For three consecutive cycles — 2023, 2024 and 2025 — NEET PG was a 200-question paper worth 800 marks. Every score benchmark that circulated in coaching circles, every “you need 550+ for a safe branch” rule of thumb, every mock test series on the market was built on that arithmetic.
In 2026, NBEMS changed it. The paper is now 180 questions for a maximum of 720 marks, split into five sections of 36 questions each instead of 40. The exam ran on 30 August 2026 in a single shift from 9:00 AM to 12:30 PM, with a re-examination on 5 September for candidates affected by a power failure at a Jaipur centre. Over 2.65 lakh aspirants sat the paper, and with the provisional answer key and response sheets now due at natboard.edu.in, almost all of them are about to run the same calculation: what does my raw score actually mean this year?
The honest answer is that it means something different from what it meant last year — and not in the direction most people assume.
The change that nobody expected: time went up, not down
The instinctive reading of “20 fewer questions” is that the exam got easier or shorter. It didn’t. The total duration stayed at 3 hours 30 minutes — 210 minutes — and the sectional timing stayed at 42 minutes per block.
Do the division. Under the old format, 200 questions in 210 minutes gave you about 63 seconds per question. Under the new one, 180 questions in the same 210 minutes gives you roughly 70 seconds. That is about seven extra seconds per question, which across a full paper adds up to something close to half an hour of additional effective thinking time.
This was deliberate. Rather than lengthening an already exhausting exam, the committee reduced the question count so candidates would get more breathing room per question without sitting in the hall any longer.
The consequence for your rank is the part worth sitting with. When every candidate gets more time per question, the paper stops rewarding raw speed as heavily and starts rewarding accuracy. Under a tight clock, a certain number of strong candidates lose marks purely to rushing — misreading a stem, picking the second-best option, running out of time in the final block. Loosen the clock and a chunk of those errors disappear across the entire cohort.
That compresses the top. If the people around you are making fewer careless mistakes, the distance between rank 500 and rank 2,000 narrows in absolute marks. Two or three questions that would have separated you comfortably in 2024 may separate you by a hair in 2026.
Why you cannot compare your score to last year’s cutoffs
This is the single most common mistake being made right now, in WhatsApp groups and on every forum where scores are being shared.
A score of 560 out of 800 is 70%. A score of 560 out of 720 is 77.8%. These are not the same performance, and the second one is considerably stronger. Yet the cutoff lists everyone has bookmarked — the 2025 closing scores for Radiology, for Dermatology, for General Medicine at a particular institute — are all denominated in the old 800.
If you want a rough translation, convert to percentage rather than comparing absolutes. Multiply your 720-scale score by 1.111 to get its approximate 800-scale equivalent, then check that against historical data. A 500 this year is roughly equivalent in percentage terms to a 555 last year.
But treat even that as a very loose guide, for two reasons. First, cutoffs are not fixed properties of a branch — they move with the difficulty of the paper, the number of candidates, and seat availability in a given year. Second, the compression effect described above means percentile relationships may not map cleanly onto percentage relationships. Your percentile is what determines your rank, and your percentile depends entirely on how everyone else did, not on how you would have done against a previous cohort.
The only number that will genuinely tell you where you stand is your actual rank, and that arrives with the result.
The guessing arithmetic is unchanged, and it still decides ranks
The marking scheme did not move: +4 for a correct answer, −1 for an incorrect one, 0 for an unattempted question.
The gap between a wrong answer and a blank is therefore five marks — the four you failed to earn plus the one deducted. On a 720-mark paper, five marks is a meaningful fraction of the distance between two ranks in the crowded middle of the distribution.
The expected-value maths is straightforward. On a four-option question where you have no information at all, you have a 25% chance of +4 and a 75% chance of −1, which comes to an expected value of +0.25 marks. Technically positive, but so marginal that it is effectively noise. Eliminate one option and your odds improve to one in three, giving an expected +1.33. Eliminate two and you are at a coin flip, expecting +1.5 marks per attempt. That is where guessing stops being a gamble and becomes correct strategy.
The practical rule that falls out of this: blind guessing across twenty questions will gain you roughly nothing and may cost you. Educated guessing on twenty questions where you have genuinely ruled out two options will gain you around thirty marks. Those thirty marks, on a 720-mark paper in a compressed field, are ranks.
If you are reconciling your response sheet against the answer key now, count your wrong attempts separately from your blanks. The ratio tells you something useful about your exam temperament, even if it is too late to change this year’s outcome.
Sectional lock-in: the rule that punishes poor pacing
The five sections — Groups A through E — are time-bound. Each block gets 42 minutes. When the timer expires, the interface locks that section permanently and moves you to the next one. You cannot go back to review, change an answer, or return to a question you flagged.
With 36 questions in 42 minutes, that is 70 seconds per question within each block, with no ability to bank surplus time from an easy section and spend it on a hard one.
This structure interacts with the reduced question count in a way that matters for rank. Because there is more time per question than in previous years, candidates who manage their blocks well now have a genuine opportunity to finish a section and use the remainder to re-check. Candidates who panic and burn fifteen minutes on four difficult questions in Group B lose that section irrecoverably. The pattern change rewards discipline more than it ever has.
There is no normalisation, because there is only one shift. Your raw score is your raw score.
What actually determines your rank from here
Three things, in this order.
Your final score after the objection process. The provisional answer key is exactly that — provisional. NBEMS is releasing it for the first time following a Supreme Court directive, and there is a window during which candidates can challenge specific questions for a per-question fee. Successful challenges can result in questions being dropped or keys being corrected, which shifts scores across the board. Your score today is an estimate, not a result. There is also a single combined answer key covering both the 30 August exam and the 5 September re-test.
The shape of the cohort’s distribution. With 2.65 lakh candidates and a paper widely described as moderate in difficulty, expect a dense cluster in the middle. In a dense cluster, four marks — one question — can move you several hundred ranks. This is not a reason to panic; it is a reason to stop calculating your rank to three decimal places from an unofficial key.
Tie-breaking. The information bulletin specifies the order in which identical scores are resolved. With more candidates packed into fewer available marks this year, ties are statistically more likely than they were on an 800-mark paper. If you are near a branch boundary, read the tie-break clause in the bulletin rather than relying on a summary of it.
What to do this week
Download your response sheet the moment it is available and check it against the provisional key carefully, question ID by question ID. Calculate your score honestly using +4 and −1, without rounding in your own favour.
Then use that number for one purpose only: deciding which counselling rounds to register for and which colleges to research. Do not use it to predict your exact rank, do not compare it against last year’s 800-mark cutoffs without converting, and do not let a rank predictor built on the old pattern set your expectations.
The result is due by 30 September, and going by previous years it may well arrive earlier. Until then, the most productive thing you can do is stop refreshing the calculation and start building a realistic college preference list — because the candidates who get the best outcome in counselling are rarely the ones who scored highest. They are the ones who understood the process before the clock started.
