RUN DETAIL
Selected run
Loading the run…
CALIBRATION
Reliability
Reliability curve
Predicted probability against what happened. Points on the dashed diagonal are honest; above it the model was underconfident, below it overconfident.
Every bin: predicted range, answers that landed in it, observed rate.
Performance
Hover for exact values.
Drawdown
Decline from each peak.
Allocations
Weights after each decision. An invalid answer set holds the previous book.
What this run costs
Estimates are list-price equivalents, not invoices.