CAT Negative Marking: "Don't Guess" Is the Wrong Rule
Every CAT page says the same thing: avoid negative marking. It sounds prudent. It is also the wrong objective, because it treats the penalty as the quantity to minimise, when the quantity that decides your score is expected marks — and under the real marking scheme those two goals pull apart far more often than candidates expect.
The entire topic reduces to one equation. Everything else on this page is a consequence of it.
where p is your probability of getting it right. Set EV to zero and you get p = 0.25 — which is exactly the chance of a blind guess among four options.
Read that last line again. It is the whole design of the scheme. A blind guess is worth exactly nothing — not a loss, nothing — because the penalty is calibrated precisely so that filling every box at random gains you no ground. It stops the paper being a lottery. It was never built to punish uncertainty.
Three terms this page leans on
- p — your real probability of being right
- Not your confidence. If you have eliminated one of four options and are choosing among the rest, p is one in three, whatever it feels like.
- EV — expected marks per attempt
- What the attempt is worth on average across many similar decisions. Positive EV means attempt; zero means it makes no difference; negative means leave it.
- TITA — type in the answer
- Non-MCQ questions where you type a value instead of selecting an option. These carry no negative marking, which changes the arithmetic completely.
What one eliminated option is worth
This is the number competitors never publish, and it is the practical content of the entire topic.
Eliminating a single option converts a break-even gamble into a positive one. Not an improvement at the margin — a change of sign. Eliminate two and the attempt is worth a full mark on average: a third of what a correct answer pays, on a question you cannot solve.
The rule, as a decision you make once
Decide this now. Not at minute thirty-four with your pulse up — now, in August, on a quiet afternoon, because the entire value of a rule is that it removes deliberation from the one moment where your deliberation is least trustworthy.
Notice where that puts the threshold. Not "am I more than half sure". The line sits at twenty-five per cent, and almost nobody sets it there — which is exactly why these marks are still lying around.
Why you will still under-attempt
Knowing the arithmetic changes less than it should. A one-mark loss feels heavier than a three-mark gain. That asymmetry is not a character flaw and it does not respond to being told the numbers — it produces one specific, predictable error, which is that candidates leave blank the questions where they genuinely had a read. The mark they might lose looms larger than the three they might win. So they collect neither.
It is a rule fixed in advance. One option eliminated, I answer. Decided in preparation, executed without deliberation, because the moment of decision in a closing section is the exact moment your weighing of gains against losses stops being reliable.
The three sections do not behave the same
The rule is constant. How easily you can satisfy it is not, and this is where general advice stops being any use.
Be stricter — eliminate two
Options are built to be close: true-but-unresponsive, or correct until the final six words. Elimination is genuinely hard here, so a VARC "guess" is more often truly blind than it feels. Raise your own bar.
Check whether it's a guess at all
You can often back-substitute options into the question and verify. Many quant guesses are not guesses — they are a slower solving method. Ask whether substitution settles it in forty seconds before treating it as uncertain.
Partial structure is elimination
Much of this section is TITA, so the penalty often does not apply. And where a set is part-built, the structure frequently rules options out even though it has not fully resolved. That counts.
CAT Mock Test — find your real accuracy on attempted questions
The rule only pays if you know where you actually sit. A full paper separates attempts from accuracy and shows both.
Four things candidates believe that cost them marks
One case the arithmetic does not settle
The last five minutes of any section
- Fill every empty typed-answer box. No exceptions, no thought required — a blank TITA is a guaranteed zero and a guess is not.
- Return to MCQs where you eliminated an option and then left them blank. Those are the marks you already earned and did not collect.
- Do not open questions you have never read.
- Do not revisit questions you answered confidently.
That first line is the most reliably ignored free expected value in the whole paper. It takes seconds. It cannot cost you anything. Candidates leave marks there every year.
Are you under-attempting?
Very high accuracy on a low attempt count is the signature of loss aversion, and it shows up clearly in a full paper.
- Attempts vs accuracy, per section
- Questions left blank
- Sectional percentiles
Questions candidates actually ask
What is the negative marking in CAT?
Multiple-choice questions carry three marks for a correct answer and deduct one mark for a wrong one. Non-MCQ questions, where you type the answer instead of selecting an option, carry the same three marks for a correct answer and no penalty at all for an incorrect one. That difference is not a detail — it means the two question formats need opposite strategies under time pressure.
Should I guess in CAT?
Yes, once you have eliminated at least one option. Expected marks per attempt work out to four times your chance of being right, minus one — so a blind guess among four options is worth exactly zero, while eliminating a single option makes the attempt worth about a third of a mark on average. Eliminating two makes it worth a full mark. On typed-answer questions you should always answer, because nothing is deducted.
Does a blind guess lose marks in CAT?
Not on average — it is worth exactly nothing. The penalty is calibrated so that guessing randomly among four options breaks even, which stops the paper being gameable by filling every box. It does add variance without adding expected marks, so it is not worth doing, but the common belief that random guessing systematically drains your score is simply wrong.
Is there negative marking for TITA questions in CAT?
No. Typed-answer questions deduct nothing for a wrong response, which makes attempting them profitable at any non-zero chance of being right and makes leaving one blank a guaranteed zero. There is no circumstance in which an empty typed-answer box is the correct decision. In the final two minutes of every section, sweep for empty ones and fill all of them.
How sure should I be before answering a CAT MCQ?
The break-even threshold is twenty-five per cent, not fifty. Any genuine reduction of the option set past a blind choice makes the attempt worth taking. Most candidates set their internal bar somewhere near fifty per cent, which discards an entire band of profitable attempts — and because that band contains the questions where you had a real read, it is a costly place to be conservative.
Why do I leave questions blank even when I have a good idea?
Because a one-mark loss feels heavier than a three-mark gain. That asymmetry is a well-documented feature of how people weigh losses against equivalent gains, and it produces a predictable error under exam pressure. Knowing the arithmetic does not fix it. What fixes it is deciding the rule in advance — one option eliminated, I answer — so that no weighing happens at the moment when your weighing is least reliable.
Does the guessing rule change by section?
The rule is constant; how easily you can satisfy it is not. VARC options are built to be close, so elimination is genuinely hard and a stricter bar of two eliminations is sensible. In quant you can often back-substitute options and verify, which means many apparent guesses are really a slower solving method. In DILR much of the section is typed-answer, and a part-built structure often rules options out even before it fully resolves.
Is high accuracy better than more attempts?
Only up to a point, and candidates routinely sit past it. Ninety-five per cent accuracy on twelve attempts returns 33.6 marks; seventy-eight per cent on twenty returns 42.4. What matters is accuracy on the questions you chose to attempt, combined with attempting everything that clears the twenty-five per cent bar. Very high accuracy on a low attempt count is usually a symptom of under-attempting rather than a sign of precision.
Like what you read? Share with a friend
Senior Content Writer
Co-Founder at PrepGrind, working on education content, product development, and student-focused learning resources. I also write articles and guides for MBA/MBS entrance exams, covering preparation strategies, exam updates, and useful resources for students.
