Scoring yourself

Scoring yourself needs rubric thinking. A rubric row is a behavior, and behaviors are countable. For a self-grade, mark your answer:

• +1 for each obligation from the spec faithfully implemented
• +1 for each edge case (empty/absent/boundary) correctly handled
• +1 for the correct return shape and type
• −0 for style — nobody scores style
• −1 candidate: any obligation silently dropped (the most common loss)

Honesty anchors: most rubric rows for a 9-point question map to ~6 obligations plus ~3 edge/shape rows. If you cannot enumerate what your code does per row, you cannot self-score — enumerate first.

The simulate→score→repair loop is the final training: simulate one set, score one set, repair the one biggest loss, repeat. Repair the BIGGEST loss only — time spent polishing your best behavior while a blank behavior remains is exactly the exam's time trap in miniature.