Ichimoku and Supertrend as EAs: do they actually win?

Adding Ichimoku and Supertrend indicators to my trading framework did not provide the diversification I was looking for, despite both showing decent individual performance. I recently tested these two trend-following indicators using a portfolio of eight currency pairs across a clean dataset from 2015 to 2024. To ensure the results were reliable, I used a forward-testing approach with fixed parameters. The results for the H4 timeframe were promising; both strategies outperformed my baseline Breakout Long strategy.

Strategy	Total Return	PF	Sharpe Ratio	Profitable Years
Baseline Breakout Long	13.6%	-	0.22	-
Ichimoku	42.9%	1.11	0.45	7 of 10
Supertrend	49.1%	1.14	0.48	6 of 10
Note: PF (Profit Factor) is gross profit divided by gross loss; a value above 1 indicates a profitable system.
While these numbers look good, the daily correlation between these indicators and my existing Breakout strategy is extremely high, ranging between 0.82 and 0.86. In other words, I am simply using different math to capture the exact same trend edge I already have. Because they do not offer a new, uncorrelated edge, they fail to provide the portfolio diversification I need to reduce overall risk.

Why individual performance is deceptive

I also experimented with replacing the entry logic of my current system with Ichimoku signals. The results confirmed that a strong individual indicator does not necessarily make for a stronger overall system.

Metric	Baseline (Breakout)	Ichimoku Entry
Monthly Return	0.47%	0.38%
Drawdown (DD)	-9.0%	-12.6%
MC Pass Rate	85.1%	72.8%
PF	1.30	1.15
Note: MC (Monte Carlo) pass rate is the probability of passing prop-firm rules based on resampling daily returns.
When I swapped the entry logic, the drawdown exceeded my limits and the MC pass rate dropped significantly. The Ichimoku version generated 2.3 times as many trades, which increased my total transaction costs and led to frequent exits that hurt the bottom line.
This confirms a recurring theme in my research: price-based logic tends to converge on the same trend-following behavior. Even if an indicator looks impressive on its own, it often falls apart when integrated into a robust, multi-timeframe system designed for prop firm requirements. I will be keeping my current system exactly as it is.

How this connects

This verification builds on earlier ones (what failed before and what I tried this time, comparisons between approaches).

Code to reproduce

You can reproduce this with the following scripts (see repo).

scripts/research/study_final_ichimoku.py