Negative marking and accuracy in IBPS PO

Every IBPS PO aspirant knows there is negative marking. Far fewer know what it actually does to their score. The rule is simple: a wrong answer costs one-fourth of the marks assigned to that question, and a question you leave costs nothing. The consequences are less simple, because the questions are no longer all worth the same, because most have five options, and because one wrong assumption in a puzzle can turn five answers wrong at once.

This guide works through the arithmetic once, properly, so that on exam day you follow a few rules without having to think about them.

What each question is worth now

In the 2026 cycle, the marks per question differ by section. If marks are spread evenly within a section, one question is worth:

The full pattern is in the IBPS PO exam pattern guide. The point here is that the penalty is always a fixed fraction of the reward. That gives one rule that holds in every section.

The maths of guessing with five options

Suppose you are down to k options and pick one at random. The chance of being right is 1/k. For a question worth m marks, the expected gain is:

m × [1/k − (k − 1)/k × ¼] = m × (5 − k) / 4k

Put in the values of k and you get this table:

Three things follow.

• A blind guess gains nothing on average. Over many questions, blind guesses leave your score where it was. They only add randomness.
• The break-even point is 20% confidence. If p is your real chance of being right, the expected gain is m × (5p − 1)/4. That is positive only when p is above 1 in 5. The break-even does not depend on the section, because the penalty is always a quarter of the reward.
• Elimination is what makes guessing pay. Ruling out one option barely helps. Ruling out two helps, and getting down to two options is well worth a guess.

Four worked examples

Example 1: English prelims. Two candidates attempt the 30-question English section.

• Candidate A: 24 attempted, 22 right, 2 wrong. Score = 22 − (2 × 0.25) = 21.5.
• Candidate B: 29 attempted, 22 right, 7 wrong. Score = 22 − (7 × 0.25) = 20.25.

B attempted five more questions, got none of them right, and lost 1.25 marks. Those five extra attempts were blind guesses in effect, or worse.

Example 2: Reasoning prelims, where attempts and accuracy trade off exactly.

• Candidate C: 28 attempted, 25 right, 3 wrong. Score = 25 × 8/7 − 3 × 2/7 = 200/7 − 6/7 = 194/7 ≈ 27.71.
• Candidate D: 33 attempted, 26 right, 7 wrong. Score = 208/7 − 14/7 = 194/7 ≈ 27.71.

D did five more questions and got one of them right. One right answer and four wrong ones net to exactly zero, the rule from the Key idea above. D spent the extra minutes for no gain at all.

Example 3: one puzzle gone wrong. A five-question puzzle in prelims reasoning is worth 5 × 8/7 ≈ 5.71 marks if you solve it. If you misplace one person and all five answers are wrong, you lose 5 × 2/7 ≈ 1.43. The gap between solving it and getting it wrong is about 7.1 marks. The gap between getting it wrong and skipping it is 1.43. That is why a puzzle you have not fully settled should not be answered on hope.

Example 4: mains data analysis. Candidate E answers 22 questions: 18 right, 4 wrong. Score = 18 × 1.5 − 4 × 0.375 = 27 − 1.5 = 25.5. Candidate F answers 28, but misreads the base year in one caselet and gets all five of its questions wrong along with four others: 19 right, 9 wrong. Score = 28.5 − 3.375 = 25.125. Six more attempts produced a lower score.

Accuracy versus attempts

Each attempt earns, on average, m × (5a − 1)/4 marks, where a is your accuracy. This table shows the net marks per attempt as a share of the question's value:

Suppose you have 20 minutes and can either do 20 questions at 90% accuracy or 25 at 68%.

• 20 questions at 90%: 18 right, 2 wrong. 18 − 0.5 = 17.5 (one-mark questions).
• 25 questions at 68%: 17 right, 8 wrong. 17 − 2 = 15.

Fewer attempts, done carefully, gave 2.5 more marks. The lesson is not "attempt less". It is: push attempts up only while accuracy holds. A sensible benchmark is 90% accuracy in mocks before you chase more attempts, and the table shows why: below about 70%, each extra attempt is worth little more than half a question.

Where negative marking bites hardest

Rules for the exam

1. Answer what you know first, in one pass through the section.
2. For a doubtful question, try to rule out at least two options before you guess.
3. Never answer questions from a puzzle or arrangement you have not fully solved.
4. Do not change a confident answer without a clear reason, such as a misread number.
5. Use "mark for review" and return only if time allows.
6. Do not fill in unread questions in the last minute.

For how these rules fit into timing each section, see the prelims strategy.

Try this

In your next three mocks, mark every guess with a small "G" on your rough sheet, and note how many options you had ruled out. After the result, count how many guesses were right.

• If fewer than one in three of your "two options ruled out" guesses were right, your elimination is weaker than you think.
• If your guesses as a group score close to zero, stop guessing below three options ruled out.
• If your accuracy is below 85% overall, cut attempts by two or three per section and see whether the score rises.

This turns negative marking from a vague fear into a number you know about yourself. Log it with the rest of your mock analysis.