AI Signal Filter
Overview
The AI Signal Filter is an advanced feature designed for users who want to refine their trading strategy's performance. It uses machine learning to analyze market patterns and historical trade data to filter out low-quality or false signals. The goal is to improve the consistency and profitability of the strategy by only taking trades that meet a higher standard of confidence.
This feature is intended for advanced users. Most strategies can be highly profitable without it, but it provides an extra layer of analysis for those seeking maximum optimization.
Settings
Main Configuration
These settings control the overall behavior of the filter.
AI Signal Filter Mode
Determines the method the AI uses to filter signals.
Disable: The filter is turned off.
Signal: Analyzes the characteristics of each potential signal.
Trade Performance: Analyzes patterns from the strategy's historical trades.
Similarity Metric
Controls how the filter measures similarity between the current market pattern and past patterns.
Standard: Gives equal weight to all parts of the pattern.
Trend-Weighted: Focuses more on trend-related data.
Volatility-Weighted: Focuses more on volatility-related data.
Comparison Analysis
Defines how the filter combines the results from similar past situations to make a decision.
Weighted Average: Averages the outcomes, giving more weight to more similar patterns.
Outcome Consensus: Looks for a general agreement among the outcomes of similar patterns.
Core Filter Parameters
These inputs fine-tune the filter's core logic.
Min Confidence Score
Sets the minimum confidence score required for a signal to be approved. Higher values make the filter stricter.
Range: 0.0 - 1.0 Recommended: 0.30 - 0.70
Comparison Count
The number of similar past events to compare against. More events provide more confirmation but take longer to process.
Range: 3 - 50 Recommended: 10 - 30
Historical Lookback
The number of recent bars to use as a historical reference.
Range: 50 - 2000 Recommended: 500 - 1500
Outcome & Pattern Analysis
These settings define how the filter evaluates outcomes and recognizes patterns.
Outcome Mode
Controls how the period for evaluating a signal's success is determined.
Static: Uses a fixed period.
Adaptive: Adjusts the period based on market volatility.
Trend-Adjusted: Adjusts based on trend strength.
Complexity-Adjusted: Adjusts based on market predictability.
Outcome Period
The number of bars to look into the future to define a successful or unsuccessful outcome.
Range: 1 - 50 Recommended: 3 - 15
Pattern Mode
Defines how the filter selects patterns for comparison.
Static: Uses a fixed pattern approach.
Adaptive: Adjusts patterns based on market conditions.
Market-Weighted: Weights patterns by market regime.
Dominant-Factor: Focuses on the most important patterns.
Shape-Based: Emphasizes geometric patterns.
Predictability-Based: Focuses on patterns with the highest predictive power.
Pattern Sensitivity
Controls the lookback period for the pattern recognition logic. Lower values are more responsive.
Range: 1 - 10 Recommended: 3 - 7
Advanced Thresholds
These are additional on/off filters that can be enabled for more specific criteria.
Similarity Distance
When enabled, this filter only allows signals where the current market pattern is highly similar to historical patterns. You can set the Max Distance to control how strict the similarity match must be (lower is stricter).
Range: 0.5 - 15.0 Recommended: 3.0 - 8.0
Potential Move
When enabled, this filter analyzes the potential price movement seen in similar past situations. It only approves signals with a sufficient historical move, defined by the Min Move (as an ATR multiplier).
Range: 0.5 - 10.0 Recommended: 1.5 - 4.0
Risk/Reward
When enabled, this filter evaluates the historical risk-to-reward ratio from similar patterns. It only approves signals with a favorable ratio, defined by the Min Ratio threshold.
Range: 0.5 - 10.0 Recommended: 1.5 - 4.0
Best Practices & Usage
Start with the Filter Disabled: First, find a profitable base strategy using the other settings. The AI Signal Filter should be used for fine-tuning, not for fixing a broken strategy.
Enable and Test: Once you have a solid strategy, enable the filter using the
SignalorTrade Performancemode. Start with the recommended default values.Adjust One Setting at a Time: The filter is complex. To understand the impact of each setting, only change one at a time and observe the results in the Strategy Tester.
Use the Confidence Score as Your Main Control: The
Min Confidence Scoreis the most direct way to control how strict the filter is. Increase it to approve fewer, higher-quality signals, or decrease it to allow more signals through.Balance Performance and Sample Size: A very strict filter might produce a high win rate and profit factor but on a very small number of trades. Ensure you have enough
Total Tradesfor the results to be statistically meaningful.
Last updated
Was this helpful?