JEE Main Marking Scheme Explained: The Current Pattern and How It Got Here
A lot of JEE Main advice still in circulation describes a pattern that no longer exists: "numericals have no negative marking, so always fill something in" or "pick any 5 of the 10 numerical questions." Both were true at one point, and both stopped being true years ago. Here's the marking scheme as it stands now, how it got here, and what that means for how you practise.
The current pattern (Paper 1, B.E./B.Tech)
- 75 questions in 3 hours, 25 each in Physics, Chemistry and Mathematics, for a total of 300 marks
- Each subject has Section A with 20 multiple-choice questions and Section B with 5 numerical-value questions
- All 75 questions are compulsory. There is no choice within any section
- Marking is the same in both sections: +4 for a correct answer, -1 for an incorrect answer, 0 for a question left unanswered
How it got here: two changes worth knowing
The numerical-value section arrived in JEE Main 2020, with 5 numerical questions per subject and no negative marking on them. In 2021, during the pandemic, NTA expanded Section B to 10 numerical questions per subject, of which candidates answered any 5. The two changes that produced today's pattern came after that:
- 2022: negative marking extended to numerical questions. NTA's information bulletin for JEE Main 2022 initially said Section B had no negative marking, but NTA clarified in a notice in March 2022 that it would: every wrong numerical answer costs 1 mark, the same as a wrong MCQ. That has been the rule in every cycle since
- 2025: the optional choice removed. In a public notice dated 17 October 2024, NTA discontinued the "any 5 of 10" choice in Section B, which had been introduced as a COVID-era measure. From JEE Main 2025, Section B has exactly 5 compulsory numerical questions per subject
| Year | Section A: MCQs | Section B: numericals | Wrong numerical answer |
|---|---|---|---|
| 2020 | 20, all compulsory | 5, all compulsory | 0, no penalty |
| 2021 | 20, all compulsory | 10 given, answer any 5 | 0, no penalty |
| 2022 to 2024 | 20, all compulsory | 10 given, answer any 5 | -1 |
| 2025 onward | 20, all compulsory | 5, all compulsory | -1 |
Why negative marking hits numericals harder than MCQs
The penalty is the same -1 either way, but the odds behind a guess are not. On an MCQ, a blind guess has a 1-in-4 chance at +4, which works out to an expected +0.25 marks, slightly better than leaving it blank (the same maths behind guessing strategy on any +4/-1 exam). A numerical-value question has no options to choose from. You either compute the right value or you don't, so a rough estimate has a very small chance of being exactly right and a large chance of costing a mark. Under the current scheme, submitting a number you aren't confident in is almost always a losing bet.
What this should change in your prep
- Stop treating numerical practice as done once you recognise the method. The current scheme rewards finishing the calculation cleanly, and a fumbled last step now costs a mark instead of nothing
- In mocks, leave a numerical blank rather than submit a rough estimate, unless you're genuinely confident in the value
- With all 25 questions per subject compulsory, there's no skipping the two ugly numericals any more. Practise the full spread of a chapter's question types, not only the comfortable ones
- On MCQs, the usual rule still applies: guessing pays only when your chance of being right is above 20%, so eliminate options before you guess
How to spot outdated advice
If your notes, practice sets or a senior's tips say "numericals have no negative marking" or "choose 5 of 10", they describe a pattern from 2021 or earlier. Before building a strategy around any rule, check it against the current year's information bulletin on jeemain.nta.nic.in, since NTA publishes the full marking scheme there for every cycle. And if you're timing numerical practice to see how often you finish cleanly versus guess under pressure, Studyloaf's free timer tags sessions by subject so that pattern shows up in your own data.