The mock-analysis engine
This is where your score is actually built. The protocol below, run after every mock, is what compounds you to 99%ile.
The analysis protocol (3+ hrs, every mock)
1Re-attempt untimed. Before looking at any solution, redo every question you skipped or got wrong, with no clock. Many "wrong" become "I actually knew this", that's a strategy/time error, not a knowledge gap. This distinction is everything.
2Classify every question into one of four buckets:
| Bucket | Meaning | Action |
|---|---|---|
| Correct & fast | Ideal | Nothing, keep it up |
| Correct but slow | Method too slow | Learn a faster approach |
| Wrong | Concept / calc / misread? | Log it + fix the cause |
| Skipped | Right call or missed mark? | Refine your selection |
3Update the error log with every wrong/slow question and its lesson.
4Review your selection. Did you pick the right RC passages / DILR sets / QA questions? At this level, most lost marks come from bad selection, not missing knowledge. This review is gold.
5Write 3 fixes for next mock, e.g. "abandon DILR set 3 at 10 min," "skip the 4th RC," "double-check % calculations." Test them next time.
Your two living documents
Error log
For every mistake record: topic · question type · what went wrong · correct approach · one-line lesson. Review the whole log each weekend. Within a month you'll see your 3-4 recurring failure patterns, fixing those is the single highest-leverage move you can make.
Insight library (DILR + RC)
Separate from errors: a running list of "the key insight that cracked this set / passage." DILR and RC are pattern-recognition games. After ~20 mocks this library means you've seen the shape before, exactly how toppers move fast.
Mock targets & mindset
Volume: aim for ~25-30 full mocks by exam day. One series, analysed deeply, not five, sampled shallowly.
Convert every mock to percentiles via the Percentile vs Marks table and track the trend, not any single score.
.gs builders already in your toolkit, ask any time you want a fresh one for the week.