estigrade
Grading the estimate, not just the exam
Estigrade is an experimental grading rule designed to develop students' metacognitive skills by rewarding accurate self-assessment. Before an exam, a student estimates their expected score, and the final grade is adjusted by how close that prediction turned out to be.
The formula balances two terms: a reward for a close estimate and a penalty for a far one. Predict your performance well and you gain bonus points; miss badly and points are deducted.
The idea is to make self-knowledge a graded skill, so students build realistic self-assessment that carries beyond the classroom. Whether it actually improves learning, and whether honest estimation is even the score-maximizing strategy, are open questions the assumptions panel keeps separate from the formula itself.
Model changelog
- estigrade calculator: final = exam + reward*(100-|gap|) - penalty*|gap|, with live inputs for the estimate, exam score, and reward/penalty factors.
- split into Settings and Viewer components; the final grade is derived from a shared logic module.
- calibration reproduces hand-computed final grades for perfect, maximal-miss, and no-adjustment cases.
- assumptions flag that the final grade is unclamped and that incentive-compatibility is not guaranteed.