The short answer
Harmonic patterns are five-point price shapes (X, A, B, C, D) whose legs follow Fibonacci ratios. The four most traded on gold are the Gartley, Bat, Butterfly and Crab. Traders place orders where point D should complete, with a stop just beyond it.
We built an automatic detector (an ATR-based ZigZag plus standard ratio rules) and tested it honestly, acting only on information available at the time. The result was too few trades to judge:
- Hourly, two years: 74 patterns detected, 31 filled, +$84 per ounce after costs (profit factor 1.19).
- 4-hour, two years: 6 trades, all 6 lost.
- Daily, since 2015: 8 trades, roughly break-even.
So this page mainly teaches the patterns and shows why testing them is hard. Treat the numbers as a description of a small sample, not as evidence.
The four patterns
All four start with an impulse leg XA. The pattern is defined by where B retraces and where D completes, both measured against XA. The BC leg retraces part of AB, and CD runs to the completion zone.
| Pattern | B retracement of XA | D (completion) relative to XA | Typical stop | Character |
|---|---|---|---|---|
| Gartley | 61.8% | 78.6% retracement | Beyond X | The original; D sits inside the XA leg |
| Bat | 38.2% – 50% | 88.6% retracement | Beyond X | Deep retracement close to X |
| Butterfly | 78.6% | 127% extension | Beyond the 141.4% extension | D goes beyond X |
| Crab | 38.2% – 61.8% | 161.8% extension | Beyond the 200% extension | The most extended; sharp reversals |
In a bullish pattern, X is a low, A a high, B a low, C a high and D a low where you buy. A bearish pattern is the mirror image.
The Potential Reversal Zone (PRZ)
Experienced harmonic traders do not use the XA ratio alone. They look for a cluster of projections around D:
- the XA ratio for the pattern (for example 88.6% for a Bat)
- an extension of the BC leg (often 1.618 to 2.618)
- an AB=CD leg, where CD equals AB in length
Where these overlap is the PRZ. The tighter the cluster, the more confident the trader. Our Fibonacci lesson covers retracements and extensions on gold in detail, and chart patterns on XAUUSD covers the classic shapes harmonics grew out of.
How we detected and traded them
Swings. A ZigZag that confirms a swing high or low once price reverses by 2 x ATR(14) (variant: 3 x ATR). A swing is only used from the bar where it is confirmed, never earlier. Ratios. B retracement windows: Bat 0.382–0.50, Crab 0.50–0.58, Gartley 0.58–0.66, Butterfly 0.75–0.82 (narrowed so that each set-up maps to one pattern). BC must be 38.2% to 88.6% of AB. D from the XA ratio only; we did not require BC extensions or AB=CD, which strict harmonic traders do. Orders. When C is confirmed, a limit order is placed at D. It is cancelled if price breaks beyond C first, or after twice the time from X to C. Exits. Stop at X (Gartley, Bat), at 141.4% (Butterfly) or 200% (Crab) of XA; target 38.2% of AD (variant: 61.8%). Data. COMEX gold futures (GC=F): 1-hour bars October 2024 to October 2026, 4-hour candles built from them, daily bars since 2015. Cost. $0.30 per ounce per round trip.
The results
| Data and ZigZag | Patterns detected | Trades filled | Win rate | Net (after costs) | Profit factor |
|---|---|---|---|---|---|
| H1, 2 x ATR, target 38.2% of AD | 74 | 31 | 35.5% | +$83.80 | 1.19 |
| H1, 2 x ATR, target 61.8% of AD | 74 | 31 | 22.6% | +$174.00 | 1.36 |
| H1, 3 x ATR, target 38.2% | 49 | 17 | 41.2% | −$50.00 | 0.87 |
| H4, 2 x ATR, target 38.2% | 24 | 6 | 0.0% | −$189.20 | 0.00 |
| H4, 3 x ATR, target 38.2% | 5 | 2 | 50.0% | +$335.10 | — |
| D1 since 2015, 2 x ATR, target 38.2% | 27 | 8 | 50.0% | +$4.70 | 1.03 |
| D1 since 2015, 2 x ATR, target 61.8% | 27 | 8 | 50.0% | +$95.10 | 1.68 |
| D1 since 2015, 3 x ATR, target 38.2% | 16 | 4 | 50.0% | +$2.30 | 1.03 |
By pattern (H1, 2 x ATR, 38.2% target)
| Pattern | Detected | Trades | Win rate | Net |
|---|---|---|---|---|
| Bat | 22 | 7 | 57.1% | +$149.90 |
| Crab | 17 | 4 | 50.0% | +$48.10 |
| Gartley | 18 | 8 | 12.5% | −$220.70 |
| Butterfly | 17 | 12 | 33.3% | +$106.50 |
What the results say, in plain English
- Real harmonic patterns are rare. Two years of hourly data gave 74 candidate patterns and only 31 trades, because many never reached D before breaking C or timing out. The daily chart gave eight trades in almost twelve years.
- The numbers swing with tiny changes. Changing the ZigZag from 2 to 3 ATR turned +$84 into −$50 on H1. One 4-hour trade made $368 and decided that row on its own.
- Per-pattern results mean nothing yet. The Gartley lost on 8 trades and the Bat won on 7. A few trades either way would reverse that.
- Costs are irrelevant here. Harmonic trades are large relative to a $0.30 cost; the problem is the sample, not the spread.
The honest verdict is not tested enough. If someone shows you a harmonic win rate on gold, ask how many trades it is based on and how the swings were chosen.
Why harmonic patterns are hard to test honestly
- Swings are subjective. Two traders will draw different X, A, B and C on the same chart. A ZigZag makes it objective, but a different threshold gives different patterns.
- C is only known late. Point C is confirmed only after price has moved away from it toward D. Backtests that use the final shape of the chart quietly use future information.
- Tolerance creep. Allowing 5% on every ratio turns almost any wave into some pattern.
- The PRZ is a cluster. Our test used one ratio for D. Real traders often wait for several projections to agree, which would mean even fewer trades.
How to trade harmonics more safely
- Use them as zones, not prices. Treat the PRZ as an area to watch for a reaction, then enter on confirmation such as a lower-timeframe structure break.
- Combine with context. A bullish Bat completing at a daily support level or a weekly open carries more weight; see our weekly open test.
- Respect X. If price breaks X on a Gartley or Bat, the pattern is wrong. Exit.
- Count your sample. Keep a list of every pattern you identify in real time, not just the ones that worked. Some traders combine harmonics with Elliott wave counts; the same hindsight warning applies.
Risk sizing
The average stop on the hourly test was about $20.64 per ounce. Risking $100:
- $100 / ($20.64 + $0.30) is about 4.8 ounces (0.05 lots).
- For a Crab or Butterfly, the stop beyond the extension can be much larger; recalculate every time with the position size calculator.
Checklist before you trade a harmonic pattern
- Use a consistent swing method, such as a fixed ZigZag setting on a fixed timeframe, so that X, A, B and C are not chosen after the fact.
- Write the pattern name and the PRZ before price reaches D, with the XA ratio, the BC projection and any AB=CD leg you use.
- Confirm the B point sits inside the pattern’s window. A B at 70% of XA is neither a Gartley nor a Butterfly.
- Place the stop where the pattern is wrong: beyond X for a Gartley or Bat, beyond the next extension for a Butterfly or Crab.
- Check the reward. If the 38.2% retracement of AD is less than twice the distance to the stop, the trade may not be worth taking.
- Record every pattern, including those that break C or never reach D. Without the misses, any win rate is overstated.
Test it yourself
If you trade harmonics, run a forward journal: every time a pattern’s C point forms, record X, A, B, C, your PRZ and your planned entry before price reaches D. After 50 or more patterns you will have a sample that means something. The strategies tested hub explains our method for the other 13 strategies in this series.
Practise it with other traders
Harmonic patterns reward careful, shared review because everyone draws them a little differently. In the XAUUSD Global community, traders post their XABCD drawings before D completes and compare outcomes. The first month is free, then $19.99 a month.
Educational content only, not financial advice. Trading gold with leverage carries a high risk of loss.
Frequently asked questions
Do harmonic patterns work on gold?
We could not show that they do or do not. Detected automatically with a ZigZag and standard ratios, the patterns produced only 31 trades on two years of hourly gold (+$84 per ounce after costs), 8 on daily data since 2015 and 6 on two years of 4-hour data (all six lost). Those samples are far too small for a conclusion.
What are the main harmonic patterns?
The Gartley (B at 61.8% of XA, D at 78.6%), the Bat (B at 38.2% to 50%, D at 88.6%), the Butterfly (B at 78.6%, D at 127% extension of XA) and the Crab (B at 38.2% to 61.8%, D at 161.8% extension). Each has a bullish and bearish version.
What is the PRZ in harmonic trading?
The Potential Reversal Zone, where point D is expected to complete. Traders build it from the XA ratio for D plus other projections, such as the BC extension and an AB=CD leg. Entries are usually limit orders in or near the PRZ.
Where do harmonic traders put stops and targets?
Stops usually go beyond X for the Gartley and Bat, and beyond the next extension (for example 141.4% or 200% of XA) for the Butterfly and Crab. Targets are often 38.2% and 61.8% retracements of the AD leg.
Why are harmonic patterns hard to backtest?
Swing points are subjective, the ratios have tolerances, and point C is only known after price has already started moving toward D. Any test must decide how swings are found and when the pattern is known, and different choices give different results. We used an ATR-based ZigZag and only acted after each swing was confirmed.
Which harmonic pattern did best in your test?
On hourly data, the Bat (7 trades) and Butterfly (12 trades) were positive and the Gartley (8 trades) lost. With so few trades per pattern, those differences are almost certainly not meaningful.
Sources and further reading
- CME Group — Gold futures overview — Contract specifications and trading hours for COMEX gold
- CFTC — Commitments of Traders — Weekly futures positioning, including gold
- Investor.gov — Stop order — How stop orders work, and their limits in fast markets
External sites are independent of XAUUSD Global. Check dates and data yourself.