Open Harmonics · Pattern guide
The pattern guide
Seven harmonic patterns, defined exactly as our TradingView script detects them, with ten years of results on about 500 US large-cap stocks.
All patterns share one skeleton: five turning points X, A, B, C and D, found by a zigzag. The pattern is decided by the ratios between the legs (Fibonacci ratios such as 0.618 and 1.618, within ±6%), and C must retrace at least 50% of AB. The strategy buys at D once price confirms a reversal.
Every pattern exits the same way: a stop just beyond the lower of D and X, half off at 1.5R, the rest at 2R. Read how the exits work and what we've tested →
Mean R per setup, by pattern and timeframe
| Pattern | 15-minute | 1-hour | 4-hour | Daily |
|---|---|---|---|---|
| Gartley | +0.05 n=512 | +0.14 n=441 | +0.49 n=125 | +0.31 n=71 |
| Bat | +0.00 n=664 | +0.25 n=491 | +0.35 n=144 | +0.28 n=72 |
| Butterfly | +0.03 n=1,280 | +0.15 n=957 | +0.36 n=275 | +0.25 n=141 |
| Crab | +0.11 n=565 | +0.16 n=336 | +0.30 n=98 | +0.21 n=28 |
| Deep Crab | +0.19 n=465 | +0.22 n=352 | +0.48 n=91 | +0.49 n=44 |
| Cypher | +0.09 n=830 | +0.25 n=694 | +0.08 n=222 | +0.39 n=113 |
| Shark | +0.09 n=12,180 | +0.19 n=9,937 | +0.20 n=3,035 | +0.26 n=1,629 |
Instrument-level differences are mostly sampling noise: on every timeframe the spread of per-instrument mean R is close to what chance alone produces, and an instrument's odd-numbered setups do not predict its even-numbered ones. Do not read the instrument table as a ranking. The same caution applies to small differences between patterns: their 95% ranges overlap heavily.
Backtests on today's Nasdaq-100 / S&P 500 members (survivorship bias), TradingView Strategy Tester, 0.01% commission, no slippage, all entry filters off. Education and research only — not financial advice.