GATE EE Negative Marking: MCQ, MSQ and NAT, and What Guessing Costs

GATE EE is 65 questions and 100 marks in three hours, and only one kind of question can cost you marks: the multiple-choice question (MCQ). What most guides miss is how much the paper has shifted towards MCQs. In the official 2024 EE paper, 32 of the 85 subject marks were MCQ; in 2026 it was 55 of 85. More marks exposed to negative marking means guessing matters more — and in 2025, 5,263 candidates finished GATE EE with a negative total. This guide sets out the marking rules and what the numbers say about attempting.

GATE 2027 IB; official EE answer keys; GATE 2025 Statistical Report.

You may use only the on-screen virtual scientific calculator; a physical calculator, including one on a watch, is prohibited. Scribble pads are issued one at a time and returned at the end. You may take your seat from 40 minutes before the start and log in to read the instructions from 20 minutes before; you cannot enter more than 15 minutes after the start or leave before the end. Malpractice can lead to a ban from GATE for three years.

In 2024, NAT questions — which cannot lose you marks — were worth 41 of the 85 subject marks; by 2026 they were worth 22, while MCQs rose to 55. We cannot know whether 2027 continues the trend, but the recent papers are clear: the share of the paper where a careless guess costs marks has grown sharply. Three questions in the three papers were awarded “marks to all” in the final keys (2024 GA Q7 and EE Q18; 2026 EE Q58, a two-mark NAT on economic load dispatch), which is the organisers’ remedy for a flawed question.

A negative total means that, over the whole paper, the penalties for wrong MCQs exceeded everything earned. In 2025 almost one EE candidate in thirteen did that. The arithmetic is simple: on a four-option MCQ a pure guess has a 1-in-4 chance of +1 and a 3-in-4 chance of −1/3, an expected value of exactly zero — but with real variance, so a string of blind guesses usually costs marks. Eliminating even one option makes a guess worth attempting on average.

The expected value of a blind guess is exactly zero, but the spread is what hurts: ten blind 2-mark guesses most often produce two or three right — two right and eight wrong is a net loss of about 1.3 marks, three right and seven wrong a net gain of about 1.3 — so the outcome is a coin toss with real downside. As soon as you can eliminate one option, the average turns positive and attempting is worth it. The same logic explains the negative totals in the statistics: a candidate who attempts every MCQ without knowing the material is likely to lose marks, not gain them.

1. Enter every NAT and MSQ answer you have worked out. A wrong MSQ or NAT scores zero, the same as a blank, so a worked answer should never be left unentered — but a random entry wastes time you could spend elsewhere.
2. For MCQs, attempt when you can eliminate at least one option, and leave true blind guesses. The calculator above shows why.
3. Watch the rounding on NATs: compute with full precision on the virtual calculator and round only at the end, to the number of decimal places the question states.
4. For MSQs, choose only the options you are sure of — there are no partial marks, so a wrong or incomplete set scores zero; unlike an MCQ, it does not deduct marks.
5. Practise on the virtual calculator, not a physical one. Its layout slows many candidates down in the first mock.

Most question banks were built when NATs were the bigger share of the EE paper. If the 2026 balance holds, a practice set should now look like the 2026 paper: roughly two-thirds of subject marks as MCQs, a handful of MSQs and the rest NATs. Score every set with the penalty applied, and keep a separate count of how many MCQ marks you lost to wrong answers — if that number is regularly above three or four marks, the fix is attempt discipline, not more study. The 2024 paper, with its heavy NAT share, is still excellent practice for numerical accuracy; just do not use it as your guide to how much guessing costs.

The time budget is three hours for 65 questions, a little under three minutes each. GA is usually the fastest 15 marks on the paper, and Engineering Mathematics is fixed at 13 marks by rule, although our reading of the 2026 paper found about 18 marks of questions that were essentially mathematics (the divergence theorem, an ODE classification, a skew-symmetric quadratic form, orthogonal matrices, a contour integral, probability and a definite integral). Securing those before the long power-system and machines problems is a sound order of work.