NASDAQ Mean Reversion

Forums ProRealTime English forum ProOrder support NASDAQ Mean Reversion

Viewing 15 posts - 1 through 15 (of 38 total)
  • #176000

    Hello,

    A work in progress strategy based on a simple mean reversion idea inspired by Kevin Davey that I have adapted from a daily to a 15 minute timeframe.  It looks for low volume and low price over N bars, then goes long.  Only two entry criteria and one initial filter, and works well but needs improving.

    The focus at this stage is to work on reducing the Drawdown, and potentially the number of trades, by use of an entry filter.  Thought I would share this early version in the hope that people might have some useful ideas for a suitable filter.

    NASDAQ

    15m

    Spread 2

     

    Thank you,

    S

    7 users thanked author for this post.
    #176003

    With screenshot that was dropped

    #176035

    Really impressive, thanks for sharing! 🙂

    #176043

    Wow. Impressing. No short entrys?

    #176048

    One thing you might want to look at is the section starting at line 73 above. In a 10 year backtest no trades have gone to 1600 bars so those values are doing nothing. This is also the case for the Breakout code you posted.

    This is the snippet I use for that function, closes very long-running positions whether they’re winning or losing. You have to optimise for b1 and b2 (b3, b4 for short).

    1 user thanked author for this post.
    #176051

    whats the different between B1 and B2 flr longmarket?

    #176052

    b1 if it’s winning, b2 if it’s losing.

    but I should add that this isn’t always advantageous, you should also try switching it off, with some algos it’s better to just let them run.

    1 user thanked author for this post.
    #176053

    B1 is for pp > 0 and B2 is for pp < 0.

    I have successfully used above on several of my Algos from my early coding days, I will get back on it again, now I see you’re using it Nonetheless! 🙂

    I also optimised using broad values for positionperf (in lieu  of 0 as above).

    2 users thanked author for this post.
    #176054

    but I should add that this isn’t always advantageous, you should also try switching it off, with some algos it’s better to just let them run.

    Without this remark I already wanted to add :

    I have attempted so many means of exiting because of too long “useless” trades. They always and always net lose. Otherwise this will be highly subject to curve fitting, IMO.

    BUT

    Only last week I applied a “strange” means for this : let shrink the SL per bar (or once in the so many, also OK). This seems counterproductive, but the contrary appears true. This starts with allowing for a higher SL (avoid peaks at entry) which is psychologically good. Next the losses because or running into the SL in practice are not that high at all. But moreover, this works out. For me it does …
    Warning : I am almost sure that we should not change the StopLoss per bar because the broker may not like that. Eh, *if* possible at all. So it needs to be your own exit (like the example code in the earlier post) and NOT changing the StopLoss each time.

    #176055

    Visualisation of what I was saying : when blue and yellow meet, it’s over.

    1 user thanked author for this post.
    #176068

    One thing you might want to look at is the section starting at line 73 above. In a 10 year backtest no trades have gone to 1600 bars so those values are doing nothing. This is also the case for the Breakout code you posted.

    This is the snippet I use for that function, closes very long-running positions whether they’re winning or losing. You have to optimise for b1 and b2 (b3, b4 for short).

    Thank you very much, I shall add this in and retest.  Much appreciated.

    1 user thanked author for this post.
    #176069

    I also optimised using broad values for positionperf (in lieu of 0 as above).

    Would u like to show an example

    #176074

    I played a bit with the strategy, on DAX 1€, 200K units, 5 min-TF (mtf), changing just TP & SL and slightly the Trailing Stop to use PointsToKeep as sort of a trailing step. Performance is nice and WF is around 25%. I also commented out exiting on RSI:

     

    2 users thanked author for this post.
    #176077

    @samsampop, what is the result about kevin’s initial strategy on daily timeframe ? thx

    #176078

    Hello MAKSIDE

    I put the source for the original link at the top of the code, it’s pretty decent but I tend not to trade algos on a daily basis, only US OTC stocks.

    Source:  https://www.youtube.com/watch?v=D_P_XqB5nHs

    Thanks

    1 user thanked author for this post.
Viewing 15 posts - 1 through 15 (of 38 total)

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