GATE ME Negative Marking: MSQs Vanished, MCQs Grew — What to Attempt

GATE’s marking rules have not changed in recent brochures: multiple-choice questions lose a third of their marks for a wrong answer, while multiple-select and numerical-answer questions carry no penalty. What has changed is how many of each type the ME paper contains. Counting the official final keys, the subject marks carrying a penalty rose from 32 of 85 in 2023 to 47 in 2026, and multiple-select questions disappeared. This guide sets out the rules, the real question mix year by year, what it did to scores, and a calculator for the marks you put at risk.

GATE 2027 brochure; our count of the official ME keys; GATE 2026 statistical report.

Adding General Aptitude’s 15 MCQ marks, 62 of the 100 marks in the 2026 paper carried a penalty, against 47 in 2023. Numerical questions — where guessing costs nothing but also almost never works — fell from 43 subject marks to 38. The paper has moved toward questions where a careless answer can cost marks, and there is no sign yet whether 2027 will continue the trend.

Two practical consequences follow. A candidate who used to bank on MSQs — where a partly-known question could be attempted without risk — now faces that share of the paper as penalty-carrying MCQs. And the numerical questions that remain are where a well-prepared candidate scores without any penalty risk, so NAT practice is still the safest source of marks.

In 2026, more than 3,200 candidates finished with a negative total — 5.4% of those who sat, about one in eighteen and a half. That is the visible end of a wider effect: every wrong MCQ drags down scores all the way up the distribution. With 62 penalty-carrying marks on the paper, a candidate who attempts every MCQ and gets a third of them wrong loses about 7 marks to penalties alone (our arithmetic: a third of 62 marks wrong, times one-third). For a candidate scoring around 50 in 2026, 7 marks is the difference between about rank 900 and about rank 1,800 — our interpolation of the GATE 2026 report’s mark table, where 888 scored 50 or more and 2,521 scored 40 or more.

General Aptitude deserves its own note. Its 15 marks were all MCQs in every year from 2023 to 2026 — five 1-mark and five 2-mark questions — so a quarter of the penalty exposure on the paper sits outside the engineering subject altogether. GA questions are usually shorter than subject questions, which makes them a good place to be both fast and careful.

With four options and a penalty of one-third, a blind guess on an MCQ has an expected value of exactly zero: a 1-in-4 chance of the full mark against a 3-in-4 chance of losing a third. Eliminating even one option tips it positive; eliminating two makes a guess clearly worth taking. The calculator below does the arithmetic for 1- and 2-mark questions. On NAT and MSQ questions nothing is lost by answering, so every one should be attempted if time allows — but a guessed numerical answer is very unlikely to be exactly right.

When a question is found to be flawed, GATE awards its marks to every candidate. For ME this happened to one question in 2023 (Q55, a NAT), to five in 2024 (General Aptitude Q7 and Q10; ME Q19, Q34 and Q63), and to none in 2025 or 2026. It is a reminder that a question that seems impossible sometimes is — leave it, move on, and let the key sort it out.

Numerical-answer questions carried 38 subject marks in 2026 and between 43 and 48 in the three years before. They cannot be guessed, carry no penalty, and are marked right only if the answer falls in the key’s range — so the risk is arithmetic, not guessing. Only the on-screen virtual calculator is allowed, which is slower than a physical one; practising NATs with the GATE virtual calculator, and carrying enough decimal places through the working, recovers marks that are otherwise lost to rounding.