Harmonic patterns are a family of five-point price structures (labelled X-A-B-C-D) in which each leg is required to retrace or extend the prior leg by a specific Fibonacci ratio, within a small tolerance. The premise — originated by H.M. Gartley (Profits in the Stock Market, 1935), formalized into named patterns with strict ratios by Scott Carney (and popularized by Larry Pesavento, Bryce Gilmore) — is that markets trace geometric, Fibonacci-proportioned swings, and that when a swing completes a recognized harmonic shape, the final point D marks a high-probability reversal zone. The core tension is sharp: the patterns are highly specified (multiple precise ratio constraints per pattern), which makes them feel rigorous, but that same specificity creates large researcher degrees of freedom and selection bias, and there is essentially no rigorous independent evidence that the geometry confers a tradeable edge. Treat a harmonic pattern as a structured confluence zone to watch, never as a standalone signal.

How it's formed

Every pattern is an XABCD five-point structure: an initial impulse leg XA, a retracement AB, a counter-move BC, and a final leg CD that terminates at point D, the reversal point. What distinguishes one pattern from another is the combination of Fibonacci ratios each leg must satisfy. The canonical ratios (Carney's definitions; cross-checked across IG, TradingSim, and harmonic-trading texts):

• Gartley — AB = 0.618 retracement of XA; BC = 0.382–0.886 of AB; CD = 1.13–1.618 of BC; D = 0.786 retracement of XA (D sits inside the XA leg). This is the original; the 0.786 D and 0.618 B are Carney's defining numbers.
• Bat — AB = 0.382–0.50 of XA (shallower B than Gartley); BC = 0.382–0.886 of AB; CD = 1.618–2.618 of BC; D = 0.886 retracement of XA. Known for a tight, deep PRZ.
• Butterfly — AB = 0.786 of XA; BC = 0.382–0.886 of AB; CD = 1.618–2.618 of BC; D = 1.272–1.618 extension of XA (D projects beyond X). Attributed to Bryce Gilmore.
• Crab — AB = 0.382–0.618 of XA; BC = 0.382–0.886 of AB; CD = 2.24–3.618 of BC; D = 1.618 extension of XA — the most extended, "deepest" pattern (Carney).
• Shark — a 5-point (O-X-A-B-C) variant Carney introduced; C completes near a 0.886–1.13 reciprocal of OX with a 1.618–2.24 BC extension. Often a precursor to the 5-0 pattern.
• Cypher — a four-leg variant (Darren Oglesbee): AB = 0.382–0.618 of XA; C = 1.272–1.414 extension of XA; D = 0.786 retracement of XC.

Tolerances of a few percent around each ratio are universally applied but are subjective and vendor-dependent — there is no canonical tolerance band.

The Potential Reversal Zone (PRZ)

The PRZ is the price zone where the final leg (D) is expected to complete — the confluence of two or more Fibonacci projections (the XA retracement/extension that defines D, plus the BC extension and sometimes an AB=CD equality). Harmonic traders do not treat the exact D price as a trigger; they treat the PRZ as a zone to watch, then require an independent confirmation (a reversal candle, momentum divergence, or shift in order flow) before acting. Stops are conventionally placed just beyond X (or beyond the pattern's extreme); first targets are commonly the 0.382/0.618 retracement of the AD leg. This makes the patterns inherently reversal/mean-reversion tools — they fight the prevailing move, so they suit ranging conditions and tend to fail in strong trends.

Adoption, debate & evidence (read this carefully)

Harmonic patterns are widely taught — they are a staple of retail trading education, charting-platform indicators (TradingView, MT4/5), and paid courses/software — but they are largely absent from institutional and quant practice and from the peer-reviewed literature. The honest evidence picture:

• No rigorous independent edge has been demonstrated. I found no peer-reviewed study establishing that harmonic geometry predicts reversals better than chance after costs. Bulkowski — the standard source for measured chart-pattern base rates — does not validate harmonics the way he does classical patterns. Absence of evidence here is closer to evidence of absence than for better-studied tools.
• Severe researcher degrees of freedom. Choosing the X-A impulse leg is subjective (any chart has many candidate swings), tolerances around each ratio are adjustable, and patterns can be found at any timeframe. With enough free parameters, you can fit a "pattern" to almost any wiggle — a textbook over-fitting risk.
• Selection / survivorship bias in the examples. Educational material and many "repainting" scanners display only completed, successful patterns; failed instances are not shown, inflating perceived reliability. Repainting indicators also redraw the pattern after the fact, which makes honest backtesting nearly impossible.
• Vendor "win-rate" claims should be discounted. Reported accuracy figures (commonly cited ranges run roughly 55–80% depending on pattern and source) come almost entirely from vendors selling indicators or courses, are unaudited, rarely net of costs, and often produced by the repainting tools above. Do not treat any of these numbers as established. Treat them as marketing until reproduced by an independent, point-in-time backtest.
• Weak theoretical basis. Critics (e.g. Rayner Teo) note there is no market-mechanics reason price should reverse at, say, 0.886 of XA beyond self-fulfilling prophecy — unlike a consolidation-breakout, which has an order-flow rationale.

Strengths & limitations

The legitimate value of harmonics is as a risk-defined, confluence-based framework: the structure forces a precise entry zone and a logical stop (just past X), giving favorable potential risk/reward on clean setups — a discipline benefit independent of any predictive edge. Limitations dominate, though: heavy subjectivity in leg selection, no demonstrated standalone edge, poor performance in trends (they're reversal trades), and a strong tendency for practitioners to see patterns that aren't statistically meaningful. The single most common misuse is treating point D / the PRZ as an automatic entry without independent confirmation, and trusting vendor win-rates. Use them, if at all, only as one zone-of-interest input combined with confirmation and strict risk control.

System relevance

For the Augustus trade-setup agent: a harmonic pattern is not a standalone trigger and must never be weighted as predictive on its own. At most, treat a completed PRZ as a structured confluence zone — a place to look for a setup that other, better-evidenced inputs (trend regime, support/resistance, volume, momentum) already favor, and only with independent confirmation at D. Defer all swing-specific operational mechanics (exact entry timing, stop sizing, hold period, scaling) to the Swing Trading branch — don't duplicate them here. Cross-link the Fibonacci retracement/extension definition node for the underlying ratio mechanics.

Sources

• IG — "Top 7 harmonic patterns every trader should know" (per-pattern XABCD ratios, PRZ)
• TradingSim — "Harmonic Patterns: Gartley, Bat, Butterfly & Crab" (Gartley 0.786 D, Bat 0.886 D, Butterfly 1.272–1.618 ext, Crab 1.618 ext)
• HarmonicTrader.com (Scott Carney) — pattern definitions, PRZ / 0.886 retracement / 38.2% trailer attribution
• AvaTrade & naga.com academy — Butterfly (Gilmore, 0.786), Bat tight PRZ, Crab 1.618 XA, Cypher structure
• TradingWithRayner — "5 Problems With Harmonic Trading" (subjectivity of leg selection, weak theoretical basis, trend underperformance)
• QuantifiedStrategies / LiberatedStockTrader / Algotrading-Investment — repainting, backtest difficulty, selection bias, unverified vendor win-rates
• Historical attribution: H.M. Gartley, Profits in the Stock Market (1935); Scott Carney, Harmonic Trading Vols. I–III; Bryce Gilmore (Butterfly); Larry Pesavento (Fibonacci ratios + Gartley); Darren Oglesbee (Cypher)