Running a GATE test series online that marks MSQ and NAT correctly

The first thing to check in any GATE test series software is whether it marks the way IIT Madras does. GATE has three question types, each with its own rule. A wrong MCQ costs a third of its marks, or two-thirds on a 2-mark question. An MSQ scores only for the exact set of correct options, with no penalty and no partial credit. A NAT answer counts if it lands inside the key's range, and a few keys accept two or even three separate ranges. If the engine gets any of this wrong, every rank you publish is wrong too.

What the GATE 2027 brochure lays down

IIT Madras is organising GATE 2027, and its information brochure, revised on 27 September 2026, sets out the pattern. Every paper is a three-hour computer-based test in English, with 65 questions for 100 marks. Ten General Aptitude questions carry 15 marks and the other 55 cover the subject. In 17 engineering papers, including CS, ME, CE, EE and EC, 13 of those subject marks come from engineering mathematics. Every question is worth 1 or 2 marks.

Two other rules affect how you build the series. Candidates may use only the on-screen virtual scientific calculator, and anyone found with a physical one in the hall is disqualified, whether or not they used it. And the syllabi for most papers have been revised for 2027. There's a new Robotics and Automation (RA) paper, Textile Engineering now sits inside XE as section XE9, and the section codes for XE, XH and XL have changed. If your question bank is tagged against the 2026 syllabus and codes, re-tag it before the first full mock goes out.

The exam runs on 6, 7, 13, 14, 20 and 21 February 2027, with a forenoon session from 9:30 am and an afternoon one from 2:30 pm. Results are due on 19 March, though the brochure says all dates are liable to change. The official GATE 2027 site carries the brochure, past papers and their answer keys.

The question mix changes by branch and by year

Count the question types in the official answer keys and the branches look very different. In 2026, CS-1, the first of two computer science sessions, had 24 MSQs out of 65 questions. Mechanical had none: 45 MCQs and 20 NAT questions. Civil's first session had 28 NAT questions and only 6 MSQs, and Electrical had 5 MSQs.

The split moves between years too. CS-1 went from 13 MSQs in 2025 to 24 in 2026, while Mechanical dropped from 26 NAT questions to 20. A single "GATE mock" template copied across branches trains students for a paper they won't sit. A CS aspirant has to be at ease with MSQs, where one missed option wipes out the question. A civil or mechanical aspirant spends far more of the three hours on long numerical working with the on-screen calculator.

So build a blueprint for each branch from its last three years of keys: how many of each type, at 1 mark and at 2 marks. Aim full mocks at the range across those years rather than last year's exact split. MSQs also take more care to write than they look. Each of the four options has to be clearly true or clearly false on its own, or the key becomes arguable, and GATE students will argue it.

GATE's MSQ rule is all or nothing

Many students, and some test engines, carry habits from exams that give partial credit on multiple-correct questions. GATE gives none. An engine can go wrong by giving part marks for some correct options, by deducting marks for a wrong combination, or by refusing a single selection.

That last failure costs more than you'd expect. Five of the 24 MSQs in CS-1 2026 were keyed to a single option, and Electrical, Electronics and Civil's first session each had at least one. A student trained to assume an MSQ always has two or more right answers loses those marks, and so does an engine that won't accept one tick. Put a few single-answer MSQs into every full mock, so students learn to judge each option on its own.

MCQ penalties need the same care with fractions. If an engine stores the penalty as 0.33 rather than as a third, three wrong 1-mark answers cost 0.99 instead of 1. In a batch of a few thousand, that's enough to split students who should be tied. The ones who total their own scores will be in your inbox the next morning.

NAT answers come with a range, and sometimes three

A NAT answer is a signed real number, clicked in on the virtual keypad. The question states how many decimal places to give, and students are told to round off only at the end. The keys then accept a range. In the 2026 Mechanical paper, one answer was accepted anywhere from 102.0 to 105.0 and another from 4.20 to 4.29. A third took either -0.61 to -0.57 or 0.57 to 0.61, and Civil's first session accepted either -5 or 5 for one question. In 2025, one Mechanical key gave three separate ranges for a single answer. Computer science keys are mostly exact: 12 of the 13 NAT answers in CS-1 2026 were a single value.

Each of those cases breaks a different kind of engine. One that compares text marks 4.2 wrong against a key of 4.20. One that stores a single minimum and maximum can't hold an answer with two or three ranges. A blanket tolerance, say 1 per cent either way on every question, is too loose for exact integer answers and may be too tight for long calculations. And an engine that lets students type on a phone keyboard never makes them practise entering a minus sign or a decimal point by clicking.

Much of the fix sits with faculty, not software. Set each range from your own worked solution, wide enough for legitimate rounding at the stated decimal places and no wider. Where the sign depends on a convention, either key both ranges or reword the question so the sign is fixed. After every mock, list the NAT questions where many students landed just outside the range; that usually means the range is too tight. The official keys show the other remedy. Electrical's 2026 key marks question 58 "MTA", marks to all, so your engine also needs a clean way to award a flawed question to everyone and re-rank.

A ten-minute test to run before you publish a mock

Whatever platform you use, build one short test with two MSQs (one keyed A and C, one keyed B alone), three 1-mark MCQs and the NAT questions below. Take it yourself as a student and check each result.

1. Choosing A and C on the first MSQ gives full marks; A alone gives zero, not half.
   
   
2. Choosing A, B and C gives zero, with no negative mark.
   
   
3. The MSQ keyed B alone accepts a single selection and gives it full marks.
   
   
4. Three wrong 1-mark MCQs cost exactly 1 mark, not 0.99.
   
   
5. A NAT keyed 4.20 to 4.29 accepts 4.2, 4.20 and 4.29, and rejects 4.30.
   
   
6. A NAT keyed -0.61 to -0.57 or 0.57 to 0.61 accepts both -0.59 and 0.59.
   
   
7. A negative decimal can be entered using only the on-screen keypad.
   
   
8. Setting one question to marks-to-all after the test updates every score and the rank list.
   
   

Rerun it after every platform update, and keep it for next year.

The calculator, and sections chosen in the hall

Students who do every mock with a handheld calculator lose time on exam day hunting for functions with a mouse. The brochure says the official site will carry mock exams with the virtual scientific calculator. If your mock player has no on-screen calculator, students can keep the official one open in another window during full mocks. Either way, the handheld should be in a drawer by December.

A few papers also let candidates choose sections during the exam. In XE, everyone answers General Aptitude and XE0 Engineering Mathematics, then up to two of the nine sections XE1 to XE9, picked after seeing the questions. XL works the same way, with chemistry (XL0) plus up to two of five sections. AR, GE and RA candidates pick part B1 or B2 in the hall. XE alone allows 36 pairs, and changing a chosen section in XE or XL clears its answers after a warning. The skill is to read a few sections fast and commit early.

An engine that can't offer that choice inside one test forces you to publish dozens of separate mocks, and students never practise the decision. GG and XH are simpler, because candidates fix their part or section on the application form.

Mocks for engineers who study after office hours

GATE has no age limit and a score stays valid for three years. So batches often include engineers already in jobs, aiming at an M.Tech or a PSU post. For them, a ranked mock at 10 am on a Tuesday is useless. Open each ranked full mock over a window, Friday 8 pm to Sunday midnight for example, and freeze the rank list when it closes. Release solutions only after the window shuts, because answers travel fast while it's open; our guide to preventing test series leaks covers the rest.

On weekdays, offer 45 to 60 minute subject tests that fit after 10 pm. Full mocks still belong on a laptop with a mouse, because the keypad and calculator are both clicked. Ask students to sit a few of them in one three-hour block that starts at 9:30 am or 2:30 pm on a weekend, like a real session. Three hours of focus in daylight is a different skill for someone who usually starts studying at 10 pm.

Engineering students stop trusting a GATE series the first time it marks an MSQ or NAT answer wrongly. You can test for that before you pick a platform. Upclass Mock Test runs online test series alongside your courses and live classes, in branded apps in your institute's own name on Android, iOS, Windows and Mac. Bring a 2026 paper and its official key to a demo and put it through the checks above, or see how Upclass fits GATE and ESE coaching.

Frequently asked questions

Is there negative marking for MSQ in GATE?

No. A wrong MSQ answer costs nothing under the GATE 2027 brochure, but there are no partial marks either; the student scores only by choosing exactly the right combination. NAT questions carry no negative marking. Only MCQs lose marks, a third of the question's value for each wrong answer.

Can a GATE MSQ have only one correct answer?

Yes. The brochure says one or more of the four options may be correct. The official 2026 keys have single-answer MSQs in several papers, five of them in CS-1 alone. Students should judge each option on its merits rather than assume there must be at least two.

Is a calculator allowed in GATE?

Only the on-screen virtual scientific calculator. Physical calculators are banned in the hall, and the brochure says candidates found with one will be disqualified whether or not they use it.

How are GATE NAT answers checked?

Against the range in the official answer key. The student enters a signed number on a virtual keypad, to the decimal places the question asks for, and any value inside the range earns full marks. Some keys give two ranges of opposite sign for one question, and one 2025 Mechanical key gave three.