Lorentzian Classification Strategy

The Lorentzian Classification Strategy indicator provides a comprehensive trading system based on machine learning and multi-filter technical analysis to identify and capitalize on trending market moves. It combines advanced Lorentzian classification with trend filters and risk management tools to offer a robust backtesting and live-analysis environment.

Usage

The Usage section highlights how to interpret the signals generated by the integration of Lorentzian classification, moving averages, and volatility filters.

• Long Positions: A long signal is generated when the close price is above the 200-period EMA and the Lorentzian Classification model issues a buy signal.
• Short Positions: A short signal is generated when the close price is below the 200-period EMA and the model issues a sell signal.
• Exit Strategy: The strategy employs a dual-exit approach. Half of the position is closed at a fixed 3:1 risk/reward ratio. The remaining half is closed when either the model provides an opposing signal or the integrated Supertrend filter indicates a trend reversal.
• Stop Loss & Break Even: Initial stop losses are determined by an ATR-based stop loss or recent swing highs/lows. A break-even point is automatically triggered once a 1:1 risk/reward ratio is achieved.
• Risk Management: Users can define a specific percentage of their account to risk per trade. The indicator includes a position size calculator that displays the required USD amount in the top corner of the chart to maintain consistent risk.

Details

The strategy is built upon the Lorentzian Classification model, which utilizes a K-Nearest Neighbors (KNN) approach modified for Lorentzian space to classify market regimes.

Key implementation details include:

• Feature Engineering: The model uses five customizable technical features (RSI, WaveTrend, CCI, and ADX) to categorize market data.
• Filters: To reduce noise, the strategy incorporates a Volatility Filter, ADX Filter, and Regime Filter (Kernel Regression) to ensure trades are only taken during high-probability trending periods.
• Backtesting Utilities: The script includes advanced logic for simulating capital additions or withdrawals (fixed or percentage-based) over time, as well as leverage simulation for more realistic performance tracking.

Settings

General Settings

• Source: The data input for the ML model (usually the close price).
• Neighbors Count: The number of historical neighbors the KNN model evaluates.
• Max Bars Back: Controls the range of history used for calculations; reducing this can improve loading speeds.

Feature Engineering

• Feature 1-5: Allows the selection of RSI, WT, CCI, or ADX as inputs for the machine learning model.
• Parameter A/B: Adjusts the internal periods or settings for each selected feature.

Filters

• Use EMA/SMA Filter: Toggles the use of moving averages to ensure trades align with the long-term trend.
• Regime/ADX Threshold: Sets the sensitivity for trend detection; lower thresholds allow more trades, while higher thresholds are more restrictive.

Stop Loss & Risk Management

• ATR Multiplier: Sets the distance for the ATR stop loss.
• Stoploss Source: Switches between ATR-based stops or Swing High/Low detection.
• % Account Risk per Trade: The percentage of equity to be risked on a single trade.
• Use Leverage: If enabled, the strategy will simulate using borrowed funds if the required position size exceeds current capital.

Display Settings

• Show Supertrend/EMA: Toggles the visibility of the underlying filter indicators.
• Show Bar Predictions: Displays the raw integer output of the ML model on each bar.
• Date/Session Filters: Restricts backtesting and signal generation to specific dates or hours of the day.

FAQ

How does the Lorentzian Classification differ from standard KNN? Lorentzian Classification uses a distance metric better suited for non-linear, "fat-tailed" financial data, potentially reducing the impact of outliers compared to standard Euclidean distance.

Why are signals not appearing on every trend change? Signals only appear when all conditions—ML classification, EMA trend alignment, and volatility filters—are met simultaneously to ensure higher quality entries.

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