In my first year of teaching CLAT, I had a student who solved all 120 questions in every mock — and never scored above 75. He knew the content cold. His problem was the -0.25 rule: he was answering 120 questions at ~65% accuracy while students answering 95 questions at 78% accuracy were beating him by 25 marks. That single insight — accuracy, not attempts, wins CLAT — has guided our strategy ever since, including for our AIR 17 NLSIU and AIR 32 NLSIU selections.
💡 The rule in one line: +1 for correct, -0.25 for wrong, 0 for unattempted. Four wrong = one correct erased. Everything else in this article is about exploiting that arithmetic in your favour.
How the Marking Scheme Works
CLAT 2027 (exam: 6 December 2026) carries 120 questions and 120 minutes — exactly one minute per question on average. Every correct answer adds 1 mark; every wrong answer subtracts 0.25; skipping adds zero. There is no partial marking and no penalty for unanswered questions.
| Scenario | Marks |
|---|---|
| Correct answer | +1.00 |
| Wrong answer | -0.25 |
| Unattempted | 0.00 |
| 4 wrong answers | Equivalent to losing 1 correct answer |
Why the -0.25 Rule Is Brutal (The Math)
Two students with identical knowledge can differ by 30+ marks purely on guessing behaviour. Compare them:
| Student A (Blind Guesser) | Student B (Accuracy-First) | |
|---|---|---|
| Attempted | 120 questions | 96 questions |
| Correct | 78 | 75 |
| Wrong | 42 | 21 |
| Net score | 78 - (42 x 0.25) = 67.5 | 75 - (21 x 0.25) = 69.75 |
Student B attempted 24 fewer questions and still outscored the blind guesser — because every guess that misses is a small, repeated wound. In CLAT's dense merit list, 2 marks can separate a top NLU from a mid-ranked one.
ℹ️ Real perspective: the difference between AIR 17 NLSIU (Anish Gupta, CLAT 2024) and a rank outside the top 200 can be fewer than 10 raw marks. In our mock analysis at SLA, we have seen students recover 15–20 marks in eight weeks just by fixing negative-marking habits — no new syllabus, no new books, same knowledge.
When Should You Guess?
The honest answer: guess only when you have eliminated at least one option, and treat "educated" as a strict word. The arithmetic behind this:
- Blind guess (4 options): 25% chance of +1, 75% chance of -0.25. Long-run value: essentially zero — a coin flip that wastes exam time.
- One option eliminated (3 options): 33% chance of +1. Still marginal — attempt only if you are out of time anyway.
- Two options eliminated (2 options): 50% chance of +1 vs -0.25. Now the expected value is clearly positive — worth attempting.
- Inferred confidence: if you can explain why the remaining option is wrong, the guess is really a reasoned answer — attempt it.
The SLA Toppers' Rules for the -0.25 Rule
Rule 1: The 90-Second Cut
If a question is not nearly solved in 90 seconds, mark it for review and move on. In passage-based CLAT, the answer is either visible in the passage or it is not — staring does not manufacture information. This single rule prevents the panic-guessing that follows time pressure.
Rule 2: Attempt Sections, Not Questions
Do not pick questions randomly. Attempt a whole section pass first (e.g., all of Quantitative Techniques — the fastest section), then a second pass for the long passages. Unattempted-but-flagged questions return to you in the review round with fresh eyes.
Rule 3: Batch Your Guesses at the End
Leave deliberate educated guesses for the final 5 minutes. By then you know your remaining time and can decide honestly whether a 50/50 guess is worth it.
Rule 4: Track Accuracy After Every Mock
After every mock, compute section-wise accuracy: correct ÷ attempted. Below 70% in any section, your first fix is fewer attempts, not more speed. Above 80%, you may be skipping too much — speed up.
Section-by-Section Risk Guide
| Section | Questions (share) | Typical Risk | Recommended Accuracy |
|---|---|---|---|
| Legal Reasoning | 24% of 120 | Medium — passage traps are subtle | 75%+ |
| General Knowledge & Current Affairs | 24% of 120 | High — vague memory = wrong guess | 70%+ |
| Logical Reasoning | 22% of 120 | Low-Medium — rules are checkable | 78%+ |
| English Language | 20% of 120 | Low — passage answers are literal | 80%+ |
| Quantitative Techniques | 10% of 120 | Low-Medium — maths is precise | 75%+ |
Notice the pattern: the sections where knowledge is fuzzy (GK/Current Affairs, Legal Reasoning traps) are exactly where negatives bleed marks. English and Quant are precise — a right answer is right, a wrong one is a calculation error you can catch in review.
What the Numbers Mean — A Worked Table
Here is the scoring reality at different attempt levels, assuming a realistic 74% accuracy among attempted questions (120 questions total, each worth 1 mark):
| Attempted | Correct (74%) | Wrong (26%) | Net Score |
|---|---|---|---|
| 70 questions | 52 | 18 | 52 - 4.5 = 47.5 |
| 90 questions | 67 | 23 | 67 - 5.75 = 61.25 |
| 105 questions | 78 | 27 | 78 - 6.75 = 71.25 |
| 120 questions (all) | 89 | 31 | 89 - 7.75 = 81.25 |
Read the table the right way: the jump from 90 to 105 attempted is worth roughly 10 net marks — but the jump from 105 to 120 is worth roughly the same 10 marks at a risk of pulling accuracy below 74%. If forcing the final 15 questions drops your accuracy even to 68%, the net gain shrinks to ~4 marks — not worth abandoning the review round. The table also shows why unattempted questions are not "lost marks": they are saved marks spent on accuracy elsewhere.
ℹ️ Rank context: in CLAT's merit list, every ~2–3 net marks can shift your rank by hundreds of places among lakhs of applicants. A repeater or first-timer who adds 10 net marks through negative-marking discipline is effectively climbing the equivalent of several NLU tiers — with zero new syllabus. That is the cheapest improvement available in the entire paper.
How to Train the Habit (Mock Method)
- Mock 1–5 (diagnosis): attempt naturally; compute accuracy. You will likely find one section below 70% — that is your project.
- Mock 6–12 (restraint): cap your attempts at 100 questions; refuse blind guesses; apply the 90-second cut. Accuracy should climb 5–8 points.
- Mock 13–20 (calibration): raise attempts back toward 105–110 only where accuracy held 75%+. Never at its expense.
- Final phase: simulate exam-day conditions — same time of day, no phone, full 120 minutes, single sitting. At SLA our students take 200+ mocks in this structure, and it shows in their results.
ℹ️ One honest warning: accuracy targets shift every year with the paper's difficulty. A 2025-style paper (passage-heavy, since the format moved to passage-based questions with GK/current-affairs passages) rewards reading speed; a maths-heavy paper rewards precision. Recalibrate using your own mock percentile against the year's trend, not last year's score.
🎯 Bottom line: the -0.25 rule is not a trap, it is a filter — it filters out careless students. The students who respect it attempt fewer questions, score more marks and reach better NLUs. Respect the rule and it will respect your rank.
Get Your Mock Accuracy Analysed Free
Send us your last mock scorecard — we will tell you exactly where negatives are bleeding marks and how many questions you should actually attempt.