Ichimoku Cloud Breakouts Tested: 329 Signals on Real EURUSD Data

The cloud breakout is the one Ichimoku rule everybody agrees on: wait for price to close beyond the cloud, then trade in that direction. I ran it across 16,926 real EURUSD 5-minute bars from June 3 to August 26, 2026, with no discretion and no curve fitting. It came out marginally positive, which already makes it unusual in this series, but the reason it works is not the reason people give for it.

How was the cloud breakout tested?

Same dataset and the same measurement rules as my ADX above 25 test and my moving average crossover test, so the numbers are directly comparable.

• Indicator: standard Ichimoku 9 / 26 / 52. Senkou Span A is the midpoint of Tenkan and Kijun, Senkou Span B is the 52-bar midpoint, and both are plotted 26 bars ahead — so the cloud used for a signal is the cloud your platform actually draws at that bar.
• Signal: the previous close was inside or below the cloud and this close is above the top of it. That is a long. The mirror is a short.
• Measurement: the move in pips after a fixed hold, in the signal direction. No stop, no target, so the number describes the signal itself rather than a money management scheme layered on top of it.
• Overlap rule: after a signal the clock skips forward by the hold length, so one move is never counted twice.

What did cloud breakouts pay at each hold?

Look at the 30 minute row first. A 37.4% win rate is the worst number in this entire study, and it comes from the most popular entry in the whole system. The breakout bar itself is noise: price poking through the edge of the cloud, the move stalling, and half an hour later you are underwater. Give the same signal two hours and it flips.

Now look at the 8 hour row. Everything earned by hour two has been handed back. This is the exact opposite of what I found with ADX filtered signals, where the eight hour hold was the only profitable column. A strength filter says “a trend is running, stay in it.” A cloud breakout says “something just changed” — and change is a short lived condition.

Does a thicker cloud make a breakout safer?

This is the claim I most wanted to test, because it is stated as fact in nearly every Ichimoku tutorial: a thick cloud is strong support and resistance, so breaking through it means more. I measured cloud thickness in ATR units at the moment of the breakout. One caution before the table: these two rows are not a clean halving of the 329 signals above. Thickness is measured on a wider signal set that includes breakouts excluded from the headline test by its overlap rule, which is why 163 and 244 sum to more than 329, and why both rows can be negative while the headline number is positive. Compare the two rows against each other, not against the +0.29.

Both are negative once you isolate them, and the thick cloud is the worse of the two. The story that a fat cloud makes a break more significant is simply not visible here. If anything it points the other way: a thick cloud on 5-minute EURUSD is usually the residue of a move that already happened 26 bars ago, and breaking it is late news rather than fresh news. The median cloud thickness in this sample was 1.39 ATR, so “thick” is the normal state of the chart, not a special event worth waiting for.

How often is price actually outside the cloud?

Price is outside the cloud 85.1% of the time. That matters for anyone using “above the cloud” as a bullish condition and “below the cloud” as a bearish one: those two states cover almost the whole chart, so the condition is nearly always satisfied in one direction or the other. It describes where price has been. It does not remove much.

The 14.9% spent inside the cloud is the genuinely useful reading. That is the market saying it has no direction across the last 26 to 52 bars, and it is the one Ichimoku state I would actually act on — by standing aside.

Does confirming with Tenkan and Kijun help?

I added the obvious filter: only take the cloud breakout if Tenkan is already on the correct side of Kijun. It made things worse. 229 signals, 42.8% win rate, -0.12 pips against +0.29 unfiltered. The confirmation arrives after the information is already in the price, so it removes precisely the early signals where the edge lived. That is the same trap I found with moving average filters on crossovers: filtering a trend signal with another trend signal deletes the entries that were worth having and keeps the late ones.

What I would actually do with this

• Do not scalp cloud breakouts. Thirty minutes was the worst result in the whole study at 37.4%. If you cannot hold a couple of hours, this signal is not for you.
• Treat two hours as the working window. One hour is break even, four hours is fading, eight hours is negative.
• Ignore cloud thickness. It did not predict anything here, in either direction.
• Use “inside the cloud” as a stand aside flag. That is 14.9% of bars where the system openly admits it does not know.
• Do not stack confirmations. Adding Tenkan and Kijun agreement turned a small plus into a small minus.

Frequently asked questions

Is the Ichimoku cloud breakout profitable?

Marginally, on this data. 329 signals across 16,926 EURUSD 5-minute bars produced +0.29 pips per trade at a two hour hold, or +96.8 pips in total before any spread. Positive, but thin enough that a typical spread would erase it, which is the honest way to read the number.

Why is the cloud plotted 26 bars into the future?

By design. Senkou Span A and B are calculated from past prices and then displaced forward, so the cloud you see ahead of price today was built entirely from data that already exists. Nothing is being predicted. It is a projection of the recent midpoint range, which is also why breaking a thick cloud is often late rather than significant.

Which timeframe is Ichimoku designed for?

Goichi Hosoda built it around daily charts and a Japanese trading week that ran six days, which is where 9, 26 and 52 come from. Nothing about those numbers was chosen for 5-minute forex bars, and my Ichimoku settings guide for MT4 shows what happens when you change them.

Does the cloud work better than a simple moving average?

On the same data and the same hold, yes, and by more than I expected. The full four way comparison against EMA crossovers and Donchian breakouts is in my Ichimoku against plain trend following test.

What data was this run on?

16,926 EURUSD 5-minute bars, June 3 to August 26, 2026 — the same set behind my Stochastic overbought test, my Fibonacci retracement test and my ADX against Supertrend comparison.