OPEN-SOURCE SCRIPT
Harmonic Pattern Detection, Prediction, and Backtesting Tool
Updated
Overview:
The ultimate harmonic XABCD pattern identification, prediction, and backtesting system.
Harmonic patterns are among the most accurate of trading signals, yet they're widely underutilized because they can be difficult to spot and tedious to validate. If you've ever come across a pattern and struggled with questions like "are these retracement ratios close enough to the harmonic ratios?" or "what are the Potential Reversal levels and are they confluent with point D?", then this tool is your new best friend. Or, if you've never traded harmonic patterns before, maybe it's time to start. Put away your drawing tools and calculators, relax, and let this indicator do the heavy lifting for you.
- Identification -
An exhaustive search across multiple pivot lengths ensures that even the sneakiest harmonic patterns are identified. Each pattern is evaluated and assigned a score, making it easy to differentiate weak patterns from strong ones. Tooltips under the pattern labels show a detailed breakdown of the pattern's score and retracement ratios (see the Scoring section below for details).
- Prediction -
After a pattern is identified, paths to potential targets are drawn, and Potential Reversal Zone (PRZ) levels are plotted based on the retracement ratios of the harmonic pattern. Targets are customizable by pattern type (e.g. you can specify one set of targets for a Gartley and another for a Bat, etc).
- Backtesting -
A table shows the results of all the patterns found in the chart. Change your target, stop-loss, and % error inputs and observe how it affects your success rate.
//------------------------------------------------------
// Scoring
//------------------------------------------------------
A percentage-based score is calculated from four components:
(1) Retracement % Accuracy - this measures how closely the pattern's retracement ratios match the theoretical values (fibs) defined for a given harmonic pattern. You can change the "Allowed fib ratio error %" in Settings to be more or less inclusive.
(2) PRZ Level Confluence - Potential Reversal Zone levels are projected from retracements of the XA and BC legs. The PRZ Level Confluence component measures the closeness of the closest XA and BC retracement levels, relative to the total height of the PRZ.
(3) Point D / PRZ Confluence - this measures the closeness of point D to either of the closest two PRZ levels (identified in the PRZ Level Confluence component above), relative to the total height of the PRZ. In theory, the closer together these levels are, the higher the probability of a reversal.
(4) Leg Length Symmetry - this measures the ΔX symmetry of each leg. You can change the "Allowed leg length asymmetry %" in settings to be more or less inclusive.
So, a score of 100% would mean that (1) all leg retracements match the theoretical fib ratios exactly (to 16 decimal places), (2) the closest XA and BC PRZ levels are exactly the same, (3) point D is exactly at the confluent PRZ level, and (4) all legs are exactly the same number of bars. While this is theoretically possible, you have better odds of getting struck by lightning twice on a sunny day.
Calculation weights of all four components can be changed in Settings.
//------------------------------------------------------
// Targets
//------------------------------------------------------
A hard-coded set of targets are available to choose from, and can be applied to each pattern type individually:
(1) .618 XA = .618 retracement of leg XA, measured from point D
(2) 1.272 XA = 1.272 retracement of leg XA, measured from point D
(3) 1.618 XA = 1.618 retracement of leg XA, measured from point D
(4) .618 CD = .618 retracement of leg CD, measured from point D
(5) 1.272 CD = 1.272 retracement of leg CD, measured from point D
(6) 1.618 CD = 1.618 retracement of leg CD, measured from point D
(7) A = point A
(8) B = point B
(9) C = point C
//------------------------------------------------------
// Stops
//------------------------------------------------------
Stop-loss levels are also user-defined, in one of three ways:
(1) % beyond the furthest PRZ level (below the PRZ level for bullish patterns, and above for bearish)
(2) % beyond point D
(3) % of distance to Target 1, beyond point D. This method allows for a proper Risk:Reward approach by defining your potential losses as a percentage of the potential gains. This is the default.
//------------------------------------------------------
// Results Table / Backtesting Statistics
//------------------------------------------------------
To properly assess the effectiveness of a specific pattern type, a time limit is enforced for a completed pattern to reach the targets or the stop level. When this time limit expires, the pattern has "timed out", and is no longer considered in the Success Rate statistics. During the time limit period, if price reaches Target 1 before reaching the Stop level, the pattern is considered successful. Conversely, if price reaches the Stop level before reaching Target 1, the pattern is considered a failure. The time limit can be changed in Settings, and is defined in terms of the total pattern length (point X to point D). It is set to 1.5 by default.
Increasing the time limit value will give you more realistic Success Rate values, but will less accurately represent the success rate of the harmonic patterns (i.e. the more time that elapses after a pattern completes, the less likely it is that the price action is related to that pattern).
//------------------------------------------------------
// Coming soon...
//------------------------------------------------------
I have a handful of other features in development, including:
(1) Drawing incomplete patterns as they develop. This will allow you more time to plan entries and stops, or potentially trade reversals from point C to point D PRZ levels.
(2) Support for the Shark and Cypher patterns
(3) Alerts
Please report any bugs, runtime errors, other issues or enhancement suggestions.
I also welcome any feedback from experienced harmonic pattern traders, especially regarding your strategy for setting targets and stop-losses.
reees
The ultimate harmonic XABCD pattern identification, prediction, and backtesting system.
Harmonic patterns are among the most accurate of trading signals, yet they're widely underutilized because they can be difficult to spot and tedious to validate. If you've ever come across a pattern and struggled with questions like "are these retracement ratios close enough to the harmonic ratios?" or "what are the Potential Reversal levels and are they confluent with point D?", then this tool is your new best friend. Or, if you've never traded harmonic patterns before, maybe it's time to start. Put away your drawing tools and calculators, relax, and let this indicator do the heavy lifting for you.
- Identification -
An exhaustive search across multiple pivot lengths ensures that even the sneakiest harmonic patterns are identified. Each pattern is evaluated and assigned a score, making it easy to differentiate weak patterns from strong ones. Tooltips under the pattern labels show a detailed breakdown of the pattern's score and retracement ratios (see the Scoring section below for details).
- Prediction -
After a pattern is identified, paths to potential targets are drawn, and Potential Reversal Zone (PRZ) levels are plotted based on the retracement ratios of the harmonic pattern. Targets are customizable by pattern type (e.g. you can specify one set of targets for a Gartley and another for a Bat, etc).
- Backtesting -
A table shows the results of all the patterns found in the chart. Change your target, stop-loss, and % error inputs and observe how it affects your success rate.
//------------------------------------------------------
// Scoring
//------------------------------------------------------
A percentage-based score is calculated from four components:
(1) Retracement % Accuracy - this measures how closely the pattern's retracement ratios match the theoretical values (fibs) defined for a given harmonic pattern. You can change the "Allowed fib ratio error %" in Settings to be more or less inclusive.
(2) PRZ Level Confluence - Potential Reversal Zone levels are projected from retracements of the XA and BC legs. The PRZ Level Confluence component measures the closeness of the closest XA and BC retracement levels, relative to the total height of the PRZ.
(3) Point D / PRZ Confluence - this measures the closeness of point D to either of the closest two PRZ levels (identified in the PRZ Level Confluence component above), relative to the total height of the PRZ. In theory, the closer together these levels are, the higher the probability of a reversal.
(4) Leg Length Symmetry - this measures the ΔX symmetry of each leg. You can change the "Allowed leg length asymmetry %" in settings to be more or less inclusive.
So, a score of 100% would mean that (1) all leg retracements match the theoretical fib ratios exactly (to 16 decimal places), (2) the closest XA and BC PRZ levels are exactly the same, (3) point D is exactly at the confluent PRZ level, and (4) all legs are exactly the same number of bars. While this is theoretically possible, you have better odds of getting struck by lightning twice on a sunny day.
Calculation weights of all four components can be changed in Settings.
//------------------------------------------------------
// Targets
//------------------------------------------------------
A hard-coded set of targets are available to choose from, and can be applied to each pattern type individually:
(1) .618 XA = .618 retracement of leg XA, measured from point D
(2) 1.272 XA = 1.272 retracement of leg XA, measured from point D
(3) 1.618 XA = 1.618 retracement of leg XA, measured from point D
(4) .618 CD = .618 retracement of leg CD, measured from point D
(5) 1.272 CD = 1.272 retracement of leg CD, measured from point D
(6) 1.618 CD = 1.618 retracement of leg CD, measured from point D
(7) A = point A
(8) B = point B
(9) C = point C
//------------------------------------------------------
// Stops
//------------------------------------------------------
Stop-loss levels are also user-defined, in one of three ways:
(1) % beyond the furthest PRZ level (below the PRZ level for bullish patterns, and above for bearish)
(2) % beyond point D
(3) % of distance to Target 1, beyond point D. This method allows for a proper Risk:Reward approach by defining your potential losses as a percentage of the potential gains. This is the default.
//------------------------------------------------------
// Results Table / Backtesting Statistics
//------------------------------------------------------
To properly assess the effectiveness of a specific pattern type, a time limit is enforced for a completed pattern to reach the targets or the stop level. When this time limit expires, the pattern has "timed out", and is no longer considered in the Success Rate statistics. During the time limit period, if price reaches Target 1 before reaching the Stop level, the pattern is considered successful. Conversely, if price reaches the Stop level before reaching Target 1, the pattern is considered a failure. The time limit can be changed in Settings, and is defined in terms of the total pattern length (point X to point D). It is set to 1.5 by default.
Increasing the time limit value will give you more realistic Success Rate values, but will less accurately represent the success rate of the harmonic patterns (i.e. the more time that elapses after a pattern completes, the less likely it is that the price action is related to that pattern).
//------------------------------------------------------
// Coming soon...
//------------------------------------------------------
I have a handful of other features in development, including:
(1) Drawing incomplete patterns as they develop. This will allow you more time to plan entries and stops, or potentially trade reversals from point C to point D PRZ levels.
(2) Support for the Shark and Cypher patterns
(3) Alerts
Please report any bugs, runtime errors, other issues or enhancement suggestions.
I also welcome any feedback from experienced harmonic pattern traders, especially regarding your strategy for setting targets and stop-losses.
reees
Release Notes
Various bug fixes:- reduce probability of bar history reference RT error
- keep track of multiple pending/in progress patterns
Release Notes
Fix for not re-drawing patterns > 400 bars backRelease Notes
BIG UPDATE: Potential patterns will now be drawn after pivot point C is formed, along with projections to the PRZ levels.Release Notes
Fixed issue with updating label for pending patterns when target or stop is hit before time expires.Release Notes
Adjusted scoring so that PRZ level confluence and Point D/PRZ confluence are calculated relative to the height of the pattern (rather than the height of the PRZ). This ensures equity across pattern types (PRZ is defined differently for each type of pattern, so it's only a valid measure when comparing patterns of the same type). Release Notes
Fixed issue with incomplete patterns not getting deleted when no longer validRelease Notes
- Calculate PRZ confluence and Point D/PRZ confluence relative to the height of the XA leg (this is most equitable across pattern types)- Project reversal targets from PRZ levels (rather than all from current point D)
- Consolidated addValidPattern()
- Removed unused code
Release Notes
Removed commentsRelease Notes
(1) Added alerts(2) Now including score on potential/incomplete patterns
(3) Now showing % error for each leg retracement in the pattern tooltip
(4) Minor bug fixes
Release Notes
Improved tooltip formattingRelease Notes
Display changesRelease Notes
(1) Now showing target/stop locations for each completed pattern (2) Bug fixes
Release Notes
(1) Updated to latest version of TA library, which includes a bug fix for the XABCD pattern search algorithm(2) Added decimal places to the display of stops, targets, and PRZ levels
Release Notes
Increased default leg length asymmetry. May need to consider adjusting asym validation to consider length of XA and CD separately from AB and BC. Release Notes
Updated to latest version of TA library, which fixes a bug where intermediate lows were not being detected between points C and D.Release Notes
Color adjustmentRelease Notes
BIG UPDATE: Introducing the concept of entries. Choose to enter trades after point C, after point D, or both, based on pattern score and your specified entry level. - Entries after point C are specified relative to PRZ levels
- Entries after point D are specified as limit orders relative to point D.
The results table now includes statistics for Return %, giving much more valuable feedback regarding your inputs for stop, target, entry levels, and score weights. Play around with the settings and see how it affects your returns!
I've also added a parameter for the number of trailing bars to validate a pattern, and fixed a number of bugs.
Release Notes
bug fixRelease Notes
Bug fixesRelease Notes
(1) Linefilling background between confluent PRZ levels rather than entire PRZ(2) bug fixes
Release Notes
Added Shark and Cypher patterns. Release Notes
Updated to newest version of TA script, bug fix for Shark validationRelease Notes
- Added AD-based targets- Added Point X-based stop loss
Release Notes
Fixed bug with validation of cypher CD leg (now validating against XC rather than XA)Release Notes
(1) Updating in-progress patterns when entry window has timed out(2) Removed redundant "==true" comparisons
Release Notes
Updated to latest version of TA library, which includes bug fix for Cypher PRZ level (affects cypher entries after point C)Release Notes
Fixed spacing on incomplete pattern tooltipRelease Notes
Added .5 AD as a target optionRelease Notes
(1) Added parameter for toggling Potential/Incomplete patterns(2) Bug fix for issue with targets/stops erroneously hit on entry bar
Release Notes
A number of bug fixes including:(1) better handling of incomplete patterns
(2) more efficient entry and target checks
Release Notes
Fixed "Objects cannot be positioned more than 500 bars in the future" runtime errorRelease Notes
Disabling display of Potential/Incomplete patterns by default until issues are resolved. This will not affect trades/backtest results.Release Notes
Fixed bug with invalid incomplete patterns being addedRelease Notes
Bug fixes for drawing/updating incomplete/potential patternsRelease Notes
(1) Removed Leg Length Asymmetry component of pattern score (my research shows no correlation with pattern success). The asymmetry parameter is still available for pattern validation/filtering. (2) Fixed issue with incomplete pattern scores when PRZ level values are <0
(3) Fixed issue with incomplete pattern projection arrows. No longer showing the arrow when if it points to an entry/PRZ level at or below 0.
Release Notes
Major improvements to incomplete/potential patterns:(1) fixed bug where potential patterns were being missed when there were multiple pattern types with the same ABC coordinates. This was preventing the patterns from being drawn, and also preventing "After point C" entries to be made.
(2) now drawing incomplete pattern labels on projected point D/entry/PRZ level to reduce label overlapping:
Release Notes
Updated to use latest version of Fib library (improved efficiency for precise fibs)Release Notes
Firing alert only on bar close until issue with real-time bar alert spamming is resolvedRelease Notes
Added parameter to only include patterns that meet your "If score is above" entry requirement Release Notes
Force inclusion of completed patterns added when entry is hit after point C, regardless of score or "Only high scoring" paramRelease Notes
Added parameter to set number of bars to look back. Improves performance somewhat, but I'm unable to implement this the way I'd like to, so there's still a lot of room for improvement. Release Notes
Major rewrite to use xabcd_harmonic UDT object. Runtime performance has improved, but minimally (extremely disappointing). Compile time improved drastically (down to 20 seconds from ~2 minutes).Rewrite also includes a plethora of bug fixes and minor enhancements, including:
- more details on the pattern label tooltip (now showing Entry, Stop, T1 & T2 levels so they're always available for reference, regardless of whether they're currently drawn on the chart).
- a dashed line from point C to point D when point D is unconfirmed (i.e. pattern was created from an entry on a Potential/incomplete pattern after Point C, not from a valid pivot).
Release Notes
- Updated to use point objects (with latest version of xabcd_harmonic object)- various bug fixes
Release Notes
Big Update:- Major reduction in calculation time
- Improvements to pattern finding algorithm. Now identifies and displays incomplete patterns sooner (1 bar after C) and finds more patterns.
- "Enter after point C" is turned off by default because average return rate is better without it
- Projection arrow from entry level rather than arrows from all PRZ levels
- Bug fixes
Release Notes
Bug fixesRelease Notes
Added "None" option to disable Target 2Open-source script
In true TradingView spirit, the author of this script has published it open-source, so traders can understand and verify it. Cheers to the author! You may use it for free, but reuse of this code in publication is governed by House rules. You can favorite it to use it on a chart.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.