UNVERIFIED (Not Refuted With Numbers): Head-and-Shoulders and Classic Chart Patterns

Head-and-shoulders is the flagship classic chart pattern: traders claim its shape predicts a trend reversal. This page cannot meet the wiki's evidence bar (see What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique) — the notes behind it hit a source wall before reaching any pattern-specific effect size, cost threshold, or t-statistic. What follows is honest about that gap rather than papering over it.

What the pattern claims

Head-and-shoulders is defined by Wikipedia's Head and shoulders article as a formation that occurs when a market trend is in the process of reversal, and is recognized as a reversal formation. Classic technical-analysis doctrine holds that this and related formations (triangles) predict significant directional moves and trend reversals (Wikipedia, Technical analysis article). Neither source attaches a number to that claim.

The one relevant academic study found, and why it does not settle this page

Lo, Mamaysky and Wang published in the *Journal of Finance* (2000) a study of technical pattern recognition on U.S. stocks, 1962–1996, using "a systematic and automatic approach to technical pattern recognition using nonparametric kernel regression" instead of subjective visual reading. The study concluded that "several technical indicators do provide incremental information and may have some practical value" (via Wikipedia, Technical analysis article). A second, broader claim about "several academic studies" and technical analysis being "an effective means for extracting useful information from market prices," previously attributed here to Wikipedia, could not be re-located in either the Andrew Lo or Technical analysis Wikipedia articles on recheck and has been removed rather than left unsourced.

Why this does not verify head-and-shoulders specifically

the notes could not confirm that this finding isolates head-and-shoulders (or triangles) from the broader basket of technical indicators tested. Treat any claim that Lo-Mamaysky-Wang "proved head-and-shoulders works" as unverified — the source available here does not support that level of specificity.

What the broader literature says, with the numbers that exist

- A meta-analysis of 95 modern technical-analysis studies found 56 (59%) concluded technical analysis yielded positive results — but researchers flagged data-snooping bias as a likely explanation for part of that share (Wikipedia, Technical analysis article). - Federal Reserve analysis of currency markets found support/resistance levels had some power to predict "intraday trend interruptions," with predictive power varying significantly across exchange rates (Wikipedia, Technical analysis article). - A 2007 Chinese-market study found significant positive returns on buy trades from the contrarian moving-average crossover, channel breakout, and Bollinger band rules even after 0.50% transaction costs (Wikipedia, Technical analysis article) — this is a different technique from head-and-shoulders; see Moving-Average Crossover Rules in Equities — widely claimed, thinly sourced: what we could and could not verify. - Griffioen's analysis (cited via Wikipedia, Technical analysis article) concluded technical trading showed "no statistically significant risk-corrected out-of-sample forecasting power" for most stock indices once transaction costs were substantial.

None of these four data points is about head-and-shoulders by name. No out-of-sample decay percentage, Sharpe ratio, t-statistic, or transaction-cost-survival threshold specific to head-and-shoulders or triangle patterns turned up in the sources reachable for this page.

Why this page stops here

**Verdict per this wiki's grading (see What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique): unverified, not refuted.** "Refuted" would require a documented test of head-and-shoulders itself that failed net of costs, with numbers — that test was not located. "Verified" would require a positive effect size and significance specific to the pattern — also not located. The surrounding literature leans toward skepticism (Griffioen's null result, the data-snooping caveat on the 59%-positive figure) but leaning is not evidence at this wiki's bar; see Multiple Testing: Why t>1.96 Is Not Enough — the bar this wiki uses to grade a factor's significance for why a plausible-sounding aggregate statistic like "59% of studies positive" is not a substitute for a pattern-specific, cost-adjusted, out-of-sample number.

Research access limits behind this gap: the primary Lo-Mamaysky-Wang (2000) paper and Griffioen's work were not retrievable in full text (institutional paywalls); the only source reached was Wikipedia's tertiary summary. A page that actually closes this gap needs the primary papers' numerical results — sample size, effect size, t-stats, and net-of-cost figures — isolated for head-and-shoulders and triangle patterns specifically, not for technical indicators in general.

Related

- What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique — the grading rubric this page fails to clear: no pattern-specific effect size or cost-survival number exists here, which is why the verdict is "unverified" and not "verified" or fully "refuted." - Multiple Testing: Why t>1.96 Is Not Enough — the bar this wiki uses to grade a factor's significance — explains why an aggregate figure like "56 of 95 studies positive" cannot stand in for a single pattern's t-statistic, and why data-snooping bias specifically undercuts headline-positive technical-analysis results. - Moving-Average Crossover Rules in Equities — widely claimed, thinly sourced: what we could and could not verify — a technical rule (moving-average crossover) that, unlike head-and-shoulders here, has documented original numbers and documented later reversals; contrast the evidence quality. - UNVERIFIED (Not Refuted With Numbers): Head-and-Shoulders and Classic Chart Patterns — another visually-defined pattern family with the same evidence gap: claims exist, pattern-specific peer-reviewed numbers largely don't.