Ichimoku Kinko Hyo & Supertrend: Can They Deliver a Winning EA Logic?

A beginner-friendly summary of the verification: “Ichimoku Kinko Hyo & Supertrend: Can They Deliver a Winning EA Logic?”.

Breakout entry example (XAUUSD daily, real data): buy when price breaks above the recent high.

Breakout entry example (XAUUSD daily, real data): buy when price breaks above the recent high.

What’s the idea?

Alright, let’s talk about trend following! We’re always on the hunt for new “edges” in the market – those little statistical advantages that can help our algorithmic trading systems (EAs) make money. This time, I decided to put two well-known trend indicators, Ichimoku Kinko Hyo and Supertrend, to the test. Could they offer a fresh perspective on capturing trends, or even replace parts of our existing, proven strategies? My goal was simple: see if these indicators could identify profitable long (buy) trends on their own, and then, if they showed promise, see how they’d perform when integrated into my core multi-timeframe system.

How I tested it

To keep things fair and realistic, here’s how I set up the experiment:

The Indicators:

Ichimoku Kinko Hyo: Often called the “cloud chart,” this is a comprehensive trend-following indicator. I specifically used its Tenkan-sen (conversion line), Kijun-sen (base line), and Senkou Span A and B (leading spans) to form the “Kumo” (cloud). Crucially, the cloud was shifted 26 periods back, ensuring no look-ahead bias. This means the indicator only used past data, just like in real trading.
Supertrend: This is an ATR (Average True Range)-based trailing stop indicator. It essentially shows you the direction of the trend and helps identify potential entry/exit points with confirmed bands.
Both strategies were tested as long-only, meaning they only looked for opportunities to buy.

The Portfolio: I tested these on a diverse portfolio of 8 major currency pairs. This helps ensure the results aren’t just luck on one specific pair.
The Data & Timeframe: We used “clean” historical data from 2015 to 2024, covering a good range of market conditions. I ran tests on both the H4 (4-hour) and D1 (daily) timeframes.
The Method: I used a “walk-forward validation” approach with fixed parameters. This is like simulating real-world trading by testing the strategy on unseen data, rather than just optimizing it to fit historical data perfectly. It’s a much more robust way to see if an EA has a true edge.

What happened? (The good news… mostly)

The initial standalone results were actually quite promising!

H4 Timeframe Performance:

Ichimoku:
Total Profit: +42.9%
Profit Factor (PF): 1.11 (In other words, for every $1 lost, it made $1.11. Anything above 1 means profitable!)
Sharpe Ratio: 0.45 (This measures risk-adjusted return. A higher number is better, indicating good returns for the risk taken.)
Profitability: It was profitable in 7 out of 10 years tested.
Supertrend:
Total Profit: +49.1%
Profit Factor (PF): 1.14
Sharpe Ratio: 0.48
Profitability: Profitable in 6 out of 10 years. Compared to our baseline: Both of these performed significantly better than our simple BreakoutLong baseline strategy on the H4 timeframe, which only managed +13.6% total profit and a Sharpe of 0.22. So far, so good!

D1 Timeframe Performance:

Ichimoku: Still positive with +11.1% total profit. Not as strong as H4, but still in the green.
Supertrend: On its own, Supertrend actually failed on D1, showing a -5.4% loss. However, when paired with a simple SMA150 (150-period Simple Moving Average) filter, its performance rebounded to +10.5% with a PF of 1.35. This shows how combining indicators can sometimes rescue a struggling strategy.

What happened? (The crucial twist!)

These results might make you think we’ve found some amazing new tools, right? Well, here’s where things get interesting, and why we always dig deeper.

The Problem: High Correlation!

Despite the positive standalone performance, I found a critical issue: the daily correlation of both Ichimoku and Supertrend with our existing BreakoutLong strategy was incredibly high – between 0.82 and 0.86 for both H4 and D1 timeframes! What does high correlation mean? Imagine you have a basket of different investments. If they all move up and down together, you don’t actually have much diversification, even if they look different on the surface. It’s like having a basket full of different flavors of ice cream – they’re all still ice cream, and if the ice cream market crashes, you’re in trouble! In trading, high correlation means these new strategies are essentially capturing the same trend edge as our existing ones, just using different indicators to do it. They aren’t providing a new, uncorrelated edge that could help spread risk and make our overall portfolio more robust. (Remember from Study 50, even a correlation of 0.64 wasn’t enough for effective diversification.)

What I learned (The first pass)

My initial conclusion was clear: Ichimoku and Supertrend do possess a genuine trend-following edge – they are profitable when tested rigorously with walk-forward validation. However, they lack novelty because they are highly correlated with our existing core trend strategies. This reinforces a recurring theme in my research: it seems that most price-only logic that produces a positive walk-forward edge tends to converge on similar trend-following principles and, as a result, ends up being highly correlated with each other. It’s like rediscovering the same mountain using different paths! (This is similar to what I found in Study 34 with machine learning approaches to trend detection).

A deeper dive: Integrating into my core system

But what if we tried to integrate one of these into our established, robust multi-timeframe proprietary system? Could Ichimoku, for example, replace the Breakout entry logic in our existing FX trend system? This is the ultimate test – not just if an indicator performs well alone, but if it can improve an already optimized system. I ran an additional test (Study 56), replacing the Breakout entry with an Ichimoku entry within our core FX trend system.

The Final Verdict

The results of integrating Ichimoku into our core system were, unfortunately, a complete deterioration of performance:

Original System (FX=Breakout):
Average Monthly Profit: 0.47%
Maximum Drawdown (DD): -9.0% (The biggest peak-to-trough loss in equity)
Overall System Score (MC): 85.1% (A composite score of system health)
Profit Factor (PF): 1.30
Modified System (FX=Ichimoku Entry):
Average Monthly Profit: 0.38% (Lower)
Maximum Drawdown (DD): -12.6% (Significantly higher, and exceeded our acceptable limits!)
Overall System Score (MC): 72.8% (Much lower)
Profit Factor (PF): 1.15 (Lower)
Trade Count: 2.3 times higher! (Meaning much higher transaction costs and more “noise”) Why did this happen? Even though Ichimoku showed a decent Sharpe Ratio on the H4 timeframe in standalone tests, its entry/exit signals (specifically, frequent breaks of the Kijun-sen, or base line) led to too many trades and, more critically, significantly inflated drawdowns when integrated into the larger system. It just wasn’t as robust an entry point as our established Breakout logic. This highlights a crucial point: a great individual player doesn’t always make a great team. Just because an indicator performs well in isolation doesn’t mean it will improve a complex, multi-timeframe system. My previous research with Pyramid and Higher Time Frame (HTF) strategies has shown this repeatedly: individual indicators are not the same as robust system performance. Final decision: There will be no changes to our confirmed core system. While Ichimoku and Supertrend are interesting indicators and can show an edge, their high correlation with existing strategies and their failure to improve (or even maintain) the performance of our robust system means they won’t be integrated at this time. The hunt for truly new, uncorrelated edges continues!

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