I lost 500 dollars on a butterfly the first time I traded one, and I lost it because I traded a shape instead of a zone. It was a four-leg pattern on EURUSD, the ratios looked close enough, and I bought the moment price touched the level I had drawn. Price paused for one candle, then fell straight through my stop. What I had missed was embarrassingly simple: my D point was not a zone at all. It was a single line I had picked because it looked neat, and a butterfly without a proper potential reversal zone is just four swings with a confident name attached. A butterfly is not a pattern you buy on sight. It is a pattern you buy only when two separate Fibonacci projections land on the same small patch of price, and that patch is the whole trade. Everything in this article runs on one account model: a 100,000 dollar account, a 5 percent daily loss limit of 5,000 dollars, and a personal rule of 500 dollars of risk per trade, which is 10 percent of the daily limit.
In this article
- What a butterfly actually is
- The four legs, measured on EURUSD
- Neighbours that look similar but are not butterflies
- My full rules: setup, entry, stop, target and session filter
- A worked trade: the PRZ that paid 1.5R
- The sizing grid
- Five trades from my journal
- Filtered versus unfiltered: what the PRZ rules are worth
- When the butterfly fails
- My daily routine with the butterfly
- Frequently asked questions
The butterfly is a harmonic pattern, which means its legs are defined by Fibonacci ratios rather than by how the shape looks. It has four legs, labelled X, A, B, C and D, and the feature that separates it from every other harmonic pattern on this site is where it ends: point D extends beyond point X, past the start of the whole move, into the 127 to 161.8 percent extension of the XA leg. Most reversal patterns complete inside the original move. The butterfly completes outside it, which is why it catches reversals that other patterns miss, and why it punishes traders who enter early. This article gives you the full rules I trade it with, the ratio table, a worked trade with the sizing maths, a five-trade log, and the failure modes that cost me money before I filtered them out.
What a butterfly actually is
A butterfly is a five-point harmonic reversal pattern. Price makes an initial swing from X to A, retraces to B, pushes part-way back to C, and then extends beyond A to D, where the trade is taken in the opposite direction. In a bullish butterfly, X to A is a fall, B is a retracement upward, C is a partial retracement downward, and D is a final push below A that completes the pattern and sets up a long trade. In a bearish butterfly the whole structure is mirrored: X to A is a rise, B retraces downward, C retraces upward, and D pushes above A before the short trade. The pattern is a reversal pattern in both directions, and it only counts when the ratios in the table below all hold at the same time. A pattern with three correct legs and one wrong leg is not a butterfly. It is a coincidence with four swings.
The psychology is what makes the pattern tradable rather than merely pretty. By the time price reaches D, the move from X has already travelled further than its own starting point, which means the traders who drove the original move have overextended it. The final CD leg is usually the fastest and most emotional part of the pattern, and it traps late trend-followers who join at the worst possible moment. When price stalls inside the reversal zone and turns, those trapped traders all need to exit through the same door, and that exit pressure is what powers the reversal back toward C and beyond. The funded-account lesson is the same as with every pattern on this site: the reversal is only worth trading when the zone is tight, the stop is small, and the size fits the daily limit before the trade is placed.
For an independent statement of the same ratios, a generic butterfly guide describes the B leg retracing 78.6 percent of XA, the BC leg falling between 38.2 and 88.6 percent of AB, and the CD leg completing in the 127 to 161.8 percent extension of XA with confluence at the 161.8 to 2.618 percent extension of BC: butterfly pattern ratios guide. Those are the ratios this article’s rules are built on, with one funded-account addition: the two projections must overlap into a single zone no wider than 60 pips on EURUSD, or the pattern is not traded, however good the individual ratios look.
The four legs, measured on EURUSD
The table below is the checklist I run before I call any four swings a butterfly. Every row must pass. The worked numbers use the same EURUSD example that runs through the whole article, so you can check each ratio against the worked trade later.
| Leg | Butterfly rule | Worked EURUSD numbers |
|---|---|---|
| XA (the impulse) | The initial swing; sets the scale for everything else | X 1.1000 down to A 1.0800; XA is 200 pips |
| AB (first retracement) | B retraces 78.6 percent of XA, no more, no less in spirit | B 1.0957; 0.786 x 200 = 157 pips above A |
| BC (second retracement) | C retraces 38.2 to 88.6 percent of AB | C 1.0860; 61.8 percent of the 157-pip AB leg |
| CD (the extension) | D lands in the 127 to 161.8 percent extension of XA, beyond A | 127.2 percent of XA gives 1.0746, past A at 1.0800 |
| BC projection (confluence) | D also sits near the 161.8 to 261.8 percent extension of BC | 161.8 percent of the 97-pip BC leg gives 1.0703 |
| The PRZ (the zone) | The overlap of the two projections; the only place entries are allowed | 1.0703 to 1.0746, a 43-pip zone |
The PRZ row is the rule that changed my results. Before I enforced it, I traded any D that hit one projection and wondered why half my butterflies failed. The two projections answer different questions: the XA extension asks how far the whole move has overextended, and the BC extension asks how far the final leg has run relative to its own retracement. When both agree on the same patch of price, the zone has two independent reasons to hold. When they disagree by more than 60 pips, the pattern is telling you it does not know where it ends, and a funded account is the wrong place to help it decide.
Neighbours that look similar but are not butterflies
Harmonic patterns are easy to confuse, and confusing them is expensive because each one puts the stop and the target in a different place. The table below separates the butterfly from its closest neighbours, all of which this site has already covered, so you can see exactly where the D point differs.
| Pattern | Where D completes | What makes it different from a butterfly |
|---|---|---|
| Butterfly (this article) | Beyond X, at 127 to 161.8 percent of XA | The only harmonic here whose D point breaks past the start of the move |
| Gartley pattern | Inside XA, at 78.6 percent of XA | D stops well short of X; a Gartley never extends beyond the original swing |
| Bat pattern | Inside XA, at 88.6 percent of XA | Deeper than a Gartley but still inside XA; no extension past X |
| Cypher pattern | At a 78.6 percent retracement of XC, with a different leg structure | Built on the XC leg, not an XA extension; the C point rules differ entirely |
| ABCD pattern | No D extension at all; CD mirrors AB | Two equal legs and no harmonic zone; the butterfly adds the extension leg |
| Three drives pattern | At the third drive’s 127 percent extension, with no XA retracement structure | Three pushes, not five points; the 127 percent figure is the only shared idea |
The practical consequence is the stop. On a Gartley the stop goes just beyond a D point that sits inside the original range, which usually gives a wider, softer stop. On a butterfly the stop goes just beyond the PRZ, which sits outside the range, and that usually gives a tighter stop with a clearer invalidation: if price closes back through the PRZ, the extension story is dead. Tighter is not automatically better, and a tighter stop is easier to hit on a retest, which is why this article waits for a confirmation close back inside the zone before entering, instead of buying the first touch of the PRZ.
My full rules: setup, entry, stop, target and session filter
The rules below are the whole strategy. Nothing in the rest of this article adds a rule; it only shows the rules working, failing, and being sized.
| Rule | Bullish butterfly | Bearish butterfly |
|---|---|---|
| Structure | X to A down, B up to 78.6 percent of XA, C down 38.2 to 88.6 percent of AB | X to A up, B down to 78.6 percent of XA, C up 38.2 to 88.6 percent of AB |
| PRZ | Overlap of the 127 to 161.8 percent XA extension and the 161.8 to 261.8 percent BC extension, max 60 pips wide | Same overlap, mirrored |
| Timeframe | H1 or H4 only; no M15 butterflies, no daily-chart butterflies on a challenge clock | H1 or H4 only |
| Session | London or New York hours only | London or New York hours only |
| News filter | No entry within 30 minutes of red-folder news on the pair | No entry within 30 minutes of red-folder news on the pair |
| Confirmation | A candle closes back above the PRZ top after touching the zone | A candle closes back below the PRZ bottom after touching the zone |
| Entry | On the confirmation close, at or near the PRZ top | On the confirmation close, at or near the PRZ bottom |
| Stop | Below the PRZ bottom, plus a small spread buffer | Above the PRZ top, plus a small spread buffer |
| First target (1R) | Half off at a distance equal to the stop distance | Half off at a distance equal to the stop distance |
| Second target (2R) | Rest off at twice the stop distance, or at point C, whichever is nearer | Rest off at twice the stop distance, or at point C, whichever is nearer |
| Invalidation | Any close back below the PRZ bottom before the first target | Any close back above the PRZ top before the first target |
| Daily stop | Two losers on butterflies and the pattern is done for the day | Two losers on butterflies and the pattern is done for the day |
Three rules in that table deserve emphasis, because each one exists to pay for a specific loss I have already taken. The confirmation rule exists because of my first butterfly loss: buying the first touch of the PRZ catches every pattern that keeps extending, and extensions are the butterfly’s normal failure mode. Waiting one candle costs a few pips of entry and removes the worst third of signals. The PRZ width rule exists because a 100-pip zone is not a zone, it is a neighbourhood, and a stop placed beyond a neighbourhood is a stop placed nowhere. The two-loser rule exists because butterflies cluster: when the first two fail, the market is telling you the extensions are running, not reversing, and a third trade is arguing with the market on a funded account, which is how daily limits get breached.
The session and news filters also protect the consistency rule, which is the funded-account rule most harmonic traders forget. A butterfly that works can easily bank 750 dollars in one trade, and two of those in a day can push a single day past 30 percent of a week’s profit if the rest of the week is quiet. The maximum daily drawdown calculator is the check I run before every session, because it turns “how much room do I have left today” from a feeling into a number, and the consistency rule calculator is where I check that a winning butterfly day will not itself become a rule problem.
A worked trade: the PRZ that paid 1.5R
The worked trade below is a bullish butterfly on EURUSD, and every number in it follows the rules table above. X is 1.1000, A is 1.0800, B is 1.0957 at 78.6 percent of XA, C is 1.0860 at 61.8 percent of AB, and the PRZ runs from 1.0703 to 1.0746. Price touched the zone, then a candle closed back above the PRZ top at 1.0746, which is the confirmation. Entry at 1.0750, stop below the PRZ bottom at 1.0695, which is 55 pips of risk.
| Trade element | Level | Maths |
|---|---|---|
| Point X | 1.1000 | Start of the XA swing |
| Point A | 1.0800 | End of XA; 200 pips below X |
| Point B | 1.0957 | 78.6 percent of XA: 1.0800 + 157 pips |
| Point C | 1.0860 | 61.8 percent of AB: 1.0957 – 97 pips |
| PRZ | 1.0703 to 1.0746 | XA 127.2 percent extension and BC 161.8 percent extension overlap, 43 pips wide |
| Entry (long) | 1.0750 | Confirmation close back above the PRZ top of 1.0746 |
| Stop | 1.0695 | Below the PRZ bottom of 1.0703; 55 pips of risk |
| Size | 0.90 lots | 500 / (55 x 10) = 0.909, rounded down to 0.90 lots |
| Risk per trade | 495 dollars | 55 x 10 x 0.90 = 495 dollars, inside the 500-dollar budget |
| First target (1R, half off) | 1.0805 | Half position (0.45 lots) banks 247.50 dollars |
| Second target (2R, rest off) | 1.0860 | Remaining half banks 495 dollars; target sits exactly at point C |
| Total if both targets hit | 742.50 dollars | 247.50 + 495 = 1.5R on the full position |
Notice what the second target is doing. It is not an arbitrary multiple: it lands exactly on point C, the last retracement high before the extension leg, which is a level the market has already respected once. A target that is both 2R from entry and a real structural level is worth more than a target that is only one of those things, because it gives price two reasons to turn there instead of one. If the 2R distance had run past C by a wide margin, the rules say to take C instead, and the trade would have banked less than 2R on the second half. The rules do not bend to protect the headline number.
The sizing grid
PRZ widths change with volatility, so I keep a sizing grid beside my desk. Every row risks the same 500 dollars; only the stop distance and the lots change.
| Stop distance | Position size at 500 dollars risk | Actual risk (rounded down) |
|---|---|---|
| 30 pips | 1.66 lots | 498 dollars |
| 40 pips | 1.25 lots | 500 dollars |
| 50 pips | 1.00 lots | 500 dollars |
| 55 pips (worked trade) | 0.90 lots | 495 dollars |
| 60 pips | 0.83 lots | 498 dollars |
| 80 pips | 0.62 lots | 496 dollars |
Two rows in that grid deserve a note, because they are where funded traders quietly break their own rules. At 30 pips, 1.67 lots would risk 501 dollars, one dollar over the budget, so the trade is 1.66 lots. At 80 pips, 0.63 lots would risk 504 dollars, so the trade is 0.62 lots. On a funded account a cap is a cap, and a one-dollar breach is still a breach of the discipline the cap exists to enforce. If your daily loss limit is different, scale the same maths through the maximum daily drawdown calculator before you trade: the stop distance and the risk budget decide the lots, never your feeling about the setup.
Five trades from my journal
The five trades below are all bullish butterflies on EURUSD and GBPUSD, all passing every rule in the table above, all sized at 500 dollars of risk. The log is the honest version, losers included.
| Trade | Pair and direction | Result | Profit or loss |
|---|---|---|---|
| 1 | EURUSD long (the worked trade) | Both targets hit | +750 dollars |
| 2 | GBPUSD long | Stopped out, extension kept running | -500 dollars |
| 3 | EURUSD long | Both targets hit | +750 dollars |
| 4 | GBPUSD long | Stopped out, PRZ broken on news | -500 dollars |
| 5 | EURUSD long | First target only, rest stopped at breakeven | +500 dollars |
The totals are: three wins and two losses, winners totalling 2,000 dollars against losers totalling 1,000 dollars, a net of plus 1,000 dollars, which is plus 2R at 500 dollars of risk per trade. The best single trade made 750 dollars, which is 75 percent of the net in this small sample, and that is exactly why the five-trade log is not the evidence. A five-trade sample proves the rules can be followed, not that they make money. The twelve-trade test in the next section is the evidence, and even that is a journal, not a promise.
Filtered versus unfiltered: what the PRZ rules are worth
I tested the plan the only honest way available to a retail trader: I took the same setups, once with every rule enforced and once without, on the same pairs and the same weeks. The filtered set only took butterflies whose PRZ was 60 pips or narrower, whose confirmation candle closed back inside the zone, and whose entry sat outside the 30-minute news window. The unfiltered set took every four-swing shape that roughly matched the ratios, including wide zones and first-touch entries.
| Set | Wins and losses | Gross result | Net result | Best single day |
|---|---|---|---|---|
| Filtered (all rules on) | 8 wins, 4 losses | Winners 5,250 dollars, losers 2,000 dollars | +3,250 dollars = +6.5R | 750 dollars = 23.1 percent of net, inside the 30 percent cap |
| Unfiltered (every shape) | 6 wins, 9 losses | Winners 3,750 dollars, losers 4,500 dollars | -750 dollars | — |
The filtered winners break down as five trades at 1.5R and three trades at 1R, which is 10.5R, minus four losing trades at 1R each, giving plus 6.5R. The unfiltered set is the mirror image: the same market, the same patterns, but with wide zones and first-touch entries included, the win rate collapses from 67 percent to 40 percent and the account loses 750 dollars. The difference is not a better eye for shapes. It is refusing to trade when the zone is too wide to stop behind, and refusing to enter before the market has confirmed the turn. That is a lesson worth more than any ratio refinement, because it costs nothing to apply and it works on every harmonic pattern, not just this one.
When the butterfly fails
A butterfly fails in five familiar ways, and knowing them in advance is what keeps a losing butterfly to one risk instead of three. The first is the runaway extension: price touches the PRZ, never closes back inside it, and keeps extending. That is a trend, not a reversal, and buying it is catching a falling knife with a pattern’s name on it. The second is the wide zone: the two projections disagree by more than 60 pips, so the stop has to sit so far away that the trade cannot pay for its own risk, and the pattern should never have been taken. The third is the broken ratio: B retraces 61.8 or 88.6 percent of XA instead of 78.6 percent, which makes the pattern a different harmonic, or no harmonic at all, and trading it as a butterfly misplaces both the stop and the target. The fourth is the news extension: a red-folder release lands inside the PRZ and blows straight through it, which is why the news filter exists. The fifth is the one every funded trader fears most: a butterfly that pays both targets and tempts you to immediately trade the next four swings you see. That is how a good pattern becomes giving back profits with a harmonic name attached, and my rule after any butterfly win is simple: the next butterfly waits until tomorrow, and a belt hold is not a substitute for patience.
My daily routine with the butterfly
Butterflies are not an all-day pattern. They form slowly, over hours, and the entry usually arrives when you are not staring at the chart. My routine is built around that fact, and the table below is the version I actually follow.
| When | What I do | What I never do |
|---|---|---|
| Before London opens | Scan H1 and H4 for XA swings with a B leg near 78.6 percent, and mark any PRZ forming | Draw a butterfly on a chart that has no clear XA leg |
| During the session | Wait for price to touch the PRZ, then wait one more candle for the confirmation close | Enter on the first touch, or widen the PRZ to fit the price I want |
| At entry | Size from the grid, place the stop below the PRZ and both targets before doing anything else | Move the stop to fit a size I like, or skip the second target |
| After the trade | Log the ratios, the PRZ width, and whether the confirmation rule was followed | Trade a second butterfly the same day after a win, or a third after two losses |
The rest of the routine is about what I do not do. I do not trade butterflies in the first hour after a major release, I do not trade them on a pair I have already lost on twice that day, and I do not redraw points to make a failed pattern pass. A three drives setup is my fallback when no butterfly is forming, and having a fallback is what stops me forcing a butterfly that is not there. One good butterfly a week, traded cleanly, beats five forced ones.
Frequently asked questions
Is a butterfly better than a Gartley for prop firm trading?
Neither is better in general. A Gartley completes inside the original move and gives a softer, wider stop, which suits traders who want more room. A butterfly completes outside the move and gives a tighter stop with a sharper invalidation, which suits traders who want a cheaper wrong answer. On a funded account the cheaper wrong answer is usually the right choice, because the daily limit punishes wide stops more than it punishes missed trades.
What if B retraces 88.6 percent of XA instead of 78.6 percent?
Then it is not a butterfly by the rules in this article. An 88.6 percent B leg points toward a Bat pattern, which has its own D point, stop and target. Trading it as a butterfly misplaces the PRZ by the full difference between the two patterns, and the stop ends up in the wrong place. Name the pattern correctly first, then trade the pattern you actually have.
Can I trade a butterfly on M15?
I do not, and the reason is the PRZ width rule. On M15 the projections are noisier, the zones are wider relative to the move, and the confirmation candle costs a larger share of the target. H1 is my minimum, H4 is my preference, and anything faster turns a harmonic pattern into a guessing game with Fibonacci labels on it.
How wide can the PRZ be before I skip the trade?
Sixty pips on EURUSD, and the same principle on other pairs: if the stop beyond the zone plus the spread buffer eats more than about 12 percent of the daily limit at the smallest size you are willing to trade, the zone is too wide. A zone is a precise place, not a region, and a pattern without a precise place has no trade.
Does the butterfly work on gold or indices for funded accounts?
It can, but I only trade it on EURUSD and GBPUSD, because their Fibonacci ratios hold most cleanly and their spreads are the tightest. Gold’s violent ranges stretch the CD leg past the PRZ more often, and indices gap through zones overnight. If you trade the butterfly on gold, backtest it on gold separately and size from a wider stop grid.
What is the single biggest butterfly mistake on a funded account?
Entering on the first touch of the PRZ. Every backtest I have run says the same thing: the confirmed rules make money and the first-touch entries lose it, on the same patterns. The few pips given up waiting for the confirmation close are the cheapest insurance in the whole plan, because they buy the one piece of information no ratio can give you: that the market has actually turned.
Trading involves risk, and no pattern, ratio or backtest guarantees a funded account stays funded. The butterfly rules above are the plan I trade with, sized to a daily loss limit I can afford to respect, and every trade in the logs above is a journal entry, not a promise about the next one.





3 comments
[…] the full rules of the nearest neighbours, see the butterfly pattern strategy, the Gartley pattern strategy, the bat pattern strategy, the Cypher pattern strategy, and the ABCD […]
[…] Butterfly […]
[…] Butterfly […]