Candlesticks · Two-candle patterns
Harami candlestick pattern: the inside body and what it signals
Learn to spot a bullish or bearish harami, how TA-Lib scores it (including the ±80 variant), and what 2,278 haramis on US large caps did afterwards.
On this page
Harami is an old Japanese word for "pregnant". The pattern earns the name from its outline: a long candle, the mother, followed by a much smaller candle whose real body fits entirely inside the mother's body, like a child. After a decline it is read as a bullish harami. After an advance it is a bearish harami.
Where most reversal patterns show one side overpowering the other, the harami shows something quieter: a strong move that suddenly stops. The long first candle says the trend was running at full speed. The small second candle says the next session could not extend it, and could not even trade its body outside the previous one. Western traders will recognize the close cousin, the inside day, which applies the same idea to the full high–low range.
Below you will find how the pattern is built, the precise test Finabase applies, what the ±80 score TA-Lib sometimes returns means, two cases from 2026, and a two-year data check across 60 large US stocks and ETFs.
Mother and child: the anatomy
Three features define the bullish version:
- A long first body in the direction of the prevailing decline, so red. It shows sellers in full control at the moment the pattern starts.
- A small second body whose open and close both sit between the first candle's open and close. The second session typically opens with a gap back into the mother's body and then goes nowhere.
- A preceding decline. As with every reversal pattern, the trend is what gives the shape its meaning.
The bearish version mirrors it: a long green body after a rally, then a small body inside. Textbooks prefer the child to have the opposite color to the mother, a small green inside a red body for the bullish form, but most authors accept either. When the child is a doji, so small that open and close nearly match, the pattern gets its own name: the Harami cross: a doji inside a long candle, rules and results.
The exact rules Finabase uses
Finabase runs TA-Lib's CDLHARAMI on daily bars. TA-Lib measures "long" and "small"
against the stock's own recent candles rather than fixed percentages:
| What is checked | TA-Lib default |
|---|---|
| First candle long | Real body longer than the average real body of the 10 candles before it |
| Second candle short | Real body no longer than the average real body of the 10 candles before it |
| Strict containment (±100) | Top of the second body below the top of the first, bottom above the bottom |
| Touching containment (±80) | Same, but one edge of the second body may equal the first body's edge |
| Direction | Opposite to the first candle's color: red mother +, green mother − |
| Second candle's color | Not tested |
| Prior trend | Not tested |
What the ±80 output means
TA-Lib first tries the strict test, in which the child's body must sit strictly inside the mother's. If that fails, it tries a relaxed version where the child's top may equal the mother's top, or its bottom the mother's bottom. A match on the relaxed test is scored 80 instead of 100. The relaxed form was rare in our sample: 27 of 1,181 bullish haramis and 48 of 1,097 bearish ones, about 3% of the total. In Finabase's scanner and gauge only the sign matters, so an 80 is treated exactly like a 100.
How TA-Lib departs from the textbook
The child's color is ignored. A small red body inside a big red body counts as a bullish harami. Many authors would call that a weaker case, or no pattern at all.
No trend is required. Only the mother's color sets the direction. A long red day in the middle of a rally followed by an inside day is reported as a bullish harami. Roughly 59% of the bullish signals in our sample came after a ten-session decline, so a large minority did not.
Every harami cross is also a harami. A doji body is nearly always shorter than the
average body, so it passes the "short" test too. All 739 harami-cross signals in our
sample were also flagged by CDLHARAMI, with the same sign, and they make up about a
third of all haramis.
What the pause tells you
A long red day is a strong statement by sellers. Ordinarily the next session would build on it, opening near the low and pressing further. In a bullish harami it does not. The second session opens back inside the prior body, and the whole day's trading stays inside the range that sellers already covered the session before. Nobody pushed.
That is the essence of the pattern: uncertainty after conviction. It does not tell you buyers have taken over. It tells you sellers have stopped pressing, at least for a day. Volatility often contracts after a harami, and the next sizable move, in either direction, decides what the pause meant. That is why the pattern lends itself to confirmation better than to anticipation.
Confirmation and context
- A break of the mother's range. For a bullish harami, a close above the long red candle's open, or at least above the harami pair's high. For the bearish version, the mirror. Our study uses the simpler rule of a close beyond the second candle's high (bullish) or low (bearish).
- The size of the mother matters. A harami after an exceptionally large first candle is a bigger shift in tempo than one after a merely above-average candle.
- Oversold or overbought readings. A bullish harami while the Relative Strength Index (RSI): how to read and use it sits under 30, or a bearish one above 70, at least puts the pause where a pause is plausible.
- Compare with the Doji star candlestick: a gapping doji after a long candle: if the small candle gaps outside the mother's body instead of sitting inside it, the pattern is a star, not a harami.
Two haramis from 2026
IBM, February 24, 2026 (bullish, worked). On February 23 IBM opened at $254.37 and closed at $223.35, a 13% fall from the previous close of $257.16 and a very long red mother. The next session opened at $227.80 and closed at $229.32, a small green child well inside the red body. On February 25 the stock closed at $237.54, above the harami's $236.59 high, which confirmed the pattern. Ten sessions after the harami IBM was 9.1% higher, and it had briefly traded back above its pre-fall price.
Intel, April 14, 2026 (bearish, failed). Intel had risen 55% in ten sessions. On April 13 it printed a green candle from $62.18 to $65.18. On April 14 it slipped from $65.16 to $63.81, a small red body inside the green one, and TA-Lib returned −100. The next session closed at $64.94, above the harami's $62.09 low, so the pattern was never confirmed. Over the following ten sessions Intel rose another 32.5% and touched $87.10. A pause inside a powerful trend is often just that: a pause.
What the data says about haramis
We applied CDLHARAMI to 29,760 daily bars: 60 large US stocks and ETFs over October
30, 2024 to October 29, 2026. For each signal we checked whether price had moved the
pattern's way five sessions later and compared that with the share for every day.
| Group | Cases | Went its way after 5 sessions | Any-day baseline | Average 5-session return | Average 10-session return |
|---|---|---|---|---|---|
| Bullish harami | 1,181 | 53.6% | 53.1% | +0.43% | +0.99% |
| Bullish, confirmed (from confirmation) | 359 | 62.4% | 53.1% | +0.73% | |
| Bearish harami | 1,097 | 49.9% | 46.7% | +0.02% | +0.56% |
| Bearish, confirmed (from confirmation) | 317 | 47.6% | 46.7% | +0.58% |
Returns are price changes, so a positive figure after a bearish harami means the stock rose on average.
Unconfirmed, the harami was close to noise. Bullish haramis matched the 53.1% baseline. Bearish ones were a little better than their baseline, 49.9% against 46.7%, but the average stock was still flat to higher afterwards.
Confirmed bullish haramis stood out. Measured from the close that confirmed them, 62.4% were higher a week later, nine points over the baseline, though 359 cases from one rising market is not proof. Confirmed bearish haramis did not improve at all.
One day dominated the bearish count. On April 10, 2025, the session after the market surged on the US tariff pause, 41 of the 60 names printed a bearish harami: a small red day inside the previous day's huge green candle. Those are 41 entries, but they are really one market event.
Harami versus its relatives
| Pattern | Second body relative to the first | Read as |
|---|---|---|
| Harami | Small, inside the first body | Momentum stalls |
| Harami cross: a doji inside a long candle, rules and results | A doji inside the first body | Stall with more indecision |
| Engulfing candlestick pattern: bullish and bearish rules and data | Large, covers the first body | Control changes hands |
| Doji star candlestick: a gapping doji after a long candle | Doji gapping outside the first body | Exhaustion, needs a third day |
| Inside day | Whole range (shadows too) inside the prior range | Volatility contraction |
The harami and the engulfing pattern are geometric opposites. In one the second body fits inside the first. In the other the first body fits inside the second. Their messages differ in the same way: hesitation versus takeover.
An educational outline for a harami
An outline of how the pattern is often approached, not a recommendation to trade.
- Check the mother. It should be a clearly long candle at the end of a real move.
- Treat the harami as a pause. Wait for a close beyond the small candle's range in the reversal direction.
- Set the invalidation beyond the mother's extreme. For a bullish harami, under the long red candle's low. A new low after the pause means sellers resumed.
- Size the position from that distance, which is often wide because the mother is long. That width is the practical cost of the pattern.
How Finabase shows haramis
- On a chart. Add Harami to any symbol (the IBM chart link has it on). Each flagged second candle is marked. Because harami crosses are also haramis, adding both studies marks some candles twice.
- In the scanner. The candlestick scanner sorts each detection by the sign of the output: positive goes to the Bullish list, negative to the Bearish list. The +80 and −80 variants are listed alongside the ±100 ones with no distinction.
- In the Signals tab gauge. On a symbol's Signals tab, such as IBM, the harami is one of the 61 voters in the Candlestick gauge. Its latest non-zero output within eight bars votes Buy or Sell. And since every harami cross is also a harami, a doji inside a long candle casts two votes in the same direction.
Frequently asked questions
Is a harami bullish or bearish?
Either. After a decline, with a long red first candle, it is a bullish harami. After an advance, with a long green first candle, it is bearish. In TA-Lib the first candle's color alone decides.
What is the difference between a harami and an inside bar?
A harami compares real bodies only, so the second candle's shadows may extend beyond the first body. An inside bar requires the whole high–low range to sit inside the previous range. The two often coincide, but neither guarantees the other.
Why does TA-Lib give some haramis a score of 80?
When the small body touches the edge of the big body, its open or close exactly equals the mother's open or close, TA-Lib calls the containment imperfect and scores 80 rather than 100.
Is the harami a strong reversal signal?
Not in its raw form. It marks hesitation. In our sample it only beat the baseline clearly when a bullish harami was confirmed by a later close above its high.
Summary
The harami is a long body followed by a small body tucked inside it, the picture of a trend that suddenly stops extending. TA-Lib takes its direction from the first candle, ignores the second candle's color and the prior trend, and scores touching bodies as ±80. On two years of large US names the raw pattern was close to random, while confirmed bullish haramis were higher a week later 62.4% of the time. Read it as a pause, wait for the market to show which way the pause resolves, and keep the mother's extreme as your line in the sand.
Sources and further reading
- Steve Nison, Japanese Candlestick Charting Techniques (New York Institute of Finance, 1991)
- Thomas N. Bulkowski, Encyclopedia of Candlestick Charts (Wiley, 2008)
- TA-Lib source code — CDLHARAMI and the default candle settings
- Finabase chart data — daily bars for 60 US stocks and ETFs, 2024-10-30 to 2026-10-29; patterns detected with TA-Lib CDLHARAMI