GATE does not have one negative-marking rule - it has three, live in the same three-hour paper, and the correct guessing decision is different for each. A wrong 1-mark MCQ costs 1/3 of a mark; a wrong 2-mark MCQ costs 2/3. MSQ (multiple-select) and NAT (numerical answer type) questions carry no negative marking at all, and MSQ additionally gives no partial credit - a partly-correct selection scores zero, the same as a blank. Work the expected value on a standard four-option MCQ and a genuinely blind guess comes out at exactly 0 - not a loss, not a gain, at both the 1-mark and 2-mark weight. Eliminate even one wrong option and it turns positive: +0.11 on a 1-mark question, +0.22 on a 2-mark one. On MSQ and NAT, since nothing is deducted for a wrong answer, the arithmetic is simpler and blunter: there is no such thing as a costly attempt, only a costly blank.
That is the whole shape of the strategy. What follows is where those numbers come from, a table for each question type, and where this reasoning - independent expected-value arithmetic applied to GATE's own published rules, not a peer-reviewed or GATE-endorsed method - stops being reliable.
The three marking regimes, stated exactly
GATE's official question paper pattern page for the current cycle states the scheme without ambiguity: GATE 2027's official pattern page, hosted by IIT Madras as this cycle's organising institute, gives "For a 1-mark MCQ, 1/3 mark will be deducted for a wrong answer" and "For a 2-mark MCQ, 2/3 mark will be deducted for a wrong answer," alongside "There is no negative marking for wrong answer(s) to MSQ or NAT questions" and "There is no partial marking in MSQ." Every number in this post is built directly on those four sentences - see our GATE time-management post for the same source's paper structure (65 questions, 100 marks, one 180-minute sitting, no mandated section time locks).
A GATE MCQ conventionally carries four options with exactly one correct - this specific detail is not restated on the current official pattern page, but is a longstanding, consistently reported convention of the exam's format (cited here via GATE's Wikipedia summary, which draws on past official information brochures). That four-option structure is what makes the marking scheme's numbers work out the way they do below - if a future cycle ever changed the option count, every figure in the next section would need recalculating.
Why a blind MCQ guess is exactly break-even, not a loss
The general formula for a multiple-choice question with n equally likely options, worth m marks correct and p marks deducted for wrong, is: expected value = (1/n) × m − ((n−1)/n) × p. Put GATE's numbers in. For a 1-mark MCQ with 4 options and a 1/3-mark penalty: EV = (1/4)(1) − (3/4)(1/3) = 0.25 − 0.25 = 0. For a 2-mark MCQ with 4 options and a 2/3-mark penalty: EV = (1/4)(2) − (3/4)(2/3) = 0.5 − 0.5 = 0. Both are exactly zero, and that is not a coincidence - the 1/3 and 2/3 fractions were set at precisely one-third of the mark value, and one-third is precisely 1/(n−1) for n=4. This is the same design principle behind any exam that sets its penalty at 1/(options−1): it makes a truly random guess neither help nor hurt.
The moment you can eliminate even one wrong option, the arithmetic tips positive - not by a small amount, but proportionally a lot, because the penalty stays fixed while your odds of being right improve:
| MCQ value | Options left after elimination | Chance of being right | Expected value per attempt |
|---|---|---|---|
| 1 mark | 4 (blind guess, none eliminated) | 25% | 0.00 |
| 1 mark | 3 (one eliminated) | 33% | +0.11 |
| 1 mark | 2 (two eliminated) | 50% | +0.33 |
| 2 marks | 4 (blind guess, none eliminated) | 25% | 0.00 |
| 2 marks | 3 (one eliminated) | 33% | +0.22 |
| 2 marks | 2 (two eliminated) | 50% | +0.67 |
Read the practical rule off that table directly: on a GATE MCQ, a genuinely random guess is not a gamble that costs you marks on average - it is a coin flip with zero edge either way. The moment you can rule out one option, however, guessing among the rest becomes a positive bet, and ruling out two makes it a strong one. The strategic question during the exam is never "should I ever guess on GATE" - it is "can I eliminate at least one option," and if the honest answer is no, a blind attempt costs nothing on average but also gains nothing, so time is the only real currency at stake, not marks.
Why MSQ and NAT invert the whole calculation
MSQ and NAT questions carry zero negative marking, which changes the decision from "is guessing worth the risk" to "is there ever a reason to leave one blank." There is not, mathematically - any attempt with a nonzero chance of being right has an expected value at or above zero, since nothing is subtracted for being wrong.
MSQ has a subtlety worth stating plainly because it is easy to get backwards: there is no partial credit. If a question has, say, two correct options out of four and you select only one of them, you score zero - identical to leaving it blank or selecting a wrong option. That means the "safe" instinct of ticking only the options you are most confident about is not actually safer than committing to your best full guess at the exact correct set. Since a near-miss scores the same as a total miss, the only losing move is not attempting when you have any partial signal - there is no reward for hedging within the answer itself.
NAT questions take open numeric entry rather than a choice among options, so "guessing" in the MCQ sense does not apply - there is no set of options to narrow down. What does apply is the same no-penalty logic: if partial working gets you to an approximate or rounded value, entering it costs nothing if wrong and gains everything if right, so an educated numeric estimate from incomplete working is always worth submitting rather than leaving blank.
| Question type | Negative marking | Blank vs. attempt | Rule |
|---|---|---|---|
| MCQ, no options eliminated | -1/3 (1-mark) / -2/3 (2-mark) | EV = 0 either way | Neutral. Attempt only if you can rule out at least one option - a real guess is rarely perfectly random, and an actively wrong intuition (falling for a designed distractor) can push EV negative. |
| MCQ, one option eliminated | same | +0.11 to +0.22 | Attempt. |
| MCQ, two options eliminated | same | +0.33 to +0.67 | Always attempt. |
| MSQ, uncertain of the full correct set | None; no partial credit | EV ≥ 0 for any real signal | Submit your best full guess at the exact correct combination. Do not hedge with a smaller "safe" subset - it scores the same zero as a blank. |
| NAT, partial working only | None | EV ≥ 0 for any real signal | Enter your best computed or estimated value. There is no downside to being wrong. |
How this differs from a CAT or NEET-style paper
Most negative-marking exams a GATE aspirant may have encountered run one uniform rule across every question - CAT deducts a flat 1/3 mark for a wrong MCQ answer across the whole paper (with no penalty on the non-MCQ Type-In-The-Answer questions, which is its own binary split rather than a third regime), and NEET deducts a flat 1 mark against 4 for correct, applied identically to all 180 questions. GATE's situation is less like either of those and more like three separate exams stitched into one sitting: the same paper asks you to be risk-tolerant on MSQ and NAT and risk-calibrated on MCQ, switching mid-question-type rather than staying constant for three hours. Treating the whole paper under one mental rule - either "never guess" or "always guess" - is wrong for at least one of the three question types no matter which rule you pick.
Turning this into a mock-review habit
The arithmetic above is only useful if you know, after a mock, whether you actually followed it - not whether you meant to. GATE's subjects in a study tracker like Karma Yogi are General Aptitude, Engineering Mathematics (on the 18 papers that carry it) and your Core Subject, and each carries its own score, attempted and correct counts when you log a full mock against it. That section-level breakdown will tell you your MCQ accuracy per subject, which is the number this whole framework depends on - if your real blind-guess accuracy on Core Subject MCQs is running below 25%, the "break-even" story above does not hold for you specifically, and guessing there is a net loss regardless of the theoretical math. The product does not currently separate MCQ from MSQ from NAT within a section's score, so the finer-grained read - "am I actually attempting every MSQ I have partial knowledge on" - is a habit to self-audit against your working, not a number the app will surface for you yet.
Where this is weak, and where it does not apply
- The four-option assumption is not restated on the current official pattern page. It is well-established by longstanding convention and cited secondary sources, but it is not the same tier of evidence as the marking-scheme sentences quoted directly from the official page. If GATE ever changes the option count, every number in the first table needs recalculating - none of it is fixed by the marking rule alone.
- Expected value is a statement about pure random guessing, not about your actual guesses. A "break-even" blind guess assumes your wrong answers are as likely to land on any of the three wrong options as any other. A well-written distractor is built to catch a specific, plausible wrong intuition - if your instinct is systematically drawn to it, your real guessing accuracy runs below the 25% the math assumes, and the true EV goes negative even though the theoretical number says zero.
- True zero-information guessing is rare in practice. Most GATE questions allow some elimination from units, order-of-magnitude plausibility, or partial recall, even when full confidence is absent. The framework matters most for the elimination-threshold decision - do I have enough to rule out one option - not for a literal coin-flip scenario, which is uncommon on a well-prepared attempt.
- This is arithmetic, not exam-day psychology. Knowing the expected value in advance does not stop time pressure from producing a rushed, badly-calibrated guess in the room. The only way this becomes instinct rather than a thing you have to recall mid-exam is rehearsing the elimination-threshold decision under timed mock conditions repeatedly, not reading this table once.
- None of this is GATE's own official guidance. It is independent expected-value reasoning applied to GATE's published marking rules, not a strategy GATE itself endorses or has tested.
Sources
- GATE 2027 official question paper pattern, IIT Madras (organising institute for the current cycle): gate2027.iitm.ac.in/question_paper_pattern - marking scheme for MCQ, MSQ and NAT quoted directly.
- Graduate Aptitude Test in Engineering, Wikipedia: en.wikipedia.org/wiki/Graduate_Aptitude_Test_in_Engineering - cited for the four-option MCQ convention, which the current official pattern page does not itself restate.
- All expected-value figures are original arithmetic worked from the two sources above, not quoted from either.
Logging accuracy per subject, mock over mock, is what turns "I think I guess well" into a real number you can check the arithmetic above against. Karma Yogi's GATE tracking lets General Aptitude, Engineering Mathematics and your Core Subject carry separate scores on every mock you log.
End of essay
- Anish Guruvelli