Predictive Power Analysis

Developing a successful trading algorithm hinges on identifying signals with genuine predictive power. Spectral Alpha provides a dedicated engine to help you measure and understand the effectiveness of your signals.

How the Engine Works

1. Price Change Measurement: The engine analyzes the price change that occurs after a signal is triggered, tracking this change for every bar up to a defined cutoff point.
2. Averaging Price Changes: It calculates the average price change at each bar across all triggered trades. This provides a clearer picture of the signal's overall impact on price movement.
3. Oversampling for Significance: To enhance statistical significance, the engine generates oversampled market datasets. This increases the number of trades triggered, providing a more robust assessment of the signal's performance.
4. Parameter Sweeping: The engine sweeps through the defined range of parameter values (from minimum to maximum). This helps eliminate the influence of lucky initial values and reveals the robustness of the signal across different parameter settings.

Controlling the Analysis

To run this analysis, you need to specify:

1. Analysis Name: Give your analysis a descriptive name for easy identification.
2. Parameter Selection: Choose the specific parameter you want to analyze.
3. Trade Lifetime: Set the maximum trade lifetime (up to 500 bars). This determines how long the engine tracks price changes after a signal.

Understanding the Results

The generated report presents separate analyses for long and short trades, providing a comprehensive view of the signal's performance in both directions.

• Y-Axis: Displays the different parameter values used for each run.
• X-Axis: Shows the bar number since the signal trigger.
• Color Scheme: Indicates whether the average price change is favorable or unfavorable to the signal direction. For example, green for long trades means the market moved up after the signal, while red indicates the market moved down.

Interpreting the Results

• Trend-Following Signals: These signals typically don't predict short-term volatility, so they might initially show unfavorable price action.
• Short-Term Signals: Signals based on cycles or mean reversion should show favorable price action soon after the trigger, followed by a decay in predictive power.
• Lack of Predictive Ability: If your long/short signals are consistently red, or are erratic, it suggests that the signal might not have any predictive ability.

By carefully analyzing the results of the Signal Predictive Power Analysis, you can gain valuable insights into the effectiveness of your signals and refine your trading strategies accordingly.