Ichimoku, Drawn Out
Ichimoku looks like the most complicated indicator in this Academy. Several lines, and some of them drawn to the right of the last candle — the first time you see it, the chart looks broken.
The arithmetic is among the simplest here, though. Three of the four lines are the exact midpoint the Donchian lesson already builds, with only the period changed. What's hard about Ichimoku isn't the maths. It's the time axis.
1. Three of the four lines are the same calculation
Tenkan is (highest high + lowest low) ÷ 2 over 9 bars. Kijun is the same over 26. Senkou B is the same over 52. All three are Donchian midpoints; only the period differs.
The remaining line, Senkou A, is simpler still: the average of Tenkan and Kijun. That's the whole indicator — no moving averages, no standard deviation, no exponential weighting anywhere.
Tenkan is the midpoint of the highest high and the lowest low of the last 9 bars — a Donchian midpoint, exactly what the Donchian lesson builds. Only the period changes.
- Price (close)
- 9-bar high / low
- Their midpoint — the Ichimoku line
BTC/USDT · 4h · Binance · 2026-05-01 → 2026-07-23
Classical Ichimoku has a fifth line, the Chikou span — the close, plotted 26 bars back.
This app's template neither computes nor uses it, so it isn't in this lesson either. If you read about the Chikou span elsewhere, treat it as something outside the strategy you'd be simulating here.
2. Why the cloud sits ahead of price
The two leading spans are not drawn where they were computed. They're shifted 26 bars to the right, which is why the cloud carries on past the last candle into space where no candles exist yet.
That's what "leading" means in their name: a value produced today is placed 26 bars into the future.
Both cloud edges are computed from a bar, then drawn 26 bars to its right — which is why the cloud runs past the last candle into empty space. The faint dashes are where each value was computed; the solid lines are where it gets drawn.
- Price (close)
- Where the span was computed
- Where it is drawn
BTC/USDT · 4h · Binance · 2026-05-01 → 2026-07-23
There's a trap here that quietly breaks strategies. The cloud drawn over bar i was not computed at bar i — it came from bar i − 26. If your code reads the Senkou value at bar i directly, it is using data that doesn't exist yet at that moment.
This app's template avoids it by reading back with `src_idx = i - displacement`, and the Pine side does the same thing with `senkouA[displacement]`. Keep that rule if you write your own: look-ahead bias flatters a backtest and never survives live trading.
3. The cloud and its colour
The cloud is the filled area between the two leading spans. Whichever is higher forms the top and the other the bottom, and when they swap places the cloud changes character.
Senkou A above Senkou B is a bullish cloud; below it, bearish. Where the two cross, the cloud narrows to zero thickness — what's usually called a twist.
Filled between the two leading spans. Bullish when Senkou A is above Senkou B, bearish when below — 41% of the 423 bars that have a cloud are bullish here, and it changes sides 8 times. Where the spans cross, the cloud pinches to nothing: a twist.
- Price (close)
- Bullish cloud (Senkou A above B)
- Bearish cloud (hatched)
BTC/USDT · 4h · Binance · 2026-05-01 → 2026-07-23
Thickness carries information too. A thick cloud means a wide range of prices was established over that stretch; a thin one means price was tightly clustered. Across this window thickness runs from 0.01% to 7.57% of price.
4. Most bullish crosses never become entries
The template's buy needs two things: Tenkan crossing above Kijun, and the close above the cloud on that same bar.
Both have to be true on one bar. And the cross doesn't repeat while the trend holds — miss it once and it's gone.
The template needs a Tenkan/Kijun cross AND price above the cloud on that same bar. Of 13 bullish crosses here, 4 qualified — 31%. 5 were below the cloud, 3 inside it, and 1 came before the cloud existed.
- Price (close)
- Cloud
- Cross that became an entry
- Cross that was discarded
BTC/USDT · 4h · Binance · 2026-05-01 → 2026-07-23
Of the 13 bullish crosses in this window, 4 became entries — 31%. Five had price below the cloud, three had it inside the cloud, and one came before the cloud existed at all. That's the same effect the Simulator's re-entry filter cites as "39% across 8 assets", reproduced independently on this data.
5. The re-entry filter
All those discarded crosses mean long stretches where the trend read is bullish but there is nothing to act on. The re-entry filter exists for exactly those stretches.
Its condition is a change of state, not a cross: it adds a buy on the bar where price crosses back above the cloud while Tenkan is already above Kijun. Crossing back above, not merely being above — as a state it would be true for most of a trend and would re-fire every single bar.
A cross is tested on one bar and never repeats while the trend holds, so a discarded one is gone for good. This block adds a buy when price crosses back above the cloud while Tenkan is already above Kijun — 2 times here, against the template’s own 3 completed trades.
- Price (close)
- Tenkan
- Kijun
- Cloud
- Cross that became an entry
- Re-entry filter trigger
BTC/USDT · 4h · Binance · 2026-05-01 → 2026-07-23
6. The warm-up is long
Senkou B needs 52 bars before it produces a first value, and the 26-bar displacement stacks on top of that. The first readable cloud is bar 78.
Backtest on a 200-candle pull and 39% of it has no cloud at all. Move the two sliders below to see how the warm-up and the surviving signal count move together.
Senkou B needs its full period before it has a value, and the cloud over bar i was cast from bar i − displacement, so the two add up. Here the first readable cloud is bar 77 of 500 — 15% of the window — leaving 13 bullish crosses and 4 entries in what remains.
- Price (close)
- Cloud
- No cloud yet
BTC/USDT · 4h · Binance · 2026-05-01 → 2026-07-23
The charts in this lesson use a longer 500-bar window than the rest of the Academy — same asset, same 4h timeframe, just more of it. On 200 bars there were 2 bullish crosses, 1 entry and 0 completed trades: nothing to show you.
You'll hit the same thing in the Simulator. Pull enough data that there's still a usable stretch left after the 78-bar warm-up.
7. What actually happened here
This 500-bar window is a 15.2% downtrend. The template completed 3 trades and all three were small losses — −1.7%, −2.8%, −2.7% — with two exits triggered by the Tenkan/Kijun cross and one by price dropping out of the cloud.
The window was chosen deliberately. This is what a long-only trend template does in a falling market, and the thing to judge isn't the return — it's whether the exit rule cut the losses short. That's the job of the sell being an OR: a downward cross OR a cloud break. Buys are an AND and therefore strict; sells are an OR and therefore loose. The asymmetry is the design.
8. In the Simulator
Ichimoku ships as one template and one filter. The two share their Tenkan, Kijun, Senkou B and displacement settings, so stacking the filter onto the template adds only the toggles.
- Ichimoku Cloud — enters when Tenkan crosses above Kijun with price above the cloud, exits on a downward cross or a drop below the cloud.
- Ichimoku Cloud Re-Entry — an optional block that adds a buy when price crosses back above the cloud while Tenkan is already above Kijun.