A strategy needs reproducible rules, not just five plotted lines. This lesson defines an author-designed research variant: a cloud-boundary crossing with a Tenkan–Kijun filter. It supplies hypothetical inputs and cash-flow alternatives, not results from an executed backtest or a recommendation to trade.

This piece is about trading the signals. If you are still learning what each line means and where the 9-26-52 parameters come from, start with the basics of the Ichimoku system, then come back here for the setups.

Keep calculation time and plotted time separate

Our test uses exactly 26 index shifts: Span A and B calculated at t are plotted at t+26, and Close at t is drawn as Chikou at t−26. The cloud beneath current price therefore comes from calculations 26 candles earlier. This is the model’s explicit indexing convention; verify the implementation rather than assuming the parameter name fixes every platform’s indexing.

Let Ut=max(At−26,Bt−26) and Lt=min(At−26,Bt−26), where A and B mean calculated span values before plotting displacement. The test uses ordinary EUR/USD D1 Bid candles and settings 9–26–52 with sufficient history for every input. Colours are display settings. Shifting known historical calculations forward does not supply future price information.

Crossing above a cloud need not mean a rising Close

The condition compares moving values. Separate price movement from boundary movement, even when using the cloud plotted beneath the current candle correctly.

Scroll sideways to see the full diagram

Price can move above a falling cloud edgeConstant Close 103 moves from below U105 to above U102. Close does not rise.ABOVE THE CLOUD WITHOUT A RISE IN C99103107t −1tClose:103 →103; price change=0.Upper edge beneath price:105 →102.Tenkan 106 > Kijun 104: specified filter.All three model conditions pass.

Hypothetical rule inputs in arbitrary units. C(t−1)=C(t)=103; the correct U beneath those candles is 105 and 102, with lower edges 101 and 99. Tenkan(t)=106, Kijun(t)=104. Cloud boundaries are supplied, not reconstructed from complete OHLC. All inputs are known after t closes; no execution or result is shown. The document TradingView — Ichimoku Cloud supports the cloud's successive calculations and displaced plotting.

In this model, Close stays 103 while U falls from 105 to 102. With the specified Tenkan> Kijun filter, all three conditions of this illustrative variant pass.

You could additionally require C(t) > C(t−1), creating another variant that rejects this case. Do not add it after observing a result merely because the name “breakout” suggests a rising Close. The main research variant here uses these three conditions; the separate four-filter illustration uses different inputs.

Price relative to the cloud — the foundation of every bias

After candle t closes, the signal requires C(t−1) ≤ U(t−1), C(t) > U(t), and Tenkan(t) > Kijun(t). C is the Bid close. This tests a crossing, not simply residence above the cloud; all three conditions belong to this variant. It does not require a simultaneous Tenkan–Kijun crossover, Chikou filter or cloud colour.

Cloud position is a recorded classification, not proof of future support or resistance. A wider cloud is a larger difference between its two boundaries; it is not a wall that necessarily stops a strong move. The trend-following approach and trend-based technical analysis tools can use other definitions, which need separate evaluation.

The Tenkan and Kijun cross — a signal you always read in context

A Tenkan–Kijun crossover and the state Tenkan > Kijun are different conditions. This variant uses the state after the signal candle closes. Adding a crossover on that same candle defines another candidate set; calling it stronger does not establish its net outcome.

The model permits one long position at a time with no pyramiding or shorts. Buy a fixed 10,000 EUR at the first available Ask after a confirmed signal; constant volume does not mean constant money risk. Skip and record the signal if the first Bid is already at or below the planned stop, or provider conditions prevent the order.

Exit and protection are separate rules

The initial ordinary stop is L(t) − 0.0002, two 0.0001 pips below the signal candle’s lower cloud boundary, and stays fixed in this variant. The buffer is chosen for illustration, not an optimal parameter. From the next candle onwards, a Bid close below that candle’s Kijun requests a sale at the next available Bid.

The first actual closing fill ends the position; remove any remaining exit instruction. There is no fixed take profit in this variant. Chikou or an additional support or resistance filter would be another rule, requiring its own run rather than an assumed last-chance improvement.

Signal inputs and alternative execution paths

Supply hypothetical inputs rather than claiming reconstruction from complete OHLC: previous C=1.0995 and U=1.1000; current C=1.1005 and U=1.1000; Tenkan=1.0990 and Kijun=1.0980. All three conditions pass. L=1.0950 gives SL=1.0948. Assume a subsequent actual Ask buy at 1.1008 with Bid above SL, a USD account and 10,000 EUR. Planned distance is 60 pips, and the chosen combined commission/funding cost is USD 4.

Alternative actual Bid fills give: 1.1068 after a prior Kijun exit condition, +USD 60 before cost and +USD 56 after; exact stop fill 1.0948, −USD 60 and −USD 64; worse stop fill 1.0938, −USD 70 and −USD 74. Use 10,000 × (Bid exit − 1.1008) − 4. The profit fill is not a preset TP. Spread is already in Ask entry and Bid exits. None is a maximum possible loss. If entry worsens to Ask 1.1013, the first two net results become +USD 51 and −USD 69.

A separate geometric example retains the earlier C=1.0900, cloud 1.0830–1.0860, Tenkan 1.0890 and Kijun 1.0875, plus assumed Chikou-above-history and green future cloud. That four-filter state is another variant and does not establish the main crossing without a previous C/U pair. Independently assuming an actual Ask fill at 1.0900 and Bid stop fill at 1.0825 gives 75 pips planned distance; hypothetical Bid 1.1050 gives 150 pips and reward/risk 2 before fees. There is no universal minimum worthwhile ratio or holding period inferred from those prices.

Practical takeaway: Specify the price field, displacement and candle completion used in every Ichimoku comparison.

Test the filter with the position state

A filter can unlock later entries when only one position is allowed. In a separate fictional schedule, signals occur at indices 1, 3 and 5. Their assumed exits, if taken, are 6, 4 and 7 respectively. Without a filter, entry 1 blocks entries 3 and 5. A filter rejecting 1 but admitting 3 and 5 yields two trades: 3→4 and 5→7. No P&L or actual Ichimoku signal data is implied.

The filtered trades are not a subset of the unfiltered trade list: rejecting the early trade freed the position state. Run each variant separately with chronological entries and exits rather than deleting rows from a completed journal. Compare net result, drawdown, trade count and exposure time under the same data and costs. Adding multi-timeframe analysis is another variant, not an assumed improvement or a reason to prefer yen pairs.

What to do tomorrow

  1. Save the instrument, timeframe, price field, 9–26–52 settings and explicit displacement convention. Verify the data available on the decision candle instead of reading a backward-plotted current Close as information known earlier.
  2. Keep the main three-condition crossing and any four-filter state as distinct variants. In the diagram, Close 103 stays constant while U falls from 105 to 102; with Tenkan 106 > Kijun 104 all three main conditions pass. An extra rising-Close condition would reject it.
  3. Reserve a separate evaluation period and record all candidates, rejected signals, skipped entries and positions still open at its end. Changing rules after seeing this period makes it part of tuning. Two months of practice is not a universal proof threshold.
  4. Compare the Tenkan–Kijun variant against a fresh simulation without that filter. Test higher costs and alternative settings, and record intrabar assumptions or finer data for ambiguous fills. History does not guarantee a future outcome, and an ordinary stop does not guarantee a loss ceiling.