Minimum confidence level the algorithm must reach before emitting a signal. Higher values produce fewer but more selective signals. Lower values increase signal frequency, including borderline setups.

Filters random price fluctuations from the tick matrix before pattern analysis. Higher values retain only clean directional moves and remove low-quality data that can bias signal generation.

Controls how deep the algorithm scans the order book before confirming entry conditions. Higher depth means stronger validation of the move and less exposure to thin liquidity zones.

Suppresses signals during abnormal volatility spikes caused by news events or macroeconomic releases. Reduces exposure on gaps and unpredictable price action that bypasses the detection model.

Neural network weights update in real time to adapt to the current session's market behavior. Improves accuracy when the market structure shifts mid-session without a full model reload.

Delays the entry point by 1-2 seconds to confirm that momentum is fully aligned before emission. Slightly reduces signal speed in exchange for better entry timing on fast-moving instruments.

Limits signal analysis to London and New York trading sessions only. Eliminates low-liquidity Asian session periods where spread anomalies and thin order books reduce signal reliability.

Controls how the model balances historical pattern weight against live market flow data. Conservative minimizes false positives with proven setups. Aggressive captures faster-forming patterns with higher variance.

Input timeframe for the prediction matrix. M1 provides the fastest reaction to real-time market changes. M15 delivers greater stability on slower pairs and during low-volatility sessions.

Historical dataset depth the engine references when building signal predictions. Longer memory improves accuracy on rare formations but adapts more slowly to recent market structure shifts.