Signal Backtest Engine
Test how any GMIIE signal dimension correlates with article importance across historical data. Select a dimension, set a threshold, and measure whether a high signal score actually predicted high-impact articles.
Hit Rate
Percentage of articles in the lookback window where the chosen dimension crossed your threshold.
Importance Lift
Difference in average importance score between triggered articles and the baseline (below-threshold) group.
Top Hits
The highest-importance articles from the triggered group — the signal's best calls in the period.
Parameters
050100
Select a signal dimension and threshold above, then run the backtest to see how that signal correlated with article importance scores.
Interpretation Notes
- Signal scores (0–100) are generated per-article when content is processed. A score of 60+ generally indicates meaningful relevance for that dimension.
- Importance scores (0–10) reflect the article's actual significance as assessed by the pipeline. Scores ≥ 7 are high-impact; ≥ 9 are critical.
- A high lift (triggered avg > baseline avg) suggests the dimension is a useful leading indicator. Low or negative lift suggests the signal needs calibration.
- Use the lookback window to test signal stability across different time periods — a dimension that performs well in one period may be noisier in others.