BharatNotes Start sectional tests at month 4 of serious prep, full-length mocks 4 months before exam. Aim for 20-25 full-length mocks + 15-20 sectional tests, with PYQs solved twice. Each 2-hour mock needs 2-3 hours of analysis. Quality of analysis matters 10x more than count.
When to start
Beginners often plunge into mocks too early and break their confidence, or too late and miss strategic calibration. The right timeline, anchored to CSE 2026 Prelims on 24 May 2026:
| Months before exam | Calendar window for CSE 2026 | Test activity |
|---|---|---|
| 12-9 | May 2025 → Aug 2025 | No mocks. Build foundations. Solve PYQs subject-wise only after finishing that subject's first read. |
| 9-6 | Aug 2025 → Nov 2025 | Subject-wise sectional tests (Polity test after Laxmikanth, etc.). One per week. |
| 6-4 | Nov 2025 → Jan 2026 | Begin half-length mocks (50 Qs). Once a week. |
| 4-1 | Jan 2026 → Apr 2026 | Full-length mocks (100 Qs, 2 hours, OMR-style). 1-2 per week. |
| Last month | 24 Apr → 24 May 2026 | 1 mock every 3-4 days, with massive analysis time. |
How many is enough
- 15-25 full-length mocks — sweet spot for serious aspirants.
- 15-20 sectional/half-length tests — for plugging holes.
- Last 10 years of PYQs — solved twice, the second time as an elimination-only drill.
More than 30 full-lengths usually signals avoidance — you are testing instead of revising. Stop and go back to your notes.
Topper voice — Shubham Kumar (AIR 1, CSE 2020)
"In my 2019 attempt I took 70-75 mock tests; in 2020 I took 40-45. One should attempt both sectional and full-length mocks — sectionals from one institute, full-lengths from three or four institutes. In the last month, at least 20 tests, but avoid attempting any mock in the week immediately before Prelims."
Notice the year-over-year reduction — 70 to 45. Shubham did not need more tests in the year he became AIR 1; he needed deeper analysis on fewer. That is the trajectory every aspirant should aim for.
The mock that does nothing
If you finish a mock, check the score, feel happy or sad, and move on — you wasted 2 hours. Mocks teach almost nothing without structured post-mortem.
The 5-bucket analysis template
After every mock, categorise every wrong/skipped question into one of five buckets:
- Conceptual gap — I genuinely didn't know this. → Add to revision notes.
- Silly mistake — I knew it but mis-read or mis-bubbled. → Time-management / fatigue issue.
- Elimination failure — I had 50/50 and picked wrong. → Drill elimination logic on similar Qs.
- Over-confidence — I was sure, but my static fact was outdated. → Re-read that source.
- Should have skipped — I attempted blindly and lost marks. → Tighten attempt discipline.
Six metrics to track per mock
| Metric | Why it matters |
|---|---|
| Net score | Surface number. |
| Attempts | Risk appetite signal. |
| Accuracy = correct ÷ attempts | The master variable. |
| Skipped | Tracks discipline. |
| Subject-wise accuracy | Reveals weak silos. |
| Time-per-question (if practising on app) | Catches dawdling and rushing. |
Maintain a Google Sheet across mocks 1 to 25. By mock 15, the patterns are obvious — "I keep dropping marks in Environment" or "My accuracy crashes after question 70". Fix patterns, not individual questions.
Worked scenario — calibrating to a real cutoff
You take 20 mocks. Your sheet shows mean score 96, median 92, range 78-112, mean accuracy 71%. Mapping against the 2013-2025 cutoff swings (low 75.41 in 2023, high 116 in 2016), your 78 floor would have failed CSE 2016 badly, scraped CSE 2024 (87.98) and CSE 2025 (92.66), and cleared CSE 2023 easily. The conclusion isn't "you're safe" — it's "you need to lift the floor". Spend the last 8 weeks driving the floor from 78 to 90 by attacking the two subjects with the worst sectional accuracy.
Topper voice — Anudeep Durishetty (AIR 1, CSE 2017)
"Give every test the seriousness of the final exam — strict time limits, OMR-mode, no breaks. Take 10-15 extra minutes to finish, you are cheating only yourself."
Choosing a test series
Do not enrol in three series. One is enough — Vision IAS, Insights, Vajiram, Forum, or any reputed name. Solve every test on time, in OMR mode, without breaks. The all-India rank is irrelevant; your delta over weeks is what counts.
Recent policy hook for mock-test takers
Because UPSC will release the provisional answer key within days of the actual CSE 2026 Prelims (24 May 2026) and entertain candidate objections, mock-test analysis culture should mirror this. After every mock, treat the answer key as provisional — challenge two questions per mock with sourced reasoning. This builds the same muscle UPSC now formally rewards: precise, source-backed disagreement with a published key.
Mentor takeaway
The aspirant who took 12 mocks and analysed each one for 3 hours beats the aspirant who took 40 mocks and analysed none. Mocks are diagnostic instruments, not study material.
