Chart Pattern Confidence Scores: What the Numbers Mean

Not All Pattern Detections Are Equal

You have probably seen chart pattern scanners flag dozens of setups on any given day. But here is the uncomfortable truth: most of those detections are noise. A bull flag with a 30/100 confidence score and a bull flag with a 90/100 score might look similar at first glance, but their expected outcomes are dramatically different.

A chart pattern confidence score is a numerical assessment of how well a detected pattern matches the ideal formation. Our scoring system evaluates multiple dimensions of each detection — shape, proportion, volume, and context — and distills them into a single number from 0 to 100. Here is exactly how it works and how to use it.

How Confidence Scores Are Calculated

Our detection engine evaluates each pattern across five dimensions:

1. Shape Quality (30% of score)

How closely does the detected formation match the ideal pattern shape? A cup and handle with a smooth, rounded bottom scores higher than one with a jagged, erratic bottom. A head and shoulders with symmetric shoulders scores higher than one with wildly uneven shoulders.

The algorithm measures this using the deviation of actual price points from the ideal mathematical curve of the pattern.

2. Proportions (20% of score)

Every pattern has ideal proportions. For a cup and handle, the cup depth should be 15-35% of the prior uptrend, and the handle should form in the upper third of the cup. For a head and shoulders, the two shoulders should be at roughly the same height, and the head should extend clearly above both.

Patterns with proportions close to the textbook ideal score higher. Those with distorted proportions (handle too deep, shoulders too uneven) receive lower scores.

3. Volume Profile (25% of score)

Volume confirms or denies a pattern's validity. Our engine analyzes the volume pattern across the formation and compares it to the expected profile for that specific pattern type.

For example, a bull flag should show declining volume during the flag consolidation and increasing volume on the breakout. If the flag has rising volume (buyers still selling), the volume score drops.

4. Duration and Timeframe (10% of score)

Patterns that form over the appropriate time period for their type score higher. A head and shoulders that develops over 3 months on a daily chart scores higher than one that forms in 3 days. A pennant that lasts 7 days scores higher than one that drags on for 6 weeks.

5. Trend Context (15% of score)

Patterns do not exist in isolation. The engine evaluates whether the pattern appears in the correct context:

• Continuation patterns score higher when they form during a strong trend
• Reversal patterns score higher when they form after an extended trend
• Patterns that form in the direction of the larger timeframe's trend receive a bonus

Interpreting the Score

80-100: High Confidence

The pattern matches the textbook formation closely across all dimensions. Shape is clean, proportions are correct, volume confirms the pattern, and the trend context supports the setup.

How to trade: These are your A-grade setups. Use full position sizing (within your 1-2% risk limit). The measured move target has the highest probability of being reached.

60-79: Moderate Confidence

The pattern is valid but has one or two imperfections — perhaps the volume profile is not ideal, or the proportions are slightly off. The formation is recognizable but not textbook-perfect.

How to trade: These are B-grade setups. Use 50-75% of your normal position size. Consider tightening your target to 75% of the measured move.

40-59: Low-Moderate Confidence

The pattern has notable flaws. The shape may be rough, the volume may not confirm, or the context may be questionable. The pattern is detectable but not convincing.

How to trade: These are C-grade setups. Either pass or use minimal position size (25-50% of normal). Require additional confirmation (indicator alignment, sector strength) before entering.

Below 40: Low Confidence

The algorithm has detected a pattern-like formation, but the quality is poor. Shape is messy, volume is contradictory, or the context is wrong.

How to trade: Skip these. The probability of the pattern completing as expected is too low to justify the risk.

Confidence Scores and Win Rates

Our backtesting data shows a clear relationship between confidence scores and pattern completion rates:

• 80+ confidence: ~75% completion rate (pattern reaches measured move)
• 60-79 confidence: ~60% completion rate
• 40-59 confidence: ~45% completion rate
• Below 40: ~30% completion rate

This data covers thousands of pattern detections across stocks and crypto over multiple market conditions.

How to Use Confidence Scores in Practice

Filter, Don't Just Sort

Instead of trading the highest-scoring pattern you find, set a minimum confidence threshold (I recommend 65+) and then look for other confirming factors within that filtered list.

Compare Scores Within Pattern Types

A 70-score cup and handle and a 70-score bull flag may have different expected outcomes because the patterns themselves have different base win rates. Use the confidence score as a quality filter within each pattern type, not as a comparison across pattern types.

Watch for Score Changes

Our scanner updates confidence scores as patterns develop. A pattern that starts at 55 and improves to 75 over several days is showing increasing quality — the market is confirming the formation. A pattern that degrades from 80 to 60 is a warning sign.

The Confidence Score on TradingPatterns.io

Every pattern detection on our platform includes a confidence score from 0 to 100. You can filter and sort by score on the scanner page to focus on the highest-quality setups.

Free accounts receive confidence scores for the three free-tier patterns (cup and handle, head and shoulders, double bottom). Pro and Premium accounts get scores for all 46 or 78 pattern types.

Start scanning with confidence scores — sign up free and see the scoring system in action.