Bullish Harami — What Backtesting 160+ Trades Reveals

A systematic look at the Bullish Harami pattern across daily bars — win rate reality, optimal contexts, oscillator gate testing, and a critical lesson about config bugs in automated systems.

What is a Bullish Harami?

The Bullish Harami is the bullish counterpart to the Bearish Harami — a two-candle reversal pattern appearing at the bottom of a downtrend.

Structure:

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

The psychology: after a strong down move (Candle 1), the next session gaps up and holds, closing higher within the prior candle’s range. The sellers who drove Candle 1 are unable to push lower. The market is absorbing supply.

Like its bearish equivalent, it signals a potential reversal but requires context and confirmation to be tradeable.

Identification Rules

1. Prior downtrend required. At least 3–5 bearish candles, or a sustained downtrend. A Harami after sideways action has no reversal context.
2. Candle 2 body is inside Candle 1 body. Open and close of Candle 2 must fall within the body of Candle 1. Shadows can extend beyond.
3. Candle 2 is bullish. Close must be above open.
4. Relative size. Candle 2 should be clearly smaller — ideally 25–50% of Candle 1’s body size. Large Candle 2 reduces containment meaningfulness.
5. Position within the body. Candle 2 forming near the lower portion of Candle 1’s body is more convincing — buyers absorbed the bulk of selling pressure from the prior day.

What Backtesting 160+ Trades Shows

Our systematic backtest of 162 Bullish Harami trades on daily bars shows results that are less impressive than the Bearish Harami but still statistically meaningful:

Overall win rate: 58.6% on the baseline configuration (no oscillator gates, prior downtrend required). Lower than the Bearish Harami’s 70.4% — bullish reversals after downtrends are harder to time precisely because downtrends can be more violent and prolonged than uptrends in individual stocks.

Where it works best:

• After 5+ consecutive bearish candles (deeply oversold, clearly exhausted sellers)
• At major support levels (prior swing lows, 52-week support zones, Fibonacci retracements)
• When Candle 1 is an exceptionally long bearish candle (high range, closes on the low) — indicates maximum pessimism
• When the broader market is in a bull trend (the downtrend is a correction, not a bear market)

Where it struggles:

• After a single sharp down day — not a confirmed downtrend
• When the broader market is in a bear market — corrections can deepen further
• On stocks with fundamental problems (declining earnings, negative news) — technical patterns struggle to override fundamental deterioration

Oscillator Gate Testing — A Critical Lesson

This is where the Bullish Harami research produced its most important finding — and it involves an automated system bug that fundamentally changed the analysis.

When testing whether RSI, StochRSI, or MACD gates improve Bullish Harami performance, the process requires careful config management. During testing, the following was found in a live strategy config:

macd:
  enabled: true    # ← actual value
  # DISABLED — MACD gate also unhelpful for BuH  ← comment said this

The comment clearly documented that the MACD gate had been disabled. The actual value was true — meaning the gate was actively running. The system was behaving differently from what both the comment and the operator believed. This is a classic automated system bug: documentation and reality diverged.

The lesson for anyone running automated strategies: The value in the config file is what the system uses. Comments are what the developer intended. They can diverge, and when they do, the system wins. Always verify active parameter values from logs, not from comments.

After identifying and fixing this bug (setting enabled: false to match the intended disabled state), the clean gate-off baseline could be properly measured.

Results from clean testing:

• RSI gate: Tested at multiple ranges. In all configurations, the gate removed more profitable Bullish Harami trades than unprofitable ones. Net P&L worsened.
• StochRSI gate: Same result — improved win rate, lower absolute profit.
• MACD gate: Once correctly disabled and then re-enabled for clean testing, showed the same pattern: marginal win rate improvement, net P&L deterioration.

Conclusion: The Bullish Harami, like the Bearish Harami, performs best without oscillator gates. The pattern’s edge is in the candle structure and prior trend context, not in secondary indicator confirmation.

Why Do Gates Hurt?

This result — gates improving win rate but hurting profitability — appears consistently across multiple candlestick patterns tested. The explanation:

Oscillator gates remove trades. They remove some losing trades (that is their intended purpose) but they also remove winning trades, because oscillators are imprecise and imperfectly correlated with trade outcome. If the winning trades removed are more numerous or more profitable than the losing trades removed, the gate is net negative.

In a high-win-rate candlestick pattern context (65%+ baseline WR), the “signal” that gates try to provide is mostly noise, because the base pattern is already quite selective. Gates provide more signal value in low-win-rate patterns where the base signal is weaker and secondary confirmation is more necessary.

The implication: before adding any oscillator gate to a candlestick-based system, verify with a full backtest that it improves absolute P&L and profit factor, not just win rate.

Practical Entry and Exit Rules

Entry:

• Identify pattern after close of Candle 2
• Enter on the open of Candle 3 (next session)
• If the gap-up of Candle 2 was large, wait for Candle 3 to hold above Candle 1’s midpoint before entering — reduces risk of failed reversals

Stop:

• Stop below the low of the full two-candle pattern (Candle 1 low minus buffer)
• If the stock trades below this level, the reversal has failed — sellers remain in control

Target:

• Prior swing high before the downtrend that preceded the pattern
• Minimum 2:1 risk-to-reward required — if stop is too far from target, skip

Confirmation candle:

• The most conservative approach waits for Candle 3 to close bullish (above its open) before entering on Candle 4 open — this adds half a day of lag but filters a meaningful portion of failed patterns

Summary

The Bullish Harami is a legitimate but modestly performing reversal signal. Its win rate of 55–62% is meaningful but not exceptional — the real value comes from combining it with prior downtrend context, support levels, and proper risk-to-reward discipline. Adding RSI, StochRSI, or MACD gates consistently worsened backtested P&L by removing profitable trades alongside unprofitable ones.

The broader lesson from testing: automated config bugs can silently change system behaviour in ways that are invisible until you specifically look for them. For any automated strategy, the live config should be audited against the intended parameter values, not just against the comments.

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