HaramiCross

A stronger Harami — a large real body followed by a Doji contained within it; the total indecision after a strong move makes the reversal more potent.

Quick reference

Field	Value
Family	Candlestick Patterns
Input type	Candle (open / high / low / close)
Output type	f64 (signed signal)
Output range	{−1, 0, +1}
Default parameters	None (doji threshold fixed at 0.1)
Warmup period	2
Interpretation	+1 bullish reversal; −1 bearish reversal.

Formula

prior is a solid body (not a doji); current is a doji (|close−open| <= 0.1·range)
contained: current open AND close within [min(open,close)[-1], max(open,close)[-1]]
Bullish (+1): prior close < open (bearish body)  AND doji contained
Bearish (−1): prior close > open (bullish body)  AND doji contained
otherwise 0

The Harami Cross is a Harami whose inside bar is a Doji — the market goes from a decisive large body to complete indecision, a stronger reversal cue than a small-body inside bar. Source: crates/wickra-core/src/indicators/harami_cross.rs.

Parameters

Name	Type	Default	Valid range	Source	Description
—	—	—	—	None.	Harami Cross is parameter-free; the doji threshold is fixed at 0.1.

Signed ±1 encoding

+1.0 bullish, −1.0 bearish, 0.0 no pattern (warmup emits None).

Inputs / Outputs

From crates/wickra-core/src/indicators/harami_cross.rs:

use wickra::{Candle, Indicator, HaramiCross};
// HaramiCross: Input = Candle, Output = f64
const _: fn(&mut HaramiCross, Candle) -> Option<f64> = <HaramiCross as Indicator>::update;

A Candle in, an Option<f64> out. Python update(candle) / batch(open, high, low, close); Node update(open, high, low, close) / batch(...).

Warmup

warmup_period() == 2. The first candle seeds the prior bar (first_bar_seeds_without_signal pins this).

Edge cases

• Bullish. Big bearish body then contained doji → +1 (bullish_harami_cross pins this).
• Bearish. Big bullish body then contained doji → −1 (bearish_harami_cross pins this).
• Doji outside body → 0. (doji_outside_body_is_zero pins this).
• Second not a doji → 0. (non_doji_second_is_zero pins this).
• Finiteness. Candle::new rejects non-finite fields, so no in-method guard is needed.
• Reset. h.reset() clears the prior bar and the last value (reset_clears_state).

Examples

Rust

use wickra::{Candle, Indicator, HaramiCross};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut h = HaramiCross::new();
    h.update(Candle::new(110.0, 110.2, 99.8, 100.0, 0.0, 0)?); // big bearish body
    let sig = h.update(Candle::new(105.0, 106.0, 104.0, 105.02, 0.0, 0)?); // contained doji
    println!("{:?}", sig); // Some(1.0)
    Ok(())
}

Output:

Some(1.0)

Python

import numpy as np
import wickra as ta

h = ta.HaramiCross()
o = np.array([110, 105]); hi = np.array([110.2, 106]); lo = np.array([99.8, 104]); c = np.array([100, 105.02])
print(h.batch(o, hi, lo, c))  # [nan, 1.0]

Node

const ta = require('wickra');
const h = new ta.HaramiCross();
h.update(110, 110.2, 99.8, 100);
console.log(h.update(105, 106, 104, 105.02)); // 1

Streaming

use wickra::{Candle, Indicator, HaramiCross};

let mut h = HaramiCross::new();
let feed: Vec<Candle> = Vec::new(); // your live stream
for candle in feed {
    if let Some(s) = h.update(candle) { if s != 0.0 { /* reversal */ } }
}

Streaming update and batch are equivalent tick-for-tick (batch_equals_streaming pins this).

Interpretation

1. Strong reversal cue. The shift from conviction to indecision warns the prior move is exhausting.
2. Confirm. Wait for a confirming bar in the signalled direction.
3. Location. Strongest at a tested level or after an extended run.

Common pitfalls

• Containment, not engulfing. The doji must sit inside the prior body — opposite of an engulfing.
• Threshold sensitivity. What counts as a doji depends on the fixed 0.1 body/range ratio.
• Needs trend. A Harami Cross in a range carries little meaning.

References

Nison, S. (1991), Japanese Candlestick Charting Techniques.

See also

• Indicator-Harami — the small-body version.
• Indicator-Doji — the indecision candle.
• Indicator-Tristar — three-doji reversal.
• Indicators-Overview — the full taxonomy.