KNN Machine Learning Mean Reversion Probability

The KNN Machine Learning Mean Reversion Probability indicator utilizes a K-Nearest Neighbors (KNN) algorithm to calculate the statistical probability of price returning to its moving average within a specific timeframe. By analyzing historical price extensions and momentum exhaustion, the tool provides a probabilistic confidence score for mean reversion setups.

Usage

The Usage section describes how the script can be used to identify potential snap-back opportunities when price becomes overextended.

• Signal Interpretation: When the calculated probability crosses the user-defined threshold and price passes the "Extension Gate," a signal (▲ Rev or ▼ Rev) is generated. These signals include a label showing the exact historical probability of a reversion occurring.
• Snap Zone Fill: When a signal is active, the area between the current price and the basis MA is shaded. This serves as the target zone for the mean reversion trade. The shading disappears once price touches the basis MA.
• Bar Coloring: The script dynamically colors bars based on probability strength. Bright colors indicate probabilities above the threshold, while dimmed colors suggest approaching signal territory.
• Recommended Workflow: Users typically set the Basis MA to a preferred value (such as a 20-period EMA) and tune the Reversion Window to match their expected holding time. The Probability Threshold can be increased (e.g., to 0.75) for more selective, higher-confidence signals.

Details

The indicator follows a supervised machine-learning pipeline to generate its outputs:

• Labeling: Historical bars are labeled as "successful" or "unsuccessful" based on whether price touched the basis MA within the defined Reversion Window.
• Feature Engineering: Five normalized features are used to define the "state" of the market: Distance from MA, Bollinger Band position, RSI deviation, Candle Body compression, and Volume fade.
• KNN Engine: The algorithm uses Minkowski Distance to find the $K$ most similar historical analogs. A Gaussian Kernel is applied so that closer neighbors have a higher weight in the final probability calculation.
• Extension Gate: This logic ensures that signals only fire when price is at a minimum distance from the MA (defined by ATR). This prevents signals in low-volatility or sideways markets where mean reversion has less statistical edge.

Settings

KNN Engine

• K Neighbors: The number of historical analogs used for voting.
• Lookback Window: The number of historical bars the algorithm searches through.
• Reversion Window: The maximum number of bars allowed for a price to touch the MA for it to be considered a successful reversion.
• Minkowski p: The exponent for the distance metric (1 for Manhattan, 2 for Euclidean).
• Gaussian Bandwidth: Determines how much weight is lost as historical neighbors become less similar.
• Probability Threshold: The minimum probability required to trigger a signal.

Feature Settings

• Basis MA Type/Length: Selects the moving average type (SMA, EMA, etc.) and period used as the mean.
• Bollinger Band mult: The standard deviation multiplier for volatility-adjusted extension features.
• RSI length: The period used for the RSI exhaustion feature.
• Volume MA length: The period used to determine the volume baseline for the volume fade feature.

Extension Gate

• Require extension gate: Enables or disables the volatility-based signal filter.
• Gate band multiplier: How many ATRs price must be away from the MA to allow a signal.
• Gate ATR length: The period used for the Average True Range calculation.

FAQ

How does this differ from standard momentum KNN indicators? While most KNN indicators try to predict the direction of the next candle, this tool specifically measures the likelihood of price returning to its average "fair value" after a period of overextension.

Why are there two separate probability readings? The script tracks probabilities for "reversion from above" and "reversion from below" separately because market behavior during bull-side and bear-side extensions is often statistically asymmetrical.

How can I access this indicator? You can get access on the LuxAlgo Library for charting platforms like TradingView, MetaTrader (MT4/MT5), and NinjaTrader for free.