On this page
After JEE Main, two friends compare notes. One scored 120 marks and got a lower percentile than the other, who scored 110 in a different shift. It looks unfair until you see how NTA builds the score. This post explains the method from the official bulletin, works through the numbers, and shows what your percentile does and does not tell you.
Why does NTA normalise scores at all?
The JEE (Main) 2026 Information Bulletin explains it in Appendix VII. The exam runs in many shifts, each with its own question paper, and despite efforts to keep them equal, some papers may be harder than others. A candidate who gets a tougher paper is likely to score lower marks. To make sure candidates are "neither benefitted nor disadvantaged" by the shift they get, NTA uses a percentile-based normalisation.
The scale is not small. NTA's press release of 20 April 2026 says Session 1 of Paper 1 ran in ten shifts and Session 2 in nine, and that 15,38,468 unique candidates appeared across the two sessions.
How is the percentile calculated?
The bulletin's formula, applied separately to each shift, is:
Percentile = 100 × (number of candidates in your shift with a raw score equal to or less than yours) ÷ (total number of candidates who appeared in your shift)
Three rules follow from the bulletin:
The topper of every shift gets exactly 100, whatever their raw marks.
Percentiles are calculated to seven decimal places, to reduce ties.
The percentile for your total is calculated from your total raw score. It is not the average of your subject percentiles.
The bulletin also states plainly that the percentile is not the same as the percentage of marks.
Worked example: two shifts, two raw scores
These are sample numbers, chosen to show the method. The shift sizes are real: NTA's press release gives 1,12,909 candidates appearing on 2 April 2026, Shift 1, and 1,15,612 on 6 April 2026, Shift 2.
Riya (Shift A, 1,12,909 appeared) scores 120 raw marks. Suppose 1,00,000 candidates in her shift scored 120 or less. Her percentile is 100 × 1,00,000 ÷ 1,12,909 = 88.5668990.
Arjun (Shift B, 1,15,612 appeared) had a harder paper and scores 110. Suppose 1,04,000 candidates in his shift scored 110 or less. His percentile is 100 × 1,04,000 ÷ 1,15,612 = 89.9560599.
Arjun has fewer raw marks but a higher percentile, because he did better relative to the people who sat the same paper. That is the whole point of the method.
What happens with two sessions?
The bulletin says a candidate need not appear in both sessions, but if you do, your best NTA score of the two is used for the merit list. NTA's 20 April 2026 press release confirms that for candidates who appeared in both sessions, "the best NTA score of both the Sessions has been declared". Overall merit is then prepared by merging NTA scores from all shifts of both sessions. For planning around the two attempts, see our post on preparing for JEE Main's two sessions.
How are ties and cut-offs handled?
When two candidates have the same total NTA score, the 2026 bulletin breaks the tie in order: NTA score in Mathematics, then Physics, then Chemistry, then the candidate with the lower proportion of incorrect to correct answers overall, then the same proportion in Mathematics, Physics and Chemistry in turn. If they are still tied, they share the rank.
For eligibility cut-offs, the bulletin says that if the percentile matching the cut-off marks differs between shifts, the lowest of those percentiles becomes the cut-off for everyone. For JEE (Advanced) 2026, the press release listed the cut-off for the UR category as 93.4123549. Cut-offs change every year, so check the current notice.
Common misreadings and how to avoid them
"I got 95 percentile, so I got 95% of the marks." No. It means about 95% of your shift scored the same as you or less.
"My shift was hard, so I am disadvantaged." Not by the method: a hard paper lowers everyone's raw marks in that shift, and the percentile compares you with them.
"I can work out my exact rank from my percentile." You can only estimate. If about 1% of candidates scored above you overall, then out of 15,38,468 that is roughly 15,385 people. Treat that as a rough guide; NTA publishes the actual ranks.
"Negative marks don't matter for percentile." They do, because they change your raw score. Our post on when guessing is worth it under negative marking works through the expected value.
What should you do with this?
After each mock test, track raw marks and your position among the test-takers, not raw marks alone.
After the real exam, use the official answer key process if you spot an error. Our guide to challenging the NTA provisional answer key explains the steps.
Read the bulletin for your exam year when NTA publishes it, since the procedure, dates and tie rules can change.
Raising your percentile still comes down to raising your raw score in the chapters that cost you most. If your mock analysis shows the same kind of mistake again and again, Fermi's AI tutor can work out what you already know and build a plan for the session, so each study block starts from your weakest topic.
Open the 2026 bulletin, find Appendix VII, and redo the worked example with your own latest mock score.
