US100 mini – Overnight Swing Strategy – AI
Forums › ProRealTime English forum › ProOrder support › US100 mini – Overnight Swing Strategy – AI
- This topic has 26 replies, 6 voices, and was last updated 10 months ago by PeterSt.
-
-
11/20/2023 at 9:43 AM #223499
Dear ProRealCode members,
Background:
I am 54 years of age, have worked and obtained experience in the traditional- and alternative investment space and was until mid-2019 part of a multi national FinTech business. During the COVID lockdowns in 2020, i took the opportunity to study 47 trading books and during the last 3 years I was fortunate to implement all my work experiences and skills into developing, on a full-time basis, proprietary ProRealTime code which includes indicators and strategies.
( https://za.linkedin.com/in/niel-rossouw-055b9510 )
High-level Code Description:
I have written 1500 line code, inclusive of indicator count, that incorporates 8 different strategies ranging from deep value-, mean-reversion-, reversal- and micro-trend following strategies.
As a vivid supporter of the late Dr. Van K. Tharp, a well-known US trading coach, he made it clear in his book “Super Trader” that to quote him, “position sizing is probably responsible for 90% of your performance variability”.
During the last year, my focus moved from designing more trading codes and indicators to the design and implementation of a robust position sizing strategy making use of proprietary- and standard indicators that I fused into an array of FinTech and/or Low-level AI techniques to determine an optimal position size.
I use tight ATR-based Targets and Stops.
Results:
The full set of data for almost 13 years on PRT was used in the backtest, in tick-by-tick mode.
Results are based on a post-PRT-Walk forward simulation of one indicator to make sure that the overfitting of data is kept to a minimum.
The code is successful in trading 8 different equity market indices and plenty of prominent single stock futures supplied by the IG.com broker platform.
Details:
9-11-2023 – Optimised for GAIN, 50:50 In-sample to Out-of-sample ratio
All CFD mini contacts:
MARKET. CAGR. STARTUP CAPITAL.
FANG 90% $20k
US100 95% $20k
US500 39% $20k
DOW 19% $20k
DAX 14% $20k
CAC 8% $20k
FTSE 8% $20k
SA40 9% ZAR200kNote that the above individual indices do carry a different “Risk load” because as the market liquidity decreases, price volatility normally increases.
Conclusion:
I thought of sharing the above observations so that members can create “prompts” within the discussion framework, for all of us to understand any weaknesses in creating algo’s and for members to engage in discussions to create mutual opportunities.
Regards,
Niel
(I do struggle to upload additional screenshots to the gallery but will they in a follow-up comment section)
1 user thanked author for this post.
11/20/2023 at 9:51 AM #22394711/28/2023 at 7:57 AM #22442311/28/2023 at 8:08 AM #224425Niel, I agree with the position sizing playing an integral part of any strategy not only in regards to risk but also returns. To pick the right time to increase or decrease or be square is paramount. Also I have found that to have a well diversified portfolio of different strategies on the same instrument is sometimes better than to have one strategy on various instruments, to have various different strategies on various different instruments is the goal but margin call then comes into play. Robustness for me is to test a algo developed on a instrument on another instrument that way you can insure the different data set is really different. I also try to stay away from short time frames, too many rabbit holes, I stick to 1 hour TF but every now and again I find myself trying to code in a 5min TF, difficult to try and avoid but one should try and resist the temptation.
11/28/2023 at 9:03 AM #22443011/28/2023 at 1:33 PM #224476Hi All, as promised find the US100 mini, Walk-Forward Optimisation results that was done today together with the implementation thereof in a backtest. The only parameter that was optimised is the Risk Level.
The obvious initial questions are:
a. Whether the results are objective and achievable ?
b. Whether the broker would be able to perform on the order book ?
Let us see whether this can create a positive discussion in search of the optimal strategy.
11/28/2023 at 1:42 PM #22447811/28/2023 at 1:44 PM #22448211/28/2023 at 1:58 PM #22448511/28/2023 at 2:10 PM #22448711/28/2023 at 2:11 PM #22448911/28/2023 at 2:26 PM #22449311/28/2023 at 2:28 PM #22449511/28/2023 at 2:42 PM #22449811/28/2023 at 2:51 PM #224500 -
AuthorPosts