Neural Volatility Clustering: Filtering Market Noise on TradingView

Financial asset returns exhibit an empirical property known as volatility clustering: large price changes tend to be followed by large changes, and small price changes tend to be followed by small changes. In retail trading, indicators that do not adapt to regime transitions generate severe losses during low-volatility compression chop.

By applying neural clustering principles to Normalized True Range (NATR) and standard deviation variance, we can dynamically classify market state into three distinct regimes: Compression, Expansion, and Exhaustion.

The Three Volatility Regimes

• Compression (Chop / Distribution): Variance drops below rolling historical percentiles. All breakout and trend signals are automatically filtered out to prevent whipsaws.
• Expansion (Impulse Wave): Volatility surges above dynamic variance bands with expanding volume. This is the only regime where momentum breakouts and trend-continuation entries are valid.
• Exhaustion (Climax Run): Volatility reaches extreme standard deviations (e.g., >3.0σ) away from the mean. Trailing stops tighten automatically to lock in maximum equity.

Figure 1: Adaptive volatility regime detection filtering out false breakouts during London-NY session overlaps.

Mathematical Pine Script v6 Architecture

Here is an adaptive volatility filter in Pine Script v6 that calculates a normalized percentile rank of current market dispersion: