UPSC Prelims Strategy: Score 110+ with Accuracy-First Preparation
Updated 14 August 2026
Over the last four exams, the General-category Prelims cut-off has swung from 75.41 (2023) to 92.66 (2025), with 88.22 in 2022 and 87.98 in 2024 — calculated on GS Paper-I alone, since CSAT is only qualifying. Preparing to "just clear" a target that moves by 17 marks is how serious aspirants miss by two questions. A 110+ target gives you insurance against any realistic cut-off, and it is reached by attempting fewer questions, not more. Always confirm the current year's figures on upsc.gov.in.
Why accuracy beats attempt count
GS Paper-I has 100 questions worth 2 marks each; every wrong answer costs one-third of that, i.e. 0.66. Watch what this does to three candidates:
| Attempts | Correct | Wrong | Positive | Penalty | Net score |
|---|---|---|---|---|---|
| 95 | 57 | 38 | 114 | −25.08 | 88.92 |
| 85 | 64 | 21 | 128 | −13.86 | 114.14 |
| 75 | 66 | 9 | 132 | −5.94 | 126.06 |
The 75-attempt candidate answers nine more questions correctly than the 95-attempt candidate — and scores 37 marks higher. Accuracy is not a personality trait; it is the output of deciding, question by question, whether you have earned the right to attempt. The sections below build that decision system.
Where the marks actually are
UPSC publishes no official subject break-up, but PYQ analyses of 2021–2025 papers show a stable pattern (approximate ranges; expect year-to-year swings):
| Subject | Typical questions (2021–2025) |
|---|---|
| Polity & Governance | 14–17 |
| Economy | 14–18 |
| Environment & Ecology | 13–19 |
| History, incl. Art & Culture | 12–16 |
| Geography | 9–18 |
| Science & Tech + Current Affairs | remainder, roughly 20–30 |
Polity, Economy, Environment and Geography together supply well over half the paper and reward systematic revision. Science & Tech, by contrast, is unpredictable — read it through current affairs rather than sinking weeks into a textbook.
The PYQ drilling method
Ten years of GS-I papers give you roughly 1,000 questions — the single best predictor of what UPSC asks, because themes repeat even when questions do not. Drill them in three passes:
- Pass 1 — subject-wise, untimed. After finishing a subject's revision, solve its PYQs. For every question, write why each wrong option is wrong. This trains the examiner's mind, not just recall.
- Pass 2 — mixed, timed. Solve full papers in 2-hour sittings. Log every error in an error notebook under three heads: knowledge gap, misread question, wrong elimination.
- Pass 3 — error log only. In the final month, re-solve only your logged errors. This is the highest-yield hour of the entire preparation.
Elimination arithmetic: when to guess, when to leave
The negative-marking maths gives you a hard rule:
- Blind guess (all four options open): expected value is 0.25 × 2 − 0.75 × 0.66 ≈ 0. You gain nothing on average and add pure risk. Leave it.
- One option eliminated: expected value ≈ 0.67 − 0.44 = +0.23 per question. Marginal — attempt only if your gut leans one way.
- Two options eliminated: expected value = 0.5 × 2 − 0.5 × 0.66 = +0.67 per question. Always attempt. Over ten such questions you net about 6.7 marks — more than the gap between the 2024 cut-off and safety.
Practise this triage in every mock until it is automatic: certain, two-eliminated, one-eliminated, untouched.
The 90-day endgame
| Phase | Days | Core work | Testing |
|---|---|---|---|
| Consolidate | 1–30 | Full static revision cycle 1; subject-wise PYQs (Pass 1); start error log | 1 sectional test + 1 timed CSAT paper weekly |
| Sharpen | 31–60 | Revision cycle 2 (faster); 12-month current-affairs compilation | 2 full-length GS mocks weekly; CSAT weekly |
| Simulate | 61–85 | Revision cycle 3 from short notes; error-log re-solving (Pass 3) | Mock on alternate days at the real exam's time slots |
| Taper | 86–90 | Light revision only; no new material, no new mocks | Rest, sleep cycle aligned to exam day |
By day 60, your mocks tell you your personal accuracy rate. Calibrate attempts to it: at 85% accuracy, 75–80 attempts is optimal; below 70%, adding attempts only feeds the penalty.
Protect your CSAT
CSAT needs 33% — 66.67 out of 200 — or your GS score is never evaluated. Each question carries 2.5 marks with a 0.83 penalty, so about 27 net-correct answers out of 80 clear it. That sounds trivial until you meet a hard paper: recent CSATs have been markedly tougher, and every cycle sees GS-strong candidates eliminated by Paper-II. The protection is boringly simple: one full timed CSAT paper every week from day 1 of your 90-day plan, comprehension passages first if maths is your weakness (or the reverse), and an honest diagnosis of which of the two is bleeding marks. Never let CSAT practice lapse for more than a week — it is the cheapest insurance in this exam.
