D21-F01-A03 · four-stage decision lab · synthetic teaching data

Through-the-Cycle PD

Calculate a transparent long-run average one-year PD, retain the annual rates and pooled-rate comparison, and expose the representativeness assumptions behind through-the-cycle calibration.

Question: How can annual one-year default experience be summarized into a long-run grade-level PD without hiding yearly variation?

Stage 1 · orient

Choose the analytical question before touching the number

Decision boundary

Invariant:

Stage 2 · predict

Commit to a direction before revealing the challenge

Your prediction
Stage 3 · experiment

Move one declared driver; keep the scenario contract fixed

Complete 61-state output series and evidence trace

Scenario result traceAll independently calculated states with the selected state marked.

Stage 4 · explain

Compare the anchor and candidate before interpreting

Anchor result
Candidate result
ComponentAnchorCandidateWhat changed?

Misconception check:

Inspect the structured input and output

Input

Output


Scenario evidence index
The lab surfaces curated stops for instruction while preserving all seven scenarios × 61 independently calculated states for audit.
ScenarioDriverQuestionRangeDecision segments