Machine Learning: Optimal RSI

The Machine Learning: Optimal RSI indicator identifies the most profitable RSI length based on historical performance and applies machine learning logic to create an adaptive, smoothed oscillator. It dynamically scans multiple RSI lengths to determine which RSI-to-MA cross produces the highest return, offering a data-driven alternative to static RSI periods.

Usage

The indicator can be used to identify trend shifts and overbought/oversold conditions using an optimized RSI length that adapts to current market volatility.

• Optimal RSI Selection: The script scans a range of RSI lengths (starting from a user-defined minimum) over a specific lookback period. It tracks the simulated profitability of every RSI/MA cross and selects the length that performed best.
• Machine Learning Integration: Once the optimal RSI is found, it is processed through a Machine Learning (ML) model—either a Simple Average or a k-Nearest Neighbor (KNN) algorithm. This creates a smoothed version of the RSI that filters out noise while remaining responsive to price action.
• Dynamic Bollinger Bands: Unlike static levels (30/70), the tool provides RSI-based Bollinger Bands. These act as adaptive support and resistance levels for the RSI line, helping traders spot exhaustion points that static levels might miss.
• Filtering: Users can toggle "Filter RSI and RSI MA," which ensures the ML calculation only considers data points where the trend direction (RSI vs. RSI MA) matches the current optimal RSI direction.

Details

The script's core logic revolves around an "Optimal RSI" selection process. Traders often default to a 14-period RSI, but different assets and timeframes often respond better to different lengths.

By default, the script uses All Crossings to evaluate profit, but it can be set to Extremity Crossings. In the latter mode, the script only considers profitability when an RSI cross occurs at extreme levels (below 40 for longs or above 60 for shorts).

The KNN (k-Nearest Neighbor) functionality allows the indicator to "learn" from historical distances between fast and slow moving averages of the RSI. It identifies similar historical patterns to predict the current smoothed value, offering a more sophisticated alternative to simple moving averages. To prevent script execution timeouts on lower timeframes or data-heavy charts, the indicator includes an "Auto" adjustment feature for the lookback length and scan count.

Settings

RSI Settings

• Show Crossing Signals: Toggles visual buy/sell markers when the RSI and its MA cross.
• Show Tables: Displays a dashboard showing the currently selected optimal RSI length and its calculated profit.
• Show Bollinger Bands: Displays adaptive volatility bands around the RSI.
• Optimal RSI Type: Choose between "All Crossings" or "Extremity Crossings" for profit evaluation.
• Adjust Optimal RSI Lookback and RSI Count: Set to "Auto" to let the script manage performance or "Manual" for custom lookback periods.
• Optimal RSI Lookback Length: The number of bars the script analyzes to find the best-performing RSI length.
• RSI Count: The total number of RSI lengths to test (e.g., if min is 4 and count is 30, it tests lengths 4 through 34).
• RSI Minimum Length: The starting period for the RSI scan.
• RSI MA Length: The period of the Moving Average used for RSI crossing logic.
• Extremity Crossings RSI Backup Length: The default length used if no optimal length is found during extremity scans.

Machine Learning

• Use Rational Quadratics: Applies a rational quadratic kernel to the price source for smoother ML input.
• Filter RSI and RSI MA: If enabled, ML calculations only include historical data with the same bullish/bearish RSI-to-MA relationship.
• Machine Learning Type: Select the smoothing algorithm: KNN Average, KNN Exponential Average, Simple Average, or None.
• KNN Distance Type: Determines whether the KNN algorithm looks for Min, Max, or Both distances between data points.
• Machine Learning Length: The lookback window for the ML data set.
• k-Nearest Neighbour (KNN) Length: The number of "neighbors" (similar historical points) the KNN algorithm considers.
• Fast/Slow ML Data Length: Periods used to calculate the distances for KNN logic.

FAQ

How does the "Auto" adjustment for Lookback work? The script detects your timeframe and bar count to prevent TradingView timeout errors. It automatically lowers the scan depth on lower timeframes and increases it on higher timeframes to ensure the script executes within the platform's processing limits.

What is the benefit of using KNN over a Simple Average? A Simple Average gives equal weight to all recent data points, whereas KNN looks for historical instances that specifically resemble the current market structure (based on the distance between fast and slow RSI averages), potentially providing a more context-aware smoothing.

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