Model comparison
Decision Supervisor, LSTM, Logistic, and GJR-GARCH · Full paper evaluation sample
Aligned score histories
Decisions and observed outcomes
Current date range · Select a count to highlight its dates
Scores by observed outcome
Current date range · Each group sums to 100%
Comparison reported in the paper
Full evaluation period · Unchanged by date filters
Saved comparison from reported_comparison.csv. Values retain the paper’s reported precision.
Role results
Five published score series · Six market evaluations
Role evaluation
Full evaluation sample · Five published score series
Source: market_agent_metrics.csv. Cells link to role histories.
Supervisor by market
ROC AUC · Full sample
Risk score and threshold
Scores express downside-tail risk intensity. Dashed line: saved historical threshold. Select a date for its saved report.
Outcomes in this date range
Saved labels and threshold decisions
Alert = saved prediction 1. Observed anomaly = saved evaluation label 1.
Full-sample role metrics
These saved metrics do not change with the date filter.
Source: official role metrics.
Inspect the plotted records
Paired role scores
One point per saved target date
Absolute score gap = |horizontal − vertical|. The diagonal marks equal scores.
Month-by-role view
Target-date months; n reports available observations, including partial months.
Daily reports
Six saved role responses, prompts, and input charts per target date
[...] marks withheld source content.
Paper figures
Saved manuscript figures