Pattern Detail

Butterfly

An extension pattern that completes beyond the origin, fading exhaustion as price prints a fresh extreme.

X A B C D
  • B = 0.786 of XA
  • C = 0.382-0.886 of AB
  • D = 1.618-2.24 of BC
  • D = 1.272-1.618 of XA
Idealized bullish Butterfly. It completes at a low and projects a move up.

Bullish Butterfly

i

How to read this

Everything here is in R, the setup's own risk. 1R is the distance from the entry (the pattern's closing price) to where it would be proven wrong — its lowest low over the 4 bars that form it. So "offered 2R" means price ran twice that distance in your favor at some point before the stop. It does not assume you took profit there: a target is a strategy choice.

Room offered (≥ 1R)

66.7%

Too few to trust

Offered at least 1× its risk before the stop, vs 44.1% for a random long entry (+22.6 pts).

Move size vs normal

1.00×

Realized range over the next 20 bars vs a random bar. About normal.

Typical room (20-bar)

1.24R

Average run in favor (capped at 3R), vs 1.16R for a random long entry.

Summary

Offered ≥1R 66.7% of the time vs 44.1% for a random long entry. The 22.6-point gap is no bigger than the ±32.4-point margin of error you would get by chance from 9 occurrences. Not a reliable edge.

Room offered, this setup vs a random long entry

Only 9 occurrences. The breakdown below is shown in full, but a sample this small is anecdotal, a hint, not a measured edge. That is usually a limit of available history, not a flaw in the pattern. For a firmer read, try a lower timeframe or a more active instrument.

Outcome This setup Random entry Edge
Offered ≥ 1R 66.7% 44.1% +22.6
Offered ≥ 2R 22.2% 24.7% -2.4
Offered ≥ 3R 22.2% 15.7% +6.5
Stopped < 1R 22.2% 42.7% -20.5
Went sideways 11.1% 13.2% -2.1

9 occurrences · 10,656 random-entry controls · 20-bar horizon

Bearish Butterfly

i

How to read this

Everything here is in R, the setup's own risk. 1R is the distance from the entry (the pattern's closing price) to where it would be proven wrong — its highest high over the 4 bars that form it. So "offered 2R" means price ran twice that distance in your favor at some point before the stop. It does not assume you took profit there: a target is a strategy choice.

Room offered (≥ 1R)

27.8%

Too few to trust

Offered at least 1× its risk before the stop, vs 43.0% for a random short entry (-15.3 pts).

Move size vs normal

1.00×

Realized range over the next 20 bars vs a random bar. About normal.

Typical room (20-bar)

0.72R

Average run in favor (capped at 3R), vs 1.16R for a random short entry.

Summary

Offered ≥1R 27.8% of the time vs 43.0% for a random short entry. The 15.3-point gap is no bigger than the ±22.9-point margin of error you would get by chance from 18 occurrences. Not a reliable edge.

Room offered, this setup vs a random short entry

Only 18 occurrences. The breakdown below is shown in full, but a sample this small is anecdotal, a hint, not a measured edge. That is usually a limit of available history, not a flaw in the pattern. For a firmer read, try a lower timeframe or a more active instrument.

Outcome This setup Random entry Edge
Offered ≥ 1R 27.8% 43.0% -15.3
Offered ≥ 2R 5.6% 25.9% -20.3
Offered ≥ 3R 0.0% 17.0% -17.0
Stopped < 1R 55.6% 47.1% +8.5
Went sideways 16.7% 9.9% +6.8

18 occurrences · 10,573 random-entry controls · 20-bar horizon

The Butterfly, credited to Bryce Gilmore and Larry Pesavento, is an extension pattern. It is a five-point structure, X to A to B to C to D, defined by Fibonacci ratios between the legs. The signature is the completion: D extends to a 1.272 to 1.618 extension of the whole X-to-A leg, beyond the origin X.

How to spot it

  • A clear X-to-A trend leg, then a sequence of swings that pull back against it.
  • B retraces 0.786 of XA, C retraces 0.382 to 0.886 of AB, and D projects 1.618 to 2.24 of BC.
  • D lands at a 1.272 to 1.618 extension of XA, beyond the origin X. All four ratios hold at once.
  • The structure completes at D, a swing low in an uptrend (a buy) or a swing high in a downtrend (a sell).

The chart above marks the X, A, B, C, D points on a real occurrence, with the ratio of each leg.

Why it matters

A Butterfly is a bet on exhaustion at a fresh extreme. The X-to-A move is the trend leg, and the swings after it stretch past the prior origin into a measured extension, where the move is expected to run out of buyers or sellers and snap back. Because D pushes beyond X, the Butterfly fades an overstretched move rather than buying a dip, and the entry leans against the new extreme rather than joining it.

That places it among the extension harmonics, alongside the Crab, where the completion sits beyond the origin. The retracement harmonics, the Gartley and Bat, are the opposite case: D stays short of X, so they buy a dip inside the prior move rather than fade a new high or low. The Butterfly is the counter-trend member of the family, which is also where it is dangerous: a Butterfly that keeps extending becomes a trend the position is now against, so the stop matters.

The honest question is not whether the shape can be drawn on a chart, it always can after the fact, but whether a completion actually leads anywhere. That is what the data below measures.

Does it actually work?

A pattern is a setup, not a trade, so the question is not “did it win” but “how much room did the move offer before the structure was proven wrong.” The tabs below answer that across five futures markets (Nasdaq, S&P 500, gold, crude oil, natural gas) and seven timeframes from one minute to one day.

For each completion we measure the room price offered in units of the pattern’s own risk, then set it against what a random swing pivot of the same kind would have done. A Butterfly that turns up from a swing low is judged against ordinary swing lows; one that turns down from a swing high, against ordinary swing highs. When the completion offers more room more often than those random swings, that shows up as a real edge. When it does not, the page says so plainly.

Harmonic structures are rare, so read the sample size in view. On the slower timeframes a Butterfly completes only single digits of times, or not at all, while the faster timeframes give a larger sample. Thin samples are flagged for you.

How we measured it

  • Entry is the close of the bar that confirms the D point, the earliest a trader could act without hindsight.
  • One unit of risk, 1R, is the distance from that close to D itself, the swing extreme the structure completes at. A move back through D says the pattern has failed.
  • We follow the next 20 bars and record how far price ran in your favor, in multiples of that risk, before D gave way.
  • The comparison is a random swing pivot of the same side, which already sits at a turning point with a tight stop. That keeps it apples-to-apples: the question is whether a Butterfly swing runs further than an ordinary swing of the same shape, not whether a tight stop flatters the numbers.

What this page does not cover

  • A profit target. Harmonic tradition puts one at a Fibonacci retracement, but where you take profit is a strategy choice, so this measures only the room the structure tends to offer.
  • A trend filter, or confluence with support, resistance, or a higher-timeframe level, which traders normally use to weight a completion.
  • A guarantee. A valid shape on the chart is a setup, not a certainty. What it tends to do next is exactly what the numbers above describe.

FAQ

Do harmonic patterns like the Butterfly actually work?

Harmonic patterns are described everywhere but rarely tested honestly. This page scores every completed Butterfly by the room it offered in units of its own risk, measured against random swing pivots of the same kind, and counts a pattern only on the bar its final pivot confirms, so nothing is read in hindsight. The result is a like-for-like comparison rather than the round-number success rates usually quoted for harmonics.

What is the Butterfly pattern success rate?

There is no single reliable figure, and the often-cited round-number rates rarely specify instrument, timeframe, or test period, and almost never state what target and stop they assume. Rather than quote folklore, this page measures how far a completion runs in units of its own risk and whether that beats an ordinary swing of the same shape. The tabs show that comparison across five markets and seven timeframes.

How is the Butterfly defined here?

It is a five-point XABCD structure where B retraces 0.786 of XA, C retraces 0.382 to 0.886 of AB, D projects 1.618 to 2.24 of BC, and D extends to 1.272 to 1.618 of XA, beyond the origin, all holding at once within a 5% tolerance. Pivots use 3 bars on each side and are confirmed before they count.

Sample Butterfly Completions (20)

Based on data through Apr 29, 2026

Time Direction Risk (pts) Room offered Result
Apr 9, 2026 Bearish 69.25 0.74R Stopped
Feb 15, 2024 Bearish 84.25 0.46R Stopped
Mar 28, 2023 Bullish 38.25 3.00R Ran ≥1R
Jan 16, 2023 Bearish 20 1.95R Ran ≥1R
Apr 7, 2022 Bullish 126.75 1.50R Ran ≥1R
Mar 28, 2022 Bearish 83 1.55R Ran ≥1R
Aug 27, 2021 Bearish 30 0.58R Stopped
Sep 9, 2020 Bullish 207.5 1.05R Ran ≥1R
Aug 21, 2020 Bearish 19 0.00R Stopped
Jul 23, 2019 Bearish 39.25 0.39R Flat
Jul 9, 2019 Bearish 18.5 0.04R Stopped
Dec 26, 2018 Bearish 20 0.00R Stopped
Sep 2, 2015 Bearish 21.25 0.36R Stopped
Feb 17, 2015 Bearish 9.5 0.50R Flat
Sep 13, 2013 Bullish 9.25 3.00R Ran ≥1R
Mar 13, 2013 Bearish 4.25 2.06R Ran ≥1R
Jan 29, 2013 Bullish 12.5 1.26R Ran ≥1R
Feb 9, 2011 Bearish 4 0.44R Stopped
Aug 11, 2010 Bullish 6.25 0.04R Stopped
Oct 19, 2009 Bearish 2.75 1.73R Ran ≥1R

Backtest this pattern

Run it on your contracts, timeframes, and dates.