Bollinger Band Bounce

Yes, in strong trends price can pierce and stay outside the bands for extended periods. This is called 'walking the bands.' In an uptrend, price may ride the upper band; in a downtrend, the lower band. This is why bounce strategies need trend context - don't automatically fade band touches in strong trends.

The middle band (20-period SMA) represents the 'fair value' or average price. In mean reversion, price tends to return to this average after reaching extremes. The middle band serves as both a target for bounce trades and a reference for trend direction (sloping up = uptrend, down = downtrend).

Both are volatility-based channels, but they use different volatility measures. Bollinger uses standard deviation (more responsive to recent moves); Keltner uses ATR (smoother). Bollinger Bands expand/contract more dramatically. Some traders use both together - when Bollinger is inside Keltner (squeeze), expect breakout.

2 standard deviations is standard and captures ~95% of price action. Use 2σ as your baseline. 3σ captures ~99.7% of prices, meaning touches are rarer but more extreme. Use 3σ for very volatile instruments or when you only want to trade extreme conditions. Test both on your specific instrument.

Generally yes. A mid-candle band touch might be just a wick that closes back inside. Waiting for candle close confirms the band interaction. For the conservative approach, wait for a candle to pierce the band, then the next candle to close back inside. This confirms rejection of the extreme.

A squeeze occurs when bandwidth (Upper - Lower / Middle) reaches a multi-period low. Many platforms have squeeze indicators. Visually, bands look very narrow compared to recent history. The squeeze indicates volatility compression that typically precedes a breakout. During squeezes, avoid bounce trades - look for breakout strategies instead.

In strong trends, the momentum overwhelming overrides mean reversion. Price 'walks the bands' - continuously touches one band because trend pressure keeps pushing. The statistical mean (middle band) is a moving target in trends. ADX > 25 indicates strong trend where bounces are less reliable.

Use %B thresholds for quantifiable signals. %B < 0 means price is below the lower band (pierced). %B crossing back above 0 confirms the bounce is starting. Similarly, %B > 1 means above upper band. %B is useful for scanning and automation since it's a single number rather than visual band relationship.

The 20-period setting works across most timeframes because it measures relative volatility. However, shorter timeframes (intraday) may benefit from shorter periods (10-15) for responsiveness. Longer timeframes (weekly) may use longer periods (30) for smoothness. The deviation (2σ) usually stays consistent.

Common approaches: (1) Below the low of the candle that touched/pierced the band (specific to the setup). (2) A fixed distance below the band (e.g., 0.5× ATR beyond band). (3) Below the lower band by a percentage (e.g., 1% below lower band). The key is giving enough room for normal volatility while limiting loss if bounce fails.

Adaptive systems adjust parameters based on market conditions. For deviation: Use ATR ratio (current vs average) to scale the multiplier. For period: Use efficiency ratio or ADX to determine optimal length. Implementation requires coding custom indicators. Backtest adaptive vs static to confirm improvement, and walk-forward validate.

The squeeze indicator compares Bollinger Bands to Keltner Channels. When Bollinger is inside Keltner (BB narrower), it's a squeeze - extreme low volatility expecting breakout. When Bollinger is outside Keltner (BB wider), volatility is expanded. Use this to filter: only trade bounces when Bollinger is outside Keltner (no squeeze).

Professionals often use BB as one component in multi-factor models. They might combine %B with trend, momentum, and volume factors. Position sizing often scales with bandwidth - smaller in squeezes, larger in expanded bands. Everything is backtested across instruments and regimes with walk-forward validation.

Classification works well: predict bounce success (price reaches middle band) or failure. Use features: %B, bandwidth, bandwidth percentile, ADX, RSI, volume ratio, candlestick pattern, higher TF position. Train Random Forest or XGBoost. Set probability threshold (>60%) to filter signals. Walk-forward validate and monitor accuracy in live trading.

Key practices: (1) Use standard parameters (20, 2) as baseline. (2) Test nearby parameters for robustness - if only one exact setting works, it's overfit. (3) Walk-forward optimize: optimize on period 1, test on period 2, repeat. (4) Keep rules simple. (5) Accept somewhat lower backtested returns for real-world robustness.