Shipped a data study: we tested 17,581 Bear Flag chart patterns to see if the textbook pattern actually works in crypto.
Backstory: ChartScout detects chart patterns at scale across crypto markets. Instead of trusting textbook pattern theory, we ran real backtests 59 Binance markets, 15 timeframes, 309.7M scan windows evaluated.
Findings: only 16.59% of Bear Flags declined 5%+ overall (95% CI 16.05–17.15%), but win rate climbs to 44.49% at 15m+ timeframes. We also benchmarked against Bulkowski's classic stock-market study (55% success rate) crypto clearly behaves differently from equities.
Full breakdown (with methodology, confidence intervals, and market-regime splits): https://chartscout.io/bear-flag-win-rate-crypto-study
Would love feedback from anyone here who's done pattern backtesting before.
What makes this study more interesting than the headline result is that the data seems to push against the textbook definition rather than simply validate it. The difference between the overall result and the timeframe-specific result makes the pattern itself look much less like a fixed signal and much more dependent on context.