Signal Settings

Overview

Signal Settings for Lune Vertex allow you to control the generation and behavior of trade entry signals. These settings provide a high degree of customization, letting you fine-tune the strategy’s responsiveness by leveraging its advanced, AI-driven engine. By configuring the prediction model, selecting market features, and adjusting sensitivity, you can tailor the signal logic to fit your specific trading style and the market you are analyzing.

Signal Generation Logic

To protect our proprietary algorithms, the exact mechanics of the signal logic are not disclosed. However, the conceptual approach can be understood as follows:

Lune Vertex uses a sophisticated classification AI model that categorizes the market into three distinct states: bullish, bearish, or neutral. The strategy dynamically builds a feature vector from a wide range of market characteristics—such as volatility, trend memory, and price efficiency—based on your selections.

The model learns online, meaning it continuously updates its understanding of how these features influence future price movements, allowing it to adapt to changing market conditions. Signals are generated only when the model has a high conviction that the market is entering a strong bullish or bearish phase, and it will ignore periods of uncertainty. This adaptive approach ensures the signal logic remains effective and relevant. All signals are confirmed on the close of a price bar to ensure they do not repaint.

Settings

The following settings control the core logic of the signal engine.

General Settings

Setting
Description

Long Trades

Enables or disables the generation of long (buy) signals.

Short Trades

Enables or disables the generation of short (sell) signals.

AI & Sensitivity

These settings define the AI's prediction target and its overall responsiveness.

Setting
Description
Options / Recommended

Signal Confidence Threshold

The minimum required signal confidence for a trade entry. A higher value requires stronger confirmation from the AI before a signal is generated.

  • Range: 0.01 - 0.99

  • Recommended: 0.6 - 0.8

Analysis Lookback Window

Sets the size of the sliding data window the AI model uses for its analysis. A shorter window adapts faster to recent market changes.

  • Range: 5 - 2000

  • Recommended: 300 - 800

Prediction Horizon

Sets the number of bars ahead for the AI to predict. A shorter horizon is more reactive, while a longer horizon is better for capturing larger moves.

  • Range: 1 - 50

  • Recommended: 3 - 15

Minimum Desired Move

Sets a volatility-normalized minimum move required for a signal, using the Average True Range (ATR). A value of 1.5 means the predicted move must be at least 1.5x the current ATR.

  • Range: 0.1 - 10.0

  • Recommended: 1.0 - 3.0

Learning Rate

Controls how quickly the AI model adapts to new market data. Lower values result in slower, more stable adaptation, while higher values lead to faster but potentially less stable learning.

  • Range: 0.01 - 0.20

  • Recommended: 0.03 - 0.10

Dynamic Feature Selection

Select up to eight market features for the AI model to analyze. The model will dynamically weigh the importance of each selected feature.

Setting
Description
Options

Feature 1-8 Type

Selects the type of market feature to analyze. Each option captures a different market characteristic to enhance prediction accuracy.

Trend Filtering, Trend Strength, Trend Memory, Volatility Regimes, Volatility Analysis, Price Filtering, Volatility Smoothing, Volatility Dynamics, Pattern Detection, Price Efficiency, Distribution Skewness, Distribution Kurtosis, Distribution Kurtosis Alt, Risk Analysis, Downside Risk Ratio, Sharpe Ratio, Volume Analysis, Flow Analysis, Price Anchoring, Range Analysis, Signal Entropy, Variance Analysis, Jump Detection, Jump Analysis, Jump Filtering, Higher Moments, Higher Moments Alt, Normalized Volatility, Return Correlation, Return Memory, Return Memory Alt, Sign Correlation, Return Smoothing, Momentum Analysis, Volume-Volatility, Volume Dynamics, Time Trend Correlation

Feature 1-8 Lookback

Sets the number of bars used for the feature's calculation. A lower value is more responsive to recent data.

  • Range: 1 - 2000

  • Recommended: 50 - 250

Advanced Signal Settings

These settings enable advanced, experimental systems for signal processing. Use at your own risk.

Setting
Description
Range / Recommended

Overfitting Protection

Strength of the regularization mechanism to prevent the AI from overfitting to market noise. Higher values create simpler, more generalized models.

  • Range: 0.0 - 0.1

  • Recommended: 0.0001 - 0.01

Gradient Clipping

Clips the internal learning gradients to prevent extreme updates and improve the model's stability.

  • Range: 0.0 - 10.0

  • Recommended: 0.5 - 2.0

Best Practices & Usage

  • Select Diverse Features: When choosing features, select a varied set that captures different aspects of the market (e.g., one for trend, one for volatility, one for volume). This allows the model to build a more robust understanding.

  • Balance Confidence and Signal Frequency: A lower Signal Confidence Threshold will generate more signals but may also increase the number of false positives. Higher values provide stronger confirmation but may result in fewer trading opportunities. Find a balance that suits your risk tolerance.

  • Adjust Lookbacks for Your Timeframe: If you are trading on a lower timeframe, consider using shorter Lookback periods to make the strategy more responsive. For higher timeframes, longer Lookback periods can provide more stable and reliable signals.

  • Start with Recommended Values: The recommended values provide a solid starting point for most markets. Use them as a baseline and then carefully adjust them based on the specific asset and timeframe you are trading.

  • Tune One Thing at a Time: When optimizing the settings, adjust only one parameter at a time. This will help you clearly understand the effect of each change on the strategy's performance during backtesting.

Last updated

Was this helpful?