Bearish Harami — Backtesting a Classic Reversal Pattern

A data-driven look at the Bearish Harami candlestick pattern — win rate, optimal conditions, and what happens when you add RSI and MACD oscillator gates.

What is a Bearish Harami?

The Bearish Harami is a two-candle reversal pattern that appears at the top of an uptrend. “Harami” is Japanese for “pregnant” — the second candle is contained entirely within the body of the first.

Structure:

• Candle 1: Large bullish (green) candle — strong buying momentum, close near the high
• Candle 2: Small bearish (red) candle — opens and closes entirely within the prior candle’s body

The psychology: after a strong upward move (Candle 1), the next session gaps down and fails to recover, closing lower within the prior candle’s range. Buyers who entered on Candle 1 are now underwater. Momentum is stalling.

The pattern requires confirmation — it is a warning signal, not a guaranteed reversal. It should appear after an established uptrend, not after a single up-day.

Key Identification Rules

For a Bearish Harami to be valid:

1. Prior trend required. At least 3–5 bullish candles or a clear uptrend into the pattern. A Harami appearing after sideways price action is meaningless.
2. Candle 2 body is inside Candle 1 body. The open and close of Candle 2 must both fall within the open-to-close range of Candle 1. Shadows (wicks) can extend beyond — only the body matters.
3. Candle 2 is bearish. The close of Candle 2 must be below its open.
4. Size matters. Candle 2 should be noticeably smaller than Candle 1. If it is nearly the same size, the containment is borderline and the pattern is weaker.
5. Candle 2 position. The ideal Bearish Harami has Candle 2 forming near the upper portion of Candle 1’s body — meaning the gap down opened high but closed near the middle. This shows more convincing rejection.

What Backtesting Shows

Systematic backtesting of Bearish Harami across a large stock universe on daily bars reveals some important characteristics:

Win rate: In our systematic backtest of 155 Bearish Harami trades on daily bars (prior uptrend required, entry on next-day open, stop above the pattern high), the baseline win rate was 70.4% — higher than most candlestick patterns we have tested.

The pattern works best in the right context:

• Appearing at known resistance levels (prior swing highs, round numbers, Fibonacci levels)
• After extended uptrends (5+ consecutive bullish bars, or RSI sustained above 60 for several weeks)
• When confirmed by a third consecutive bearish candle (forming a Bearish Harami Cross variation)
• When accompanied by declining volume on Candle 2 (buyers are absent, not just hesitating)

The pattern degrades in these conditions:

• In strong bullish trends (EMA50 well above EMA200) — the primary trend overwhelms the pattern
• Near earnings dates — gaps and volatility distort the structure
• When the prior uptrend was steep and short (1–2 days) rather than sustained

The Gate Testing Question — Should You Add RSI or MACD Filters?

This is where backtesting reveals something counterintuitive.

The natural instinct is to add RSI and MACD gates: only take the Bearish Harami if RSI is in a specific range (say, 55–70, indicating overbought but not extreme), or only if MACD is declining. The logic sounds sound.

What the data shows:

We tested an RSI gate (requiring RSI between 55–70 at the pattern) across our full dataset of 155 Bearish Harami trades:

The RSI gate improved win rate by 2.9 percentage points. But the 87 trades it removed were net profitable — they generated more profit than they lost. Taking them out dropped absolute P&L by $6,787 despite the higher win rate.

We then tested a MACD gate separately. Same result: marginally higher win rate, lower absolute P&L. Both gates together produced the worst outcome of all configurations.

The lesson: Oscillator gates that improve win rate do so by removing trades, including both bad trades and good ones. If the good trades removed outnumber the bad, you have traded a better win rate for a worse result. Always evaluate gates on absolute P&L and profit factor, not win rate alone.

The Bearish Harami baseline (no oscillator gates) outperformed every gated configuration tested.

Practical Entry and Exit Rules

Entry:

• Identify the pattern after close of Candle 2
• Enter on the open of Candle 3 (the next session)
• Alternatively, wait for Candle 3 to open and confirm it is moving lower before entering intraday

Stop:

• Stop above the high of the entire two-candle pattern (Candle 1 high or Candle 2 high, whichever is greater, plus a small buffer)
• This is the invalidation point — if price recovers above it, the pattern has failed

Target:

• Initial target: the prior swing low before the uptrend that preceded the pattern
• Risk-to-reward should be at least 2:1 before entering. If the stop is too wide relative to the target distance, skip the trade.

Time stop:

• If price has not moved in your direction within 3–5 trading sessions, close the position. The thesis was not wrong but is not playing out — free the capital for the next setup.

Common Mistakes

Taking every Harami regardless of context. The pattern code can fire on flat price action after a single green candle. That is not a Bearish Harami in any meaningful sense. Prior trend is mandatory.

Confusing body with total candle range. The containment rule applies to the body (open to close), not the full high-to-low range. Many platforms and traders incorrectly require the shadows to be contained as well — this makes the pattern far too rare.

Entering without a defined stop. The Harami can fail, and it will fail regularly (30% of the time by the backtested data). You need a specific, pre-planned exit level before you enter.

Using it on stocks in strong primary uptrends. If a stock has been in a clean bull trend (above EMA50, EMA200 rising), a Bearish Harami is fighting the trend. The base rate is lower in this context. The pattern has a better track record at clear resistance levels, not in the middle of a range or trend.

Summary

The Bearish Harami is a legitimate, backtested reversal signal with a win rate in the high 60s to low 70s when applied correctly. Its edge comes from context — prior trend, clear resistance, and proper stop placement — not from the candle shape alone. Adding oscillator gates to filter it further tends to hurt rather than help, because the pattern’s natural selectivity (requiring prior trend) is already doing most of the filtering work.

This post is for educational purposes only and does not constitute financial advice.