Skip to content

Harami

Patterns candlestick reversal exhaustion

A two-candle pattern in which a small candle (the "baby") forms entirely within the body of the preceding large candle (the "mother"). It signals a potential pause or reversal after a strong directional move.

Visual Example

Harami: large red candle followed by a small green candle whose entire body (and ideally range) lies inside the prior candle's body. Annotation labels the containment relationship.

Synthetic ideal satisfying TA-Lib CDLHARAMI containment rules. Generated 2026-05-31 IST via docs/gen_candle_previews.py.

Description

The Harami reflects a sudden contraction in volatility and a loss of conviction by the dominant side. The small second candle demonstrates that the prior aggressive buying or selling has been contained.

It is a higher-probability warning when the first candle is a long body in the direction of the trend and the second candle is a Doji (Harami Cross). Traders and strategy developers use it as a trigger filter inside established Market Structure bias or as an ML feature indicating "volatility collapse after impulse."

Formula / Specification

Recognition Rules (exact implementation in QuantWave / TA-Lib CDLHARAMI):

  1. Candle 1: large body (direction determines bullish/bearish context).
  2. Candle 2: small body whose high and low (or at minimum open/close) lie strictly inside Candle 1's body.
  3. The pattern completes on the close of the second bar.
  4. Output follows TA-Lib +100/−100/0 convention.
  5. Two-bar window; the streaming wrapper retains the necessary prior bar state.

Parameters

Parameter Default Description
(none) — Pattern recognition only; no tunable parameters.

Usage Examples

Streaming (Rust)

use quantwave_core::indicators::CDLHARAMI;
use quantwave_core::traits::Next;

let mut h = CDLHARAMI::new();
for (o, h_, l, c) in &ohlcv {
    let sig = h.next((o, h_, l, c));
    ...
}

Streaming (Python) / Polars Batch — analogous to Doji examples, using CDLHARAMI / .ta.cdl_harami("open","high","low","close").

Bit-identical results across all three surfaces.

Edge Cases & Limitations

  • Requires two completed bars.
  • False positives frequent in low-volatility ranges; require trend context or volume confirmation.
  • "Inside body" vs. "inside full range" definitions vary in literature — QuantWave follows the TA-Lib body-containment rule.
  • Harami Cross (second candle is Doji) is generally stronger.
  • Most useful after a clear impulse move, not in chop.

Boundary Behavior

Condition Behavior
Warm-up Pattern functions emit 0 (no pattern) until enough bars exist.
period > len Short series returns all zeros (no pattern detected).
NaN inputs Bars with NaN OHLC are treated as no pattern (0).
Invalid params N/A for most candlestick patterns.
Empty data Empty input returns an empty integer series.

Sources & References

Primary Source: TA-Lib CDLHARAMI via quantwave-core/src/indicators/pattern.rs.

Visual: docs/gen_candle_previews.py, 2026-05-31 IST.

Context: Nison (1991) for exhaustion psychology (no boilerplate duplication). MQL5 for structure confluence.

Provenance: Next parity contract.