Chart Pattern Scanner

Valid patterns are: 1) Clear and obvious without forcing, 2) Have proper structure (e.g., H&S has head higher than shoulders), 3) Volume profile supports (usually decreasing during formation), 4) Fit the market context (reversal patterns after trends, continuation within trends).

Generally no - wait for confirmation (close beyond pattern boundary). Trading before breakout means higher risk of pattern failure. Exception: experienced traders may position early at key levels within the pattern with tighter risk management.

Daily charts are most reliable - enough data to form meaningful patterns, less noise than intraday. Weekly for major patterns, 4H for trading, 1H for day trading. Lower timeframes have more noise and false patterns.

Patterns fail due to: 1) External news/events, 2) Poor pattern quality (forced, not obvious), 3) False breaks (stops triggered, then reverses), 4) Counter-trend patterns (fighting major trend). This is why we use stops and wait for confirmation.

Continuation patterns (flags, pennants, triangles within trends) form during a trend and suggest it will continue. Reversal patterns (H&S, double top/bottom) form at the end of a trend and suggest direction change. Context matters - same shape can be either depending on where it forms.

Measured move = pattern height projected from breakout. H&S: head to neckline distance down from neckline break. Double top: peak to valley distance down from valley break. Triangle: height at widest point projected from breakout. Flag: pole length from flag breakout.

False breaks are common. Mitigate by: 1) Wait for candle close beyond level, 2) Require volume confirmation, 3) Wait for retest of broken level, 4) Use stops - accept some will fail. If stopped out on false break, the pattern is invalidated.

Counter-trend patterns have lower success rates. Reversal patterns should only be traded when: 1) After extended trend, 2) Pattern is very clear and well-formed, 3) Volume and other factors confirm. Continuation patterns with the trend are generally higher probability.

Combine patterns with: 1) S/R levels - pattern at key level is stronger, 2) Volume analysis - confirms conviction, 3) Moving averages - context for trend, 4) Options flow - smart money confirmation. Multiple confluences increase probability.

Typically: decreasing volume during formation (consolidation/indecision), then volume spike on breakout (conviction). Exception: some patterns like triangles may have irregular volume. The key is increased volume on the breakout confirming commitment.

Architecture: 1) Data layer (price feed, historical), 2) Processing (swing detection via fractal/zigzag, template matching, quality scoring), 3) Signal layer (stage tracking, breakout detection), 4) Output (dashboard, alerts, API). Scan universe regularly, filter by quality, alert on setups.

Strong features: structural ratios (shoulder heights, time proportions), volume profile (during formation vs breakout), trend context (prior trend length/strength), volatility (ATR), and market regime. Target can be binary (success/fail) or continuous (% of target achieved).

Parameters to tune: swing detection sensitivity (N bars), pattern matching tolerance, quality score threshold, volume filter. Backtest across historical data. Walk-forward test to avoid overfitting. Find robust parameters that work across market conditions.

Track: detection accuracy (valid patterns identified), signal success rate (breakouts that reach target), P&L by pattern type. Analyze: which patterns work best, optimal filters, market condition impact. Improve: refine detection rules, adjust quality thresholds based on results.

Failed patterns can be opportunities: 1) Failed H&S breaking back above neckline is bullish, 2) Failed breakout reversing is signal in opposite direction. Track failure modes. Consider strategy that trades both successful and failed patterns with appropriate direction.