PulseGrid Accuracy Tracker
The Accuracy Tracker page provides a comprehensive scorecard of how well PulseGrid's parametric signals have predicted market direction when they diverged from market consensus.
Methodology
For each divergence event (where PG Signal and Market Signal disagree), the system tracks whether the market signal subsequently moved toward PG's predicted direction:
- Correct: The market signal moved toward PG's direction within the observation window (up to 10 subsequent data points)
- Incorrect: The market signal did not converge toward PG within the observation window, and sufficient follow-up data exists
- Pending: The divergence is too recent to determine an outcome (fewer than 5 subsequent data points)
Key Metrics
| Metric | Description | Benchmark |
|---|---|---|
| Overall Accuracy | Percentage of resolved divergences where PG was correct | Above 55% is statistically meaningful |
| Total Divergences | Count of all divergence events across all instruments | Higher count provides more statistical confidence |
| Divergence Rate | Percentage of signal readings that contained a divergence | Typical range: 15-40% |
Accuracy by Severity
The severity breakdown shows accuracy rates for minor, moderate, and significant divergences separately. Higher severity divergences typically carry stronger predictive power because they represent cases where PG's parametric analysis strongly disagrees with market consensus.
Per-Instrument Accuracy
The instrument table shows accuracy rates for each tracked asset. Instruments labeled "Prophetic" (accuracy above 55%) are those where PG's contrarian calls have been most reliably validated. "Neutral" (45-55%) indicates performance near random chance. "Contrarian" (below 45%) indicates the market has consistently moved against PG's predictions for that instrument.
Accuracy Trend
The weekly trend chart shows how PG's accuracy has evolved over time. Consistent performance above the 50% random-chance line indicates genuine predictive signal. The bar chart overlay shows the volume of divergences per week for context.
Limitations
- Accuracy is computed using a fixed look-ahead window, which may not capture longer-term convergence
- The observation window (10 data points) is a heuristic; different instruments may require different windows
- Past accuracy does not guarantee future predictive performance
- Small sample sizes for individual instruments may produce unreliable accuracy estimates