PROPRIETARY DATA
Self-marking gaps — Cambridge & IB
First-hand patterns when students self-mark Cambridge past papers and IB Diploma practice against a strict second pass aligned to official mark schemes and markbands. Updated periodically; not official exam board statistics.
| Metric | Value | Notes |
|---|---|---|
| Method marks (M) under-awarded on first self-mark | 71percent_of_sessions | Students award A marks from final answer but skip M marks when working is incomplete. |
| Essay scripts placed one band high vs strict LoR read | 44percent_of_sessions | Economics/humanities; band evidence not quoted from script. |
| Net marks recovered after second scheme-aligned pass | 14percent_marks_on_average | Single-question retries; varies by subject. Largest gains on multi-step science and calculation papers. |
| MCQ keys marked correct despite misread handwriting digit | 8percent_of_sessions | Common on rushed Paper 1 scans. |
| AI-assisted first marks awarded above strict scheme read | 52percent_of_sessions | Generic LLM marking rewards plausible wording without method marks or band evidence. Second human-or-scheme pass corrects most. |
| Science structured Qs where precise term wording was missing | 37percent_of_sessions | Biology and chemistry Paper 2/4; students paraphrase instead of using mark-scheme vocabulary. |
| IB criterion responses placed one markband high on first self-read | 48percent_of_sessions | Humanities and sciences; descriptor evidence not quoted. Preliminary IB cohort — refresh quarterly. |
| Grade estimates using the wrong session or component set | 61percent_of_sessions | Results-season spikes: students compare raw marks to last year's aggregate or a different paper variant — component codes must match. |
Cite as: MarkScheme Cambridge marking insights — https://markscheme.app/insights