DOUBLE TRIPLE IMPORTANT CHANGE. AGAIN. yes again. As of June 26, 2018 I am REMOVING BOTH THE 1000 PERIOD LOOKBACK TREND FILTER AND THE STDEV 1.1 FILTER. After careful consideration, I feel that stdev was being calculated in real time on the actual entry bar, and may be the reason why so many backtest trades were not duplicated in the real world. After I removed lookahead bias (including the current entry bar in lookback trend calculation), the advantage of the 1000 bar lookback was gone - however... the max drawdown is still lower, so it may be worth offering this as an option.
IMPORTANT CHANGE: As of June 25, 2018, I ADDED TRIANGLESMA back into DifferenceEngine. As well, I changed the symbol set to:
As of today, April 8th, 2018, I have made a change to DifferenceEngine multisystem and TwentySomething and installed the new version into all Demo and Funded AlgoLabs.
Upon deeper analysis, I discovered that the 4 other smaller secondary systems that are combined with the main system in the DE multisystem are doing nothing to add to lowering the Pain/Gain ratio, and in favor of robustness, I thought that removing these systems was prudent.
I also discovered that these 4 secondary systems were the cause of many of the losing trades since this drawdown that started on March 1st, 2018, and in fact, if not for these 4 secondary systems, the period from March 1 to April 4th would be net profitable.
I am also going to take a closer look at SuperSystem, but I feel that multisystem might be negatively effected by the removal of the secondary system (there is only a single secondary system in SuperSystem).
Changes to TwentySomething were the same - I removed the other 4 systems, and then made a slight change to the 3 symbols. The 3 symbols now traded are: NQ, GC and ZS (I swapped Soy Meal out and replaced it with Soy Beans).
note: that i have NOT updated the DifferenceEngine or TwentySomething data in Performance Viewer yet.