When to Skip: The Negative-Marking Maths Nobody Teaches

Coaching wisdom says never guess under negative marking. For NEET and SSC CGL the arithmetic says the opposite, and the real skill turns out to be knowing how good your guess actually is.

By The QuizAgent Team · July 26, 2026 · 6 min read

Every aspirant has heard it. Do not guess, negative marking will destroy your score. It gets repeated in coaching halls, in strategy videos, and by well-meaning seniors who scored well in spite of the advice rather than because of it.

For most Indian competitive exams, it is arithmetically wrong.

Run the numbers on NEET

NEET UG awards 4 marks for a correct answer and deducts 1 for a wrong one. Suppose you attempt a question where your genuine probability of being right is p. Your expected return is four marks times p, minus one mark times the chance you are wrong.

Work that through and the expected value comes to five times p, minus one. It turns positive the moment p rises above 20 percent. Cerebrum Biology Academy walks through the same arithmetic for NEET aspirants and reaches the same threshold.

Now consider what a completely blind guess looks like on a four-option question. No elimination, no intuition, nothing at all. Your probability is 25 percent. That sits above the 20 percent break-even, which means even a coin-toss-in-the-dark attempt carries a positive expected value of about a quarter of a mark.

Eliminate a single option and p becomes roughly 33 percent, worth about two-thirds of a mark per attempt. Eliminate two and p is 50 percent, worth one and a half marks. Across a paper with dozens of such questions, that is not a rounding error.

The same holds for SSC

SSC CGL Tier-1 uses a different scale: 2 marks for a correct answer, half a mark deducted for a wrong one. Different numbers, identical conclusion. The break-even sits at exactly 20 percent again.

That is not a coincidence. Both schemes use a reward-to-penalty ratio of four to one, and the break-even depends only on that ratio, not on the absolute marks. Any exam that pays four times what it charges breaks even at one attempt in five.

If you can eliminate even one option, attempting is not a gamble. It is the mathematically correct move.

So why does the advice survive?

Because the calculation quietly assumes something that is almost never true. It assumes you know your own p.

The arithmetic is trivial. Knowing whether you are in the 50 percent case or the 25 percent case is the entire difficulty, and it is a completely different skill from knowing biology or quantitative aptitude. It is a judgement about your own knowledge, made under time pressure, about a question you have read exactly once.

That skill has a name in the research literature. It is called calibration, and the evidence on how well students manage it is not flattering.

Most students are badly calibrated

The general finding, familiar from the Dunning-Kruger literature, is that the weakest performers are consistently the most overconfident. They are the least equipped to recognise the gaps in what they know, so their sense of having got this one is least trustworthy exactly where it matters most.

Worse, overconfidence is self-sealing. Work on metacognitive monitoring has noted that students who overrate their understanding tend to disengage from feedback and skip the self-correction step, because by their own reckoning there is nothing to correct.

This is why blanket rules spread. Never guess is a crude prosthetic for calibration. It protects a badly calibrated student from their own overconfidence, at the cost of leaving a large pile of positive-value marks on the table.

What to practise instead

The useful reframing is that a paper contains not two categories of question but three.

1. Questions you can answer. Attempt them, and watch the clock so they do not quietly become expensive.
2. Questions where you can eliminate something. The arithmetic above says attempt these. Most students skip them, and this is where the marks leak away.
3. Questions where you genuinely have nothing. Even here the numbers mildly favour attempting on a four-option paper, though this is the one case where a strict personal rule is defensible.

The category that decides your rank is the second one. It is also the one nobody drills, because drilling it requires practising under the real marking scheme rather than on an untimed, penalty-free question bank where every attempt is free.

Practising the decision, not just the syllabus

This is the gap QuizAgent's exam presets are aimed at. The six pre-configured exams carry their real paper format: the marks, the timing, the question mix, and the negative marking. A NEET mock deducts what NEET deducts. An SSC CGL mock uses plus two and minus half.

That matters because the skip decision only exists when a wrong answer costs something. Practise on a bank with no penalty and you never rehearse the judgement you will have to make on the day.

Every question also carries a written rationale, which is worth reading even when you got the answer right. Being right for the wrong reason is a calibration failure that a bare score will never reveal to you.

QuizAgent is in free public beta and the feature set is still growing. If something would make it more useful for the exam you are preparing for, we would genuinely like to hear it.