´m on UTC+01.00 Sweden

    I´ve changed “closetime = 235500” (was 225500, row 55)

    But what should I have here below? Don´t understand the “0” (row 59)

    ctime=time>=0 and time<230000//UK time

     

    01/17/2021 at 7:46 PM #158187
    nonetheless

    Hi oboe, this is how I would check for the best Ctime settings:

    then enter a range of times, for example 0 to 150000 for t1 and 180000 to 230000 for t2

    both with steps of 10000

     

    1 user thanked author for this post.
    avatar OboeOpt
    01/17/2021 at 11:49 PM #158214
    OboeOpt

    Thanks nonetheless!

    Yes I´m familiar to the way to optimize. I just want to use the same settings as you to replicate and look at your result. If “0” is the same as “000000” I guess it will be
    ctime=time>=010000 and time<000000 in my timezone.

     

  • Author
    Posts
Viewing 15 posts - 31 through 45 (of 83 total)
1 2 3 4 5 6

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
average mfe/mae snippet

Create your free account now and post your request to benefit from the help of the community
Register or Login