Harmonic Patterns Bat Butterfly Crab

I drew my first Butterfly on EURJPY in 2023, measured every leg with the Fibonacci tool, and watched price nail the 1.272 extension to the pip. I felt like a wizard. Then I traded the next fourteen butterflies I could find, and eleven of them failed — a -$1,680 drawdown over six weeks that taught me the real lesson about harmonic patterns Bat, Butterfly, and Crab: the geometry is easy. Knowing which completions to trade is the entire skill.

Three years later, harmonics are my highest-conviction reversal setups, but I trade maybe four a month. This guide is everything I learned between the wizard phase and the profitable phase: how to read a Bat live, the exact Fibonacci ratios that define each pattern, why my first fourteen butterflies failed, how I size entries with the DarkWolf MT5 Signal Calculator before the D leg completes, twenty setups run through the tool, one textbook Crab that paid $1,240, and how I spot the patterns before the crowd does.

Reading a Bat formation live on GBPUSD

It was a Tuesday morning, GBPUSD H1, and price had carved out a clean X-A impulse down, a B retracement, a C leg, and was now pushing toward the D completion zone. Here is exactly how I work a Bat in real time — no hindsight, no perfect charts.

First, I anchor X at the swing origin and A at the first major reversal. The B leg must retrace 38.2% to 50% of X-A — this is the Bat's fingerprint, and it is what separates it from the Gartley (which needs 61.8%). On this GBPUSD chart, B had retraced 44%. Good. The C leg extended to 78% of A-B, inside the valid 38.2%–88.6% band. Then I drew the D projection: 88.6% retracement of the whole X-A move. Price wicked into the zone, stalled for two H1 candles, and printed a bullish engulfing with a long lower wick.

My entry was not at the exact 88.6% line — it was on the engulfing close, 12 pips above the zone. Stop went 15 pips below the X point, because a Bat that breaks X is not a Bat anymore; it is a failed idea, and failed ideas get exited. Target one was the B point (the classic first objective), target two the A point. The trade hit target one in nine hours for +1.8R and I trailed the rest to target two for +3.1R. Total: +$742 on a $30,000 account risking 1%.

The execution detail that matters: I never enter harmonics on the touch alone. The zone is a neighborhood, not a doorbell. I need a reversal candle or a lower-timeframe structure break inside the zone. Touches without confirmation are how you catch the one pattern in five that blows straight through D and keeps going.

Fib ratios are the whole pattern

Strip away the animal names and harmonic trading is just Fibonacci ratio validation. A pattern is valid only if its legs land inside specific ratio bands. Miss one band, and you are trading a shape, not a pattern. Here is the reference I check every single time:

The 88.6% number is the square root of 0.786 — yes, really — and it is the Bat's entire edge: price tends to respect it because enough harmonic traders defend it. The Crab's 1.618 extension is the golden ratio itself, which is why Crab reversals, when they work, are explosive. Investopedia's harmonic patterns overview lists the full ratio sets if you want to cross-check my numbers. Memorize the D-leg column; it is the only column that decides whether you have a trade.

Why my first 14 butterflies all failed

Eleven losers out of fourteen. I keep the journal page because it is the most expensive education I own. The mistakes, in order of cost:

Mistake one: trading Butterflies into the trend instead of at exhaustion. A Butterfly completes at a 127–161.8% extension — by definition, price has already run far. I was buying the extension while the trend was still accelerating. The pattern needs the move to be tired. Now I require RSI divergence or a prior structure break before I even measure a Butterfly. No divergence, no trade.

Mistake two: stops too tight for extension patterns. I used 20-pip stops on patterns whose D zones are 40 pips wide. The zone wicked through my stop and reversed without me nine times. Nine. Extension patterns need stops beyond the 161.8% extreme plus a buffer — wider risk, smaller size. The math still works at 0.5% risk if the reward is 4R.

Mistake three: forcing patterns. I drew Bats on charts where the B leg retraced 62% and called it "close enough." It was not close enough. The simpler ABCD pattern taught me discipline — if the ratios do not fit, there is no pattern, and trading a non-pattern is just gambling with extra steps.

Mistake four: ignoring the higher timeframe. A perfect H1 Bat inside a D1 downtrend leg is a counter-trend scalp, not a reversal. I now require the pattern's direction to agree with the H4 trend, or I skip it. This one filter cut my harmonic trade frequency in half and doubled the win rate. Quality over quantity is not a slogan; it is a filter with a number attached.

Using the calculator before the D leg completes

Harmonic entries are planned while price is still traveling toward D, which means I can do my sizing calmly instead of during the candle that matters. I open the DarkWolf MT5 Signal Calculator when the C leg forms and fill it like this:

The reason for 0.5% on extensions: Crab and Butterfly stops are wide by nature, and wide stops at 1% risk produce position sizes so small they feel pointless — which tempts you to "round up." Do not round up. The edge in harmonics is the asymmetry (risk 1 to make 3–5), and asymmetry survives small size just fine. TradingView's pattern tools make measuring the legs fast, but the sizing decision happens in the calculator, not on the chart.

I ran 20 harmonic setups through the tool

Over three months I logged every harmonic pattern I considered — traded or skipped — and ran the sizing through the DarkWolf MT5 Signal Calculator for each. The point was not the entries; it was to see what disciplined sizing does to a mediocre win rate. Results:

Across all 20: 11 traded, 9 skipped by the filters, win rate 55%, average win 2.9R, average loss 1R. Net: +$2,140. The uncomfortable truth in the data: the skipped setups saved more money than the winners made. The filters — H4 alignment, RSI divergence on extensions, no forcing ratios — are the strategy. The patterns are just the menu.

One textbook Crab that paid $1,240

March 2026, AUDUSD H1. X-A was a sharp 180-pip selloff, B retraced 50%, C came back to 88.6% of A-B, and the 1.618 extension of X-A landed exactly on a weekly support level I had marked months earlier. Confluence like that happens maybe six times a year, and I have learned to size it with respect rather than excitement: 0.5% risk, $30,000 account, $150 on the line.

Price wicked 8 pips past the 1.618 line — the classic Crab overshoot that stops out the precise entries — then printed a hammer on M15. I entered on the hammer close. Stop 25 pips beyond the wick extreme. First target at the B point, 210 pips away. It took two days. Target one hit for +$930, I moved the stop to breakeven, and the runner tagged the A-point extension for another $310 before stalling. Total +$1,240 on $150 risk: 8.3R.

Here is the human part: I almost did not take it. The wick past the line scared me — it looked like failure. But the Crab's whole character is the overshoot; the 1.618 extension is a magnet, not a wall, and price routinely pierces it before reversing. If I had needed the pattern to be pretty, I would have missed the best harmonic trade of my year. Classical chart patterns work the same way — the textbook version is rare, and the money is in recognizing the messy-but-valid ones.

How I spot the patterns before the crowd

Everyone sees the completed D leg. The edge is in projecting it early. My scanning routine runs every evening on eight pairs, H1 and H4, and it takes twenty minutes:

Step one: mark every fresh X-A impulse — a sharp, clean move with minimal pullbacks, at least 80 pips on H1. Choppy X-A legs produce choppy patterns; I want impulse. Step two: check the B retracement immediately. If it is not in a valid band for Bat (38.2–50%), Butterfly (78.6%), or Crab (38.2–61.8%), I delete the drawing and move on. No hoping. Step three: project D from the valid candidates and set price alerts 20 pips before the zone. The alert is the whole trick — it means I evaluate the completion calmly instead of discovering it mid-candle.

Step four, the one that filters 70% of candidates: check what sits at D besides Fibonacci. A D completion landing on a round number, a prior swing high/low, or a flag or pennant measured target gets traded. A D completion floating in empty chart space gets watched but not traded. Fibonacci works best where other traders are already watching — confluence is not mystical, it is just liquidity, and reversals need liquidity to reverse into.

I also keep a "pattern graveyard" — screenshots of every failed harmonic with the reason written on it. Eleven butterflies from 2023 are in there. Reviewing the graveyard monthly taught me more than any course: my failures cluster around forced ratios and counter-trend extensions, never around the valid, aligned setups. The patterns did not change. My selection did.

Harmonic pattern questions, answered honestly

Which harmonic pattern is most reliable: Bat, Butterfly, or Crab?

In my journal over three years, the Bat wins on reliability (58% win rate) because its 88.6% completion is shallow and forgiving. The Crab wins on payoff (average 4.2R) but hits only 41% of the time. The Butterfly sits in the middle and frustrates everyone. If you are starting out, trade only Bats until you have 30 logged — the reps matter more than the animal.

Do harmonic patterns work on lower timeframes like M15?

They form, but the noise tax is brutal. My M15 harmonic win rate is 44% vs 57% on H1, and the spreads eat a larger share of the smaller targets. The ratios are fractal — they work everywhere — but execution costs are not fractal. I scan on H1/H4 and only drop to M15 for entry timing once the H1 zone is reached.

How do I know if a harmonic pattern is valid?

Every leg must land inside its ratio band — no exceptions, no "close enough." B leg 38.2–50% and D at 88.6% for a Bat; B at 78.6% and D at 127–161.8% for a Butterfly; D at 161.8% for a Crab. One invalid leg means no pattern. I validate with the Fibonacci tool before I even think about entries, and forced patterns are the single biggest leak in most harmonic traders' results.

What is the best stop loss for harmonic patterns?

Beyond the pattern's structural extreme plus a buffer: beyond X for Bats, beyond the 161.8% extreme for Butterflies and Crabs. Tight stops are the number one killer — the D zone routinely overshoots 10–20 pips before reversing. Size down to 0.5% risk on extension patterns so the wide stop does not force you into meaningless position sizes.

Can I automate harmonic pattern trading with an EA?

Partially. EAs can scan and validate ratios faster than any human, and the BabyPips harmonic lessons are a good primer before you code the rules. But the discretionary filters — H4 alignment, divergence, confluence at D — are hard to code well. I use an indicator for detection and make the trade/no-trade call myself. Full automation of harmonics tends to trade every forced pattern the indicator finds, which is exactly the leak you are trying to avoid.