Pattern Intelligence Engine

Automated detection of 20+ technical and harmonic pattern classifications across 30m, 1H, and 4H timeframes. Each receives a 0–100 model score from structural, volume, and level-context inputs. MTF Stacking groups matching labels across timeframes for comparison; it does not establish conviction, probability, or a directional move.

What this page covers

The complete list of patterns detected and their characteristics
How model scoring works and what the displayed factors mean
How MTF stacking groups pattern labels
How to use pattern signals as study context rather than entries

Complete Pattern Library

Category	Patterns detected	Timeframes
Harmonic	Gartley, Bat, Butterfly, Crab, Shark, Cypher, AB=CD, 5-0	30m, 1H, 4H
Chart Patterns	Head & Shoulders, Inverse H&S, Double Top, Double Bottom, Triple Top, Triple Bottom	1H, 4H
Continuation	Bull Flag, Bear Flag, Bull Pennant, Bear Pennant, Ascending Triangle, Descending Triangle, Symmetrical Triangle	30m, 1H, 4H
Reversal	Rising Wedge, Falling Wedge, Cup & Handle, Rounding Bottom	1H, 4H
ICT/SMC Structures	Order Block (OB), Fair Value Gap (FVG), Break of Structure (BOS), Change of Character (CHoCH), Liquidity Void	All TF
Candlestick	Engulfing (bull/bear), Doji, Hammer, Shooting Star, Morning Star, Evening Star, Pinbar, Marubozu	30m, 1H

Confidence Scoring System

Each pattern is scored from 0–100 based on four geometric and contextual factors:

Factor	Weight	What it evaluates
Pattern geometry	40%	How precisely the pattern matches the textbook definition — Fibonacci ratios for harmonics, proportions for chart patterns, body-to-wick ratios for candles
Volume context	25%	How does volume behavior compare with the pattern label and nearby candles?
Key level context	20%	Is the pattern near a displayed support/resistance, volume-profile, or Fibonacci reference?
Timeframe field	15%	Which timeframe contains the pattern label? The field is descriptive, not a quality ranking.

MTF Stacking

MTF stacking groups the same pattern type when it appears on multiple timeframes for the same pair. Agreement across timeframes is a display relationship; it does not identify who saw the pattern or increase a known probability.

Stacking level	Description	Score bonus
Single TF	Pattern on one timeframe only	No bonus
Dual TF	Same pattern on 2 timeframes (e.g. 30m + 1H)	Displayed stack field
Triple TF	Same pattern on all 3 timeframes (30m + 1H + 4H)	Displayed stack field and feed label

Navigating the Pattern Dashboard

The Pattern Intelligence dashboard is organized as a 31-pair × 3-timeframe matrix. Each cell shows the count and model score of active patterns; colors group displayed ranges and do not establish quality or probability.

Use the heatmap view
The matrix shows all 31 supported pairs. Sort by timeframe or displayed range to compare model labels; MTF-stacked detections appear in their labeled section.
Click any cell for detail
The detail panel shows active patterns, score components, completion-zone geometry, and reference measurements. These fields describe the model output rather than suggesting an entry.
Set the display filter
Use the configurable score filter to focus a study view. The chosen value changes visibility only and is not a quality threshold.
Subscribe to MTF-stacked alerts
Toggle 'MTF Only' to display matching pattern labels. An Alert Intel notification identifies the displayed condition and does not imply action.
Review completion history
The Performance tab shows retrospective completion fields and later movement observations. Use them to document uncertainty rather than build conviction.

Pattern Follow-Through Statistics

Pattern history is retrospective and depends on definitions, sample period, data quality, and assumptions. The table below preserves the categories as study fields without presenting outcome rates or trade parameters.

Pattern Category	Displayed score field	Historical comparison field	Later-behavior notes	Timeframe field
Harmonic (Gartley, Bat)	Model output	Retrospective sample	Record later candles	4H
Harmonic (Butterfly, Crab)	Model output	Retrospective sample	Record later candles	1H, 4H
Chart Patterns (H&S, Double Top/Bottom)	Model output	Retrospective sample	Record later candles	4H
Continuation (Bull/Bear Flag)	Model output	Retrospective sample	Record later candles	30m, 1H
ICT/SMC (OB, FVG, BOS, CHoCH)	Model output	Retrospective sample	Record later candles	1H, 4H
Candlestick (Engulfing, Pinbar)	Model output	Retrospective sample	Record later candles	30m, 1H

How to Use Pattern Intelligence as Confluence

A pattern label is not a complete conclusion. It describes geometry and selected context; it does not tell you when to enter, how much to allocate, or where price will move. Use it as one study field among several:

Pattern detection
Record a hypothetical Bearish Gartley label at $2,380 on XAUUSD 1H and note its displayed score without treating it as a threshold.
Session comparison
Record whether the completion label appeared during a configured session window and compare later behavior without assigning extra weight.
Volume comparison
Compare the volume reading near $2,380 with the pattern label; do not infer institutional activity or increased confidence.
Wyckoff phase
Record the 4H Wyckoff label alongside the pattern and document agreement or conflict without assigning probability.
Retrospective review
Compare later candles with the completion zone and record alternate interpretations. No entry trigger is prescribed.

🔺 PATTERN COMPLETION DETECTED
⏰ 10:12 UTC | Kill Zone: ACTIVE (22 min remaining)

📍 XAUUSD — 4H Timeframe
   Pattern: Bat Harmonic (Bearish)
   Confidence: 84/100 ████████░░  ← HIGH
   MTF Stack: 1H + 4H (Dual — +15 confidence bonus)

📐 PATTERN GEOMETRY:
   XA leg:  $2,264 → $2,318  (+54 pips)
   AB leg:  $2,318 → $2,292  (0.386 XA retrace ✓)
   BC leg:  $2,292 → $2,311  (0.618 AB extension ✓)
   CD leg:  $2,311 → $2,285  (0.886 XA retrace ✓)
   PRZ:     $2,283–$2,287    ← completion zone

🔑 STUDY REFERENCES:
   PRZ:        $2,283–$2,287 (completion-zone label)
   Later levels: $2,276 / $2,299 / $2,311 (retrospective references)
   R:R ratio:  omitted — no trade instruction

🔗 Volume IIS at PRZ: 71 — model reading for comparison
🔗 Wyckoff Gate: 4H Markup phase (WPS 74) — displayed label

Frequently asked questions

How should I compare an 80+ confidence pattern with a 60 confidence pattern?

Treat the values as model outputs from different input matches. Compare their source factors, sample period, and later behavior; a higher score does not establish reliability, probability, or action.

Can I filter the feed to show only patterns of a specific type?

Yes. In the Pattern Intelligence dashboard, use the pattern type filter dropdown to select specific categories (e.g. Harmonic only, or Chart Patterns only) or specific patterns (e.g. only Gartley and Bat). The filter persists for your session.

Do patterns repaint after they've been detected?

No. Pattern detection only triggers on a confirmed, closed candle. An in-progress candle is never flagged. Once a pattern fires in the feed, the detection conditions will not change retroactively, even if subsequent price action invalidates the pattern's completion zone.

What's the minimum chart data required for pattern detection?

Harmonic patterns require a minimum of 20 bars of history to compute the XABCD legs. Chart patterns (H&S, double top/bottom) require 40–60 bars. Candlestick patterns need only 3–5 bars. For new pairs or assets with limited history, candlestick and ICT/SMC structures are available first.

How does the Pattern Engine handle uncertain patterns near key levels?

The model includes volume and level-context fields in its score. Review those inputs and the source candles; a lower or higher score does not prove that a pattern is false, valid, or actionable.