What are candlestick reversal “rules”?
Candlestick reversal describes a way to interpret a sequence of candlesticks as a potential change in market direction. The “rules” are not universal laws of price; they are practical, testable criteria people use to define when a reversal is claimed to be occurring.
To make those criteria independently checkable, you need three components:
- A definition of reversal you will test (for example, whether you expect the next movement to move up or down, or whether you simply label the point as a “turning attempt”).
- A rule set for identifying the pattern (what candle shapes and relationships must be present).
- A measurement plan (what time window counts as confirmation, and what data fields are used).
Because the exact market, time frame, and data quality can vary, the most reliable way to describe “rules” is to list observable conditions and the assumptions under which you would evaluate them.
Mechanism: a testable condition set (context + candle structure)
A candlestick reversal rule set typically combines context and candlestick structure.
1) Context rules (where the candle sequence occurs)
Most reversal interpretations assume that a meaningful turning attempt happens after a prior move. A simple context rule for testing is:
- Label candidate reversals only when the preceding candles show a relatively consistent direction over a defined lookback window.
A testing assumption must be explicit. For example, you might define “preceding move” as:
- Over the prior N candles, the series made higher highs (up-move context) or lower lows (down-move context).
The value of N is a variable. Different choices can change results, so treat it as part of the rule set you can swap and re-test.
2) Candle structure rules (what the individual candles look like)
Candlestick patterns vary, but the underlying “rule types” are usually observable:
- Body orientation: whether the candle closes above (bullish) or below (bearish) its open.
- Body size relative to range: larger bodies often indicate stronger buying or selling pressure.
- Wicks/shadows: long upper or lower wicks can show rejection of a price level.
- Relative positioning: whether the candle’s body or wick overlaps with the prior candle’s body.
A generic reversal-label rule set might require, for a potential downward-to-upward (bullish) reversal attempt:
- The prior context suggests a downward move.
- The candidate candle has a bullish close (close above open).
- The lower wick extends beyond what you consider typical for that candle range (a rejection of lower prices).
- The candle’s close occurs in the upper portion of its own range.
For a potential upward-to-downward (bearish) reversal, swap bullish/bearish roles:
- Prior context suggests an upward move.
- Candidate candle has a bearish close.
- Upper wick shows rejection.
- Close occurs in the lower portion of its range.
Even if you use a named pattern (for example, variations sometimes described in textbooks), you can still convert it into checkable measurements like “body must be more than X% of range” and “wick length must be at least Y% of range.” That turns interpretation into a testable rule.
3) Sequence rules (how many candles and their relationships)
Reversal rules often include a multi-candle relationship, such as:
- The candidate candle changes the direction of the close compared with the previous candle.
- Bodies overlap or do not overlap according to a specified threshold.
- A later candle closes beyond a prior reference level.
A workable approach is to define one reference level and one confirmation requirement.
- Reference level example: the previous candle’s high/low or the candidate candle’s body boundary.
- Confirmation requirement example: “a move within the next M candles that reaches beyond the reference level by at least Z.”
If you do not define “confirmation,” you will be testing ambiguity rather than rules.
Evidence or example: how you can verify a reversal label
Here is a way to test candlestick reversal rules without assuming profit.
Example rule skeleton (not a prediction)
Assume you want to evaluate whether a reversal label is followed by directional movement.
- Pick a time frame and use a consistent OHLC data source.
- Define context: consider an “up-move context” if the prior N candles had higher closes overall.
- Define a candidate candle: require the candidate to be bearish with a body that is at least a chosen fraction of its range, plus an upper wick that is longer than a chosen fraction.
- Define confirmation window: within the next M candles, measure whether price reaches beyond the reference level.
What you must record
For each candidate label, record:
- The timestamp of the candidate candle.
- The measured wick and body ratios used by your rules.
- Whether confirmation happened by your definition.
Material limitation: confirmation choice changes results
Changing M, Z, or the body/wick thresholds can dramatically alter the apparent success rate. This does not mean the “rules” are worthless; it means you must treat the rule set as a hypothesis that needs re-testing whenever assumptions change.
Also, note a key uncertainty:
- Candlestick patterns are visually descriptive; translating them into numerical thresholds introduces measurement decisions that can vary by data vendor and feed.
Limitations and failure modes you should expect
Candlestick reversal rules commonly fail in ways you can anticipate.
1) Market conditions can make candles misleading
In choppy or low-volatility periods, candle bodies and wicks may reflect noise. Your reversal “context” rule may still trigger, but the confirmation window may not show a sustained directional move.
2) Execution costs and spread-like frictions (where relevant) can dominate
Even if a reversal label correlates with future movement in raw price data, practical outcomes depend on costs and how fills are handled. Since this article assumes no real-time data and no jurisdiction-specific trading details, you should separate:
- Price-pattern evaluation (measuring movement in historical bars) from
- Trade outcome evaluation (which depends on execution, costs, and rules).
3) Overfitting by threshold tuning
If you try many combinations of body/wick thresholds, lookback N, and confirmation window M, you may find a configuration that matches historical data better than others. That can create a false sense of reliability.
A testable safeguard is to evaluate out-of-sample periods:
- Tune on one set of historical data and confirm the results on another set.
4) Ambiguity: “reversal” is not one single event
Reversal can mean:
- A short-term turn that lasts a few candles,
- A medium-term correction,
- Or a larger structural change.