Price Action Candlestick Patterns: Context and Testing

A candlestick pattern is a description of completed price bars, not a promise about the next bar. To use a pattern responsibly, define its shape before you look at the outcome, read it in market context, state what would invalidate the idea, and test the same rule across unseen historical examples.

Price action analysis focuses on recorded price movement rather than relying on calculated overlays such as RSI or MACD. That does not make it automatically earlier, more accurate, or more profitable than an indicator-based method. Open, high, low, and close are objective data points; labels such as “rejection,” “engulfing,” “support,” and “trend” are interpretations that need consistent rules.

This guide owns the pattern-focused price action intent: how to define common candlestick patterns, place them in context, describe failure, and design a replay test. For the complete no-indicator observation sequence, use the price action workflow. For OHLC geometry, forming candles, chart settings, and timeframe aggregation, use the candlestick chart reading guide. The broader candlestick pattern guide covers a larger pattern catalog.

Educational note: Technical analysis cannot guarantee direction or profit. Historical replay is hypothetical, and live results can differ because of spreads, commissions, slippage, liquidity, gaps, order handling, and trader behavior.

What a Candlestick Actually Records

A candlestick summarizes four prices for a defined bar:

• Open: the first recorded price for the bar.
• High: the highest recorded price during the bar.
• Low: the lowest recorded price during the bar.
• Close: the final recorded price for the bar.

The body shows the distance between open and close. The upper and lower wicks show the prices reached beyond the body. A candle cannot tell you by itself why those prices traded, which participants were involved, or whether the next bar will continue or reverse.

Before naming a pattern, verify the data behind it:

• instrument and venue;
• last-trade, bid, ask, midpoint, mark, or settlement data;
• regular or extended session;
• adjusted or unadjusted prices;
• timezone and session boundary;
• time-based, tick-based, volume-based, or another bar construction;
• completed versus still-forming candle.

Two charts can produce different-looking candles for the same market when their session, provider, or bar settings differ. Pattern rules are not reproducible until those settings are fixed.

A Pattern Is Not a Complete Trade Setup

A named pattern describes shape. A testable setup also needs context and execution rules.

For every setup, write down:

1. Pattern definition: the exact body, wick, range, or relationship required.
2. Market state: trend, range, transition, or unclear.
3. Location: where the pattern appears relative to a versioned level or structure.
4. Trigger: what completed event permits an entry, if any.
5. Invalidation: the observable condition that makes the hypothesis wrong.
6. Order assumption: market, limit, stop, or no-trade decision.
7. Risk rule: how planned loss and position size are determined.
8. Exit rule: the versioned target, trailing rule, time exit, or thesis exit.

Without those fields, the same candle can be labeled a successful signal after a favorable outcome and ignored after an unfavorable one. That is hindsight, not a repeatable method.

Five Pattern Families to Define and Test

The definitions below are starting templates, not universal trading rules. Choose one version, record it, and avoid changing it after seeing the result.

1. Rejection Candle or Pin-Bar Family

A rejection candle usually has a relatively small body and one wick that is large compared with the body. Traders often interpret the long wick as evidence that price traded away from the bar's closing area before returning.

Questions to define:

• How long must the wick be relative to the body or total range?
• Must the body close in the upper or lower part of the candle?
• May the opposite wick exceed a stated size?
• Does the bar need to touch or cross a level defined before it formed?
• Is the bar ignored when its range is unusually small or large?

Possible bullish hypothesis: a long lower wick at a pre-defined area may show that lower prices were rejected during that bar.

Possible bearish hypothesis: a long upper wick at a pre-defined area may show that higher prices were rejected during that bar.

Failure can occur immediately. Price may continue through the wick extreme, remain inside a range, or gap past a planned stop. The candle records one completed interval; it does not prove that buyers or sellers will remain dominant.

2. Engulfing-Candle Family

“Engulfing” can refer to the bodies only or to the full high-low ranges. Those are different rules and should not share one result set.

A body-based bullish engulfing definition might require:

• the first candle to close below its open;
• the second candle to close above its open;
• the second body to cover the first body;
• both candles to be completed;
• a minimum body size relative to recent bars.

The bearish version reverses those relationships.

An engulfing candle can reflect a sharp change within two bars, but it can also appear in noisy ranges or after an already extended move. Test whether location and follow-through add useful information instead of assuming the shape is sufficient.

3. Inside-Bar Family

An inside bar has a high below or equal to the previous bar's high and a low above or equal to the previous bar's low, depending on whether equality is allowed in the rule.

It describes contraction inside the range of the “mother bar.” It does not determine the breakout direction.

Before testing, define:

• whether equal highs or lows are permitted;
• whether the mother bar has a minimum relative range;
• whether entry requires a break, a close beyond the range, or another event;
• how false breaks are handled;
• when an untriggered setup expires.

An inside bar can precede continuation, reversal, or continued consolidation. Treat it as a range relationship, then test a separate directional hypothesis.

4. Doji and Small-Body Family

A doji has an open and close that are equal or nearly equal. “Nearly” must be expressed as a tolerance, such as a fraction of the candle's total range or the instrument's tick size.

A doji shows little net movement between open and close for that bar. It does not automatically mean indecision, reversal, or equilibrium. The same shape can occur because of low activity, a narrow session, price gaps, or the selected timeframe.

Useful context questions:

• Did it form after an extended move or inside a quiet range?
• Is its total range unusual relative to nearby candles?
• Was the bar formed during the main session or a thin period?
• What completed follow-through would confirm or reject the hypothesis?

Use the dedicated doji mistakes guide for a deeper treatment.

5. Three-Candle Reversal Families

Morning-star and evening-star labels describe three-candle sequences. Definitions vary across markets because gap behavior differs between continuously traded markets and instruments with discrete sessions.

A test specification should state:

• the direction and minimum size of the first body;
• the maximum body size of the middle candle;
• whether a gap is required;
• how far the third candle must close into the first body;
• the market and session where the definition applies.

Do not silently remove the gap rule when moving from stocks to forex or crypto. If the definition changes, create a new strategy version and keep its results separate.

How Context Changes the Question

Context does not make a pattern certain. It defines the hypothesis you are testing.

Market State

Classify the visible chart as trend, range, transition, or unclear using rules defined before the setup. A bullish candle after a pullback in an uptrend is a different test from the same candle against a downtrend or in the center of a range.

Use the market structure guide for broad HH/HL/LH/LL state classification. Keep exact BOS/CHoCH terminology with the structure verification owner.

Location

Support and resistance are zones inferred from prior price behavior, not visible order inventories. Avoid claims that a line proves “institutional orders” are waiting there.

Version the rule used to draw a level:

• prior swing high or low;
• previous session high or low;
• gap boundary;
• range edge;
• another repeatable reference.

Draw the level without revealing future bars. If a level is moved after the outcome, record that as a discretionary review rather than pretending it was the original rule. See the support and resistance guide for the separate level-construction intent.

Completed Bars and Follow-Through

A candle can change shape until the bar closes. If your method requires a completed candle, the replay or live rule must wait for that close. If it permits intrabar decisions, specify the data frequency and trigger used; otherwise the historical test may use information that was not available at the decision time.

Follow-through can be defined as:

• a close beyond a pattern boundary;
• a break followed by a retest;
• a later candle that remains above or below a stated level;
• no confirmation, with entry at the next available price.

These are different methods. Test them separately rather than selecting whichever version looks best on each chart.

Multiple Timeframes

A higher timeframe can supply context while a lower timeframe supplies a trigger, but more timeframes do not automatically improve results. They can also create contradictory signals and more opportunities for hindsight selection.

State which timeframe controls:

• market state;
• setup location;
• entry trigger;
• invalidation;
• exit.

Use only candles that were complete at the decision timestamp. The multiple timeframe analysis guide covers synchronization and look-ahead controls.

Price Action Versus Indicators

Price action and indicators are not mutually exclusive quality levels.

• Candles display recorded price data.
• Indicators transform current and historical data using a formula.
• Both approaches depend on definitions, data quality, execution assumptions, and risk controls.
• A price-only rule can lag because it may wait for a candle close or a confirmed swing.
• An indicator can be timely or delayed depending on its formula and how it is used.

The practical question is not “Which is always better?” It is “Can this exact rule be applied consistently, tested without look-ahead bias, and executed under realistic costs and liquidity?”

Removing overlays may make a chart easier to inspect. It does not create an edge by itself.

Build a No-Hindsight Replay Test

Historical replay is useful when it prevents future candles from influencing the decision. It is misleading when the rules are adjusted after each outcome.

Step 1: Freeze the Setup Version

Write the pattern, context, trigger, invalidation, sizing, and exit rules before collecting evaluation results. Give the rule set a version name.

Example fields:

Field	Example specification
Market	One named instrument and session
Data	Provider, timezone, adjustment, bar type
Pattern	Body/wick/range thresholds
Context	Versioned trend/range and level rules
Trigger	Completed close or stated order event
Invalidation	Exact observable condition
Costs	Spread, commission, slippage assumptions
Exit	Fixed, structure, time, or thesis rule
Expiry	When an untriggered setup becomes invalid

The values are examples of fields to define, not recommended trading settings.

Step 2: Separate Development and Evaluation

Use one historical segment to define the method and a later unseen segment to evaluate it. If you change the method after reviewing evaluation results, start a new version and evaluate it on another unseen segment.

This reduces, but does not eliminate, overfitting.

Step 3: Record Every Eligible Example

Log valid, invalid, skipped, winning, losing, and unfilled examples. Selectively saving attractive screenshots exaggerates apparent consistency.

Useful fields include:

• timestamp and instrument;
• screenshot at the decision point;
• rule version;
• context classification;
• pattern qualification;
• intended order and price assumption;
• planned risk;
• outcome under the stated exit rule;
• execution/cost limitations;
• rule deviation.

Step 4: Use Metrics That Match the Question

Win rate alone is incomplete. A method can win often but lose more on losing trades, or win less often with larger average wins.

Review:

• number of eligible examples;
• win, loss, scratch, skipped, and unfilled counts;
• average win and average loss;
• expectancy under stated assumptions;
• maximum observed losing sequence and drawdown;
• results by market state, instrument, session, and timeframe;
• sensitivity to spread, commission, and slippage assumptions;
• percentage of decisions that followed the written rules.

There is no universal sample size that proves a trading edge. More independent observations generally reduce uncertainty, but regime changes, dependence between trades, and repeated rule tuning can still invalidate conclusions.

Step 5: State What Replay Cannot Prove

Replay results are hypothetical. They may not reproduce:

• live bid/ask movement;
• queue position and partial fills;
• spread expansion and market impact;
• order rejection, outage, latency, or margin events;
• every fee and financing charge;
• the behavioral effect of real gains and losses.

The CFTC warns that hypothetical trading results can overstate or understate performance because trades were not executed under actual market conditions and may benefit from hindsight. Treat replay as a process-testing tool, not evidence that future profit is likely.

Stops, Targets, and Execution Are Strategy-Specific

No candlestick shape determines a universal stop distance or reward-to-risk multiple.

A stop should correspond to the method's invalidation and the instrument's execution characteristics. A stop order becomes a market order after its trigger; the execution price is not guaranteed and can differ from the stop price. A stop-limit order controls the acceptable price but may not execute.

Before testing a pattern, specify:

• what invalidates the setup;
• which order type represents the test;
• how gaps or price jumps are handled;
• whether the assumed order could have filled;
• how position size follows from planned loss.

Use the risk management guide for the broader sizing and portfolio-risk architecture. Review the Investor.gov order-types bulletin before assuming a trigger price is a guaranteed fill.

Practice the Pattern Process With ChartMini

ChartMini can help you reveal historical candles sequentially and record simulated decisions without risking real money. The simulator currently supports daily replay for stocks, forex, and crypto, plus intraday practice for forex and crypto.

ChartMini does not automatically identify candlestick patterns, validate them across multiple timeframes, label a setup as high probability, or send pattern alerts. The pattern definition and decision process remain the learner's responsibility.

A practical drill:

1. Choose one instrument, session, timeframe, and pattern version.
2. Hide future candles and define visible market state.
3. Mark the pattern only if every written condition is satisfied.
4. Record trigger, invalidation, order assumption, and planned exit before advancing.
5. Reveal the next candle or bar sequence.
6. Log the outcome and any rule deviation.
7. Review the complete set, including skipped and failed examples.

Start a historical replay session only after writing the rule you intend to test.

Common Failure Modes

Changing the Pattern Definition

If wick length, body size, gap rules, or equality rules change after an outcome, the data no longer describes one method. Create a new version instead.

Reading Future Information

Using a swing label, higher-timeframe close, level, or news outcome that was not available at the decision timestamp introduces look-ahead bias.

Treating Location as Proof

A level can organize a hypothesis, but it does not prove hidden orders or participant identity. Record the level rule and allow it to fail.

Assuming a Stop Price Is a Fill Price

Historical bars may show that a stop level was crossed without revealing the exact executable price, queue, spread, or liquidity available at that moment.

Optimizing One Market and Generalizing Everywhere

A definition developed on one instrument, session, or timeframe may behave differently elsewhere. Evaluate each material change separately.

Ignoring the No-Trade Outcome

Some patterns never trigger under the written rule. Record them as expired or unfilled instead of forcing an entry in hindsight.

Frequently Asked Questions

Are candlestick patterns profitable?

A pattern name alone does not establish profitability. Profitability depends on the complete rules, data, costs, execution, risk, and evaluation method. Historical results do not guarantee future performance.

Is price action better than indicators?

Not universally. Candles and indicators represent data in different ways. Either can be used poorly or within a carefully defined and tested method.

Which candlestick pattern is the most reliable?

There is no universal answer across markets and timeframes. Define one pattern precisely, test it under stated conditions, and report failures as well as successes.

Do I need to memorize every pattern?

No. A small number of precisely defined patterns is easier to test consistently than a large catalog of loosely defined names.

What timeframe should I use?

Use a timeframe compatible with your data, decision schedule, holding period, and execution assumptions. No timeframe is inherently best.

Can ChartMini prove that a pattern works?

No. ChartMini can support blind historical replay and repeatable practice. It cannot reproduce every live-market condition or prove future profitability.

Key Takeaways

• A candlestick records open, high, low, and close for a defined bar.
• Pattern names are interpretations; they require explicit definitions.
• Context defines the hypothesis but does not guarantee the outcome.
• Market state, location, trigger, invalidation, orders, costs, risk, and exits belong in the test specification.
• Completed-bar and multi-timeframe rules must avoid look-ahead information.
• Keep development and evaluation data separate and version every rule change.
• Record failed, skipped, expired, and unfilled examples, not just attractive charts.
• Replay tests process consistency; it does not prove future profitability.
• ChartMini provides historical replay practice, not automatic pattern recognition or alerts.

Sources and Further Reading

• CME Group: Technical Analysis — OHLC candlesticks and the subjectivity of pattern interpretation.
• CME Group: Chart Types — candlestick, bar, timeframe, and chart-construction basics.
• Investor.gov: Understanding Order Types — market, limit, stop, stop-limit, and execution-price limitations.
• CFTC: Commodity Trading Systems Sold on the Internet — limitations of hypothetical results and hindsight-based trading-system claims.