Bull Flag Success Rate: Updated Data From 24 Million Chart Patterns

Why We Updated This Analysis

In March 2026, we published our original bull flag analysis using 16 million chart embeddings. Since then, our database has grown to 24 million embeddings spanning 10 years and 19,000+ symbols. We also now have 16,000+ forward test predictions with tracked actuals — real, timestamped predictions made before the market opened.

This is not a rehash of the original post. The expanded dataset changes some of the numbers meaningfully, and the forward test data lets us report out-of-sample accuracy for the first time.

How We Define Bull Flags (and Why It Matters)

Chart Library's pattern detector identifies bull flags using structural analysis: a strong initial move (the pole) followed by a consolidation that retraces less than 50% of the pole on declining relative volume. The detector assigns confidence scores from 0 to 1.

For this analysis, we used two thresholds: high-confidence flags (score above 0.7) and moderate-confidence flags (score 0.5-0.7). The difference in outcomes between these groups is one of the most actionable findings in the data.

The Numbers: High-Confidence Bull Flags

These are bull flags with a detector confidence score above 0.7 — tight consolidation, clear pole, declining flag volume. Forward returns measured from the end of the flag consolidation:

• 1-day forward return: +0.48% average (57% win rate) vs. +0.04% base rate
• 3-day forward return: +1.02% average (56% win rate) vs. +0.12% base rate
• 5-day forward return: +1.41% average (55% win rate) vs. +0.21% base rate
• 10-day forward return: +1.89% average (54% win rate) vs. +0.43% base rate

Moderate-Confidence Bull Flags Tell a Different Story

Bull flags with detector confidence between 0.5 and 0.7 — the "it sort of looks like a bull flag" category — produce dramatically weaker results:

• 1-day forward return: +0.18% average (53% win rate)
• 3-day forward return: +0.35% average (52% win rate)
• 5-day forward return: +0.44% average (52% win rate)
• 10-day forward return: +0.61% average (51% win rate)

The Gap Between High and Moderate Confidence

The 5-day return gap between high-confidence (+1.41%) and moderate-confidence (+0.44%) bull flags is a 3x difference. This is the single most important finding for traders: not all bull flags are worth trading. The quality of the consolidation phase — how tight it is, how cleanly volume declines, how little of the pole gets retraced — directly predicts outcomes.

If you are scanning for bull flags, applying a confidence filter is more important than any other parameter. A tight flag on a strong pole is genuinely predictive. A sloppy consolidation after a weak move is noise.

Bull Flags by Market Regime

Chart Library's regime tracker classifies market conditions using historical pattern similarity on SPY and QQQ. We bucketed bull flag outcomes by the prevailing regime at the time of the pattern:

• Trending bull market (SPY above 50-day MA, VIX below 18): 5-day return +1.72%, 58% win rate
• Choppy/sideways market (SPY near 50-day MA, VIX 18-25): 5-day return +0.83%, 53% win rate
• Elevated volatility (VIX above 25): 5-day return +0.51%, 50% win rate — no edge
• Bear market (SPY below 200-day MA): 5-day return -0.12%, 47% win rate — bull flags fail in downtrends

Sector Breakdown

Bull flag performance varies significantly by sector. The strongest results come from momentum sectors, while defensive sectors produce near-random outcomes:

• Technology (XLK): +1.83% avg 5-day return, 57% win rate — the best sector for bull flags
• Consumer Discretionary (XLY): +1.52% avg, 56% win rate
• Industrials (XLI): +1.21% avg, 54% win rate
• Healthcare (XLV): +0.94% avg, 53% win rate
• Utilities (XLU): +0.31% avg, 51% win rate — essentially random
• Consumer Staples (XLP): +0.28% avg, 50% win rate — no edge

Out-of-Sample Results From Forward Testing

Chart Library runs automated forward tests every trading day, generating predictions before the market opens and tracking actual outcomes. Of our 16,000+ tracked predictions, bull flag patterns identified by the detector show 5-day direction accuracy of 54.8% out-of-sample — consistent with the in-sample numbers above.

This matters because most pattern statistics you see online are in-sample only. Our forward test data is timestamped, automated, and publicly visible on the Track Record page at chartlibrary.io/accuracy. No cherry-picking.

How to Apply This Data

The actionable takeaways from 24 million patterns of bull flag data:

• Only trade high-confidence bull flags (tight flag, declining volume, shallow retracement)
• Check the market regime first — bull flags in downtrends are traps
• Favor momentum sectors (tech, discretionary) over defensive sectors
• The 5-day horizon offers the best signal-to-noise ratio
• Use the Trade Simulator to test specific stop/target combinations on your exact setup
• Always compare to base rates — a 55% win rate is only meaningful if the base is 52%

Search Your Bull Flag

Next time you spot a bull flag, search it on Chart Library. You will see the 10 most similar historical patterns, their forward returns, and an AI summary of the setup. It takes 3 seconds and gives you a data-driven second opinion.