Goal: preserve raw paired observations, classify interruptions and compare a genuinely prior prediction when available. Proposed codes: AC9SFU03, AC9SFI02 (actual only), AC9SFI03, AC9SFI04 (prior prediction only). Prepare: paired record, approved setup if still safe or full D, Card 114.
- Source/date · 3 min. Use the true local record if obtained; otherwise D MODEL. Do not “complete” missing real entries from D.
- Prior status · 4 min. Retrieve a prediction only if recorded before actual results; else NO PRIOR PREDICTION.
- Model raw · 5 min. Read first pair exactly, preserving surface and appearance. Explain that run-off is INTERRUPTED, not a hidden mark.
- Child try · 7 min. Adult safely performs another real pair if approved and feasible, or reads D's second pair and fifth entry. Card 114 routes record raw order/validity.
- Check · 4 min. Compare the exact valid appearances with prior prediction if it exists; mixed or too few actual tries gives a limited/unclear conclusion.
- Exit · 2 min. Save unchanged raw source and one cautious first statement.