Improve CTA Performance By Tweaking Trend Parameters

The very basic CTA strategy has two key components: a) trend judge, b) timing on entry/exit/adjust. In this entry, I’m using SMA to determine the trend direction and strength, specifically it’s using SMA(60) and SMA(120) for trend filters, with the relative position suggesting the trend direction, and the slope of SMA(60) suggests the trend strength.

Now I’m trying to understand whether the choice of long term trend is sensitive to the parameters (I expect it don’t matter much); and whether the slope of the filter matters (I expect it do matter).

The comparison would be done with a) long term trend will run on 100/120/160/200, periods; b) the slope will check on 1. NIL, 2. short term, 3. long term, 4. both.

Performance measure is collected for ROR/SHA/WIN/ProfitFactor/MDD.

Results
Backtest results collected below are done for Year 2020 on 1H bars via GTJA platform. Products cover 22 different instruments in agriculture, soft commodities, metals, bulk, energy, and chemicals listed on SHFE, DCE and ZCE. Following tables show the performance matrixes (ROR, Sharpe, WinRatio, Profit Factor, MDD, Vol) under different trend periods vertically with different ichimoku settings horizontally.
Note that top performer is located at lower right corner of the matrix, for Sharpe, MDD, Volatilities, WinRation, PF.

Conclusion
Backtest proves the following intuitive thoughts:
a. SHA and MDD (and also WIN/PF) improves when trend is using longer periods, i.e. 160/200 is prefered over 100/120.
b. 2/3/4 with ichimoku filters work better than 1 which has no checks on ichimoku directions.

I would run longer period tests (Y16-Y20) for more products via GoldMiner to confirm the conclusion.