Candlestick patterns

Definition

Candlestick charting encodes each period’s open, high, low, and close into a “candle” (body and wicks) and names recurring one-to-three-candle configurations — doji, hammer, engulfing, stars — as reversal or continuation signals. Imported from Japanese rice-trading tradition and codified for Western markets by Nison (1991), the vocabulary is now universal chart grammar. The formal evidence is carried plainly: the standard peer-reviewed test (Marshall-Young-Rose, testing the full pattern catalog on Dow stocks with bootstrap methodology) found NO reliable predictive value after realistic frictions.

How it works / structure

• The encoding (the durable part): body = open-to-close (conviction of the period’s winners); wicks = rejected extremes (prices visited and refused); the encoding itself is information-dense visualization — reading rejection wicks at a level is reading intraperiod auction failure, whatever one thinks of pattern names.
• The pattern taxonomy (labeled convention): single-candle (doji — indecision; hammer — low rejected), two-candle (engulfing — period’s range consumed), three-candle (morning/evening stars); every name is a hypothesis about order flow at that location, testable only in parameterized form (candle proportions defined via indicator-atr normalization, context gates, exit rules).
• The evidence record (the load-bearing point): Marshall-Young-Rose (2006) — the catalog, tested with data-snooping controls: no value; successor studies split by market and era with mostly null results (labeled); the honest summary is that PATTERN NAMES carry no documented edge, while LOCATION + rejection logic (a hammer AT a tested level with volume) is a different, compound hypothesis each implementation must test itself.
• Why the vocabulary persists (engine-relevant): it compresses intraperiod flow into readable units; widely-watched patterns at widely-watched levels can self-fulfill briefly (bias-recency, crowd anchor effects) — a positioning observation, not a pattern endorsement.

When it applies

Chart literacy (agents parsing human commentary need the grammar); feature construction (body/wick ratios and rejection measures are legitimate parameterized inputs — stripped of names, tested as features); setup specification (disc-playbook-setups — if a candle configuration gates an entry, it enters the playbook as parameters, never as a name).

Risk profile & failure modes

• Name-based trading (the documented failure): the tested catalog carries no edge — trading “a hammer” as such is trading noise with a mascot.
• Pattern pareidolia: three-candle configurations occur constantly by chance — un-gated pattern scanning generates unlimited false signals.
• Timeframe arbitrariness: the same price path prints different patterns at different period boundaries — session-boundary sensitivity is a structural fragility (documented in the test literature’s period-dependence).
• Folklore authority: the tradition’s age and ubiquity substitute for evidence in retail materials — the KB’s citation of the null result is the antidote.

Evidence & limits

Nison (1991) is the codification (practitioner, labeled); Marshall-Young-Rose (2006) is the peer-reviewed null on the pattern catalog; successor literature is mostly consistent (labeled). The encoding- as-feature approach (proportions, rejection metrics) is legitimate hypothesis material graded per implementation.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Lower-wick rejection candles (wick > 2× body, ATR-normalized) at 20-day range lows precede above-median 5-day returns (parameterized rejection test — not a named pattern)” — falsified by the conditional distribution.
• “The named-pattern catalog, traded per Nison’s rules with realistic frictions, matches the Marshall-Young-Rose null on this platform’s universe (replication)” — falsified by a robust positive result.

Cross-references

• The location logic it needs: strategy-mean-reversion, ta-wyckoff-method (rejection grammar)
• The normalization: indicator-atr; the discipline: disc-playbook-setups
• The tradition family: ta-dow-theory, ta-elliott-wave
• The lens: lens-technical