Long only 15 min DAX

Forums ProRealTime English forum ProOrder support Long only 15 min DAX

Viewing 13 posts - 16 through 28 (of 28 total)
  • #28890

    Yes I did. The drawdown I have now is coming from when the price is consolidation around the moving average. Hopefully the system is better on a longer history now.

    #28901

    Yes, it’s a better now. 200k bars test attached.

    #28914

    Nice. Still not super good until the more volatile market starts in 2015. I added one more short term moving average (100 periods). It didn’t improve the return on the 2 year timeframe I have but it did lower the draw down which is visible on the gain/loss ratio. Backtest with 2p spread.

     

    #28919

    Better is worse than good. Too much filtering reduce trades (almost 30% less for a limited increase in gain in my 200k bars test) and also statistics about robustness. It’s always difficult to be sure of what is a good filter or not. That’s why you should only make backtest with In Sample and Out Of Sample period, like I explain in this blog article: https://www.prorealcode.com/blog/avoid-equity-curve-fitting-with-probacktest-trading-strategy-optimisation/

    (it talks about optimisation, but the idea is the same for a normal backtest).

    Try to improve the strategy on IS period with as much ideas you can imagine. Then test the strategy on OOS period to see if it has improved the strategy result in the “future”, like if you were trading the strategy in real time. If the answer is “yes”, then your new ideas may be considered to be a good ones.

    #28922

    Fair point! I have to start doing the OOS test you mention in your blog article. Very well written btw! On that point, do you know why the walk forward settings are not visible on my platform (v10.3)? (I have attached a picture of my screen).

    #28934

    Walk Forward will be available to IG customers within the next weeks. It’s only available now for prorealtime trading customers.

    #29001

    @ nicolas , could you post here the backtest on 200 k bars please .

    #29158

    This is the latest update I have on the code. Now without the longer moving average which excluded a large amount of trades. All values have been selected by the optimisation function in PRT, apart from the stop loss which was suggested to be 1.05. In my opinon were the increase of risk greater than the increase of return. I have not yet tested on an OOS period.

     

    #29166

    You should have optimised on 70% of history (IS) and test if it was correct on the last 30% by yourself (OOS) 🙂

    #29167

    haha, why didn’t I think of that myself.. thanks a lot Nicolas!

    #29220
    #29222

    just code your code with accumalation , here it is the backtest with 1 € per contract

    #30041

    I wouldn’t recommend this. Better to have several systems which you size up simultaneously.

    1 user thanked author for this post.
Viewing 13 posts - 16 through 28 (of 28 total)

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