You have 40 seconds left in the reasoning section. Three questions are unattempted. You can rule out one option in each. Do you tap an answer or leave them?
Half your group will say "never guess, negative marking will kill you." The other half will say "always guess, it is 25 percent chance." Both are working from a feeling. The real answer depends on three numbers: how many marks a right answer gives, how many a wrong answer takes away, and how many options are still alive in your head. Once you know those, the decision is arithmetic.
This post gives you the actual penalties in the main exams, the break-even math, a simple rule you can carry into the hall, and the habits that matter more than any single guess.
The penalties, exam by exam
These are the schemes as they stood in recent notifications. Patterns change, so confirm on the official notice each cycle.
| Exam / stage | Marks per correct answer | Penalty per wrong answer | Options |
|---|---|---|---|
| SSC CGL Tier 1 | 2 | 0.50 | 4 |
| SSC CGL Tier 2, Paper I | see notification | 1 mark (reported) | 4 |
| SSC CGL Tier 2, Paper II and III | see notification | 0.50 (reported) | 4 |
| RRB NTPC CBT 1 | 1 | 1/3 | 4 |
| RRB Group D CBT | 1 | 1/3 | 4 |
| IBPS PO / Clerk prelims | 1 | 0.25 | 5 |
| IBPS PO mains (objective) | as per section | 0.25 of the marks | 5 |
| SBI PO prelims and mains (objective) | as per question | 0.25 of the marks | 5 |
No exam here penalises a question you leave blank. That is the whole point. Unattempted costs zero.
Two things to notice. First, the penalty is not the same fraction in every exam. SSC CGL Tier 1 takes back a quarter of the marks a question is worth (0.5 out of 2). RRB takes back one third. IBPS takes back a quarter. Second, the number of options differs. Bank exams use five options, and SSC and RRB use four. That changes what a random guess is worth.
The one formula worth remembering
Suppose a question is worth M marks and a wrong answer costs P marks. If your chance of being right is p, then the expected gain from answering is:
Expected gain = p × M − (1 − p) × P
Set that to zero and you get the break-even chance:
p = P / (M + P)
If your real chance of being right is higher than that, answering pays on average. If lower, you lose on average. Now plug in the exams.
SSC CGL Tier 1. M = 2, P = 0.5. Break-even p = 0.5 / 2.5 = 0.20, or 20 percent. With four options a pure blind guess has a 25 percent chance. So a blind guess in SSC CGL Tier 1 is slightly positive: expected gain = 0.25 × 2 − 0.75 × 0.5 = 0.5 − 0.375 = +0.125 marks per guess.
RRB NTPC and Group D. M = 1, P = 1/3. Break-even p = (1/3) / (4/3) = 0.25, or 25 percent. A four-option blind guess is exactly 25 percent. So expected gain is zero: 0.25 − 0.75/3 = 0.25 − 0.25 = 0. Blind guessing is neutral.
IBPS PO and Clerk prelims, SBI. M = 1, P = 0.25. Break-even p = 0.25 / 1.25 = 0.20. With five options a blind guess is 20 percent. Again zero: 0.2 − 0.8 × 0.25 = 0.
So the exams are designed so that blind guessing is roughly neutral, apart from a small positive edge in SSC CGL Tier 1. The value of guessing comes entirely from what you know before you guess, meaning how many options you have eliminated. That is where the decision lives.
What eliminating options is worth
Here is the same calculation with options removed. Assume your eliminations are correct, meaning the right answer is still among the remaining ones. This is the assumption that fails most often in practice, so I will come back to it.
SSC CGL Tier 1 (4 options, +2, −0.5)
| Options left | Chance | Expected gain |
|---|---|---|
| 4 | 25% | +0.125 |
| 3 | 33% | +0.33 |
| 2 | 50% | +0.75 |
RRB NTPC / Group D (4 options, +1, −1/3)
| Options left | Chance | Expected gain |
|---|---|---|
| 4 | 25% | 0 |
| 3 | 33% | +0.11 |
| 2 | 50% | +0.33 |
IBPS / SBI (5 options, +1, −0.25)
| Options left | Chance | Expected gain |
|---|---|---|
| 5 | 20% | 0 |
| 4 | 25% | +0.06 |
| 3 | 33% | +0.17 |
| 2 | 50% | +0.38 |
You can see the story. In all three schemes, getting from four or five options down to two turns a guess from "no gain" into a real gain. Getting to three is a mild gain. Getting from five to four in a bank exam is barely worth the ink.
The failure hidden inside those tables
The tables assume your elimination is reliable. If you rule out two options and the right answer is actually one of the ones you ruled out, then you are not guessing at 50 percent, you are guessing at 0 percent.
How reliable are your eliminations? That depends on the question type, and this is the part most articles skip.
Reliable eliminations. Numerical answers where you can bound the result. If the answer must be between 40 and 50 and two options are 36 and 72, they are gone. Approximation questions where the units digit or the rough size tells you. Number series where a couple of options break the visible pattern. Sentence completion where two options are grammatically impossible. These eliminations are mechanical.
Unreliable eliminations. Options that "feel wrong" in general knowledge. Statement-conclusion questions where you dropped an option because it sounded extreme. Puzzle questions where you have half a solution. Any option you eliminated because you "don't remember hearing about it." This last one is a big trap in static GK.
My rule for myself was to count only eliminations I could justify in one sentence. "This option is larger than the total, so no" counts. "This one doesn't ring a bell" does not.
If your eliminations are 70 percent reliable, then a two-option guess is worth about 0.7 × 0.5 = 35 percent, not 50 percent. Redo the table for that. In RRB with +1 and −1/3, a 35 percent chance gives 0.35 − 0.65/3 = 0.35 − 0.217 = +0.13. Still slightly positive. In SSC CGL, 0.35 × 2 − 0.65 × 0.5 = 0.7 − 0.325 = +0.375. Positive. In IBPS, 0.35 − 0.65 × 0.25 = 0.35 − 0.1625 = +0.19. Positive.
Now if your eliminations are only 50 percent reliable, the two-option chance drops to 25 percent. RRB: exactly zero. IBPS: 0.25 − 0.1875 = +0.06. SSC CGL: 0.5 − 0.375 = +0.125. So even with mediocre elimination the guess is roughly neutral in these exams and never drastically negative. That is worth knowing: the penalties are mild enough that being somewhat wrong about your elimination does not sink you.
Where it does sink you is when you guess on questions where you have no basis at all and do it forty times. Neutral in expectation means a lot of variance and no gain. You got nothing for the effort and you spent time that could have gone elsewhere.
The rule I would actually carry into the hall
Here is a compact rule for each type of exam, built from the numbers above.
- If you can solve it, solve it. Nothing to decide.
- If you can eliminate at least two options for a reason you can state, answer. The math is positive in every exam listed.
- If you have eliminated one option or none, leave it in SSC CGL Tier 2, RRB and bank exams, unless you are in the last minute and have nothing better to do with the time. In SSC CGL Tier 1 a total blind guess is slightly positive, so it is fine.
- Do not spend time on a guess. A guess should take five seconds, not a minute of agonising. Otherwise, you lost time that could have solved another question.
There is a fifth point that overrides the rest. If a question is going to eat two minutes and you have a strong chance of solving it, that is not a guess and not a skip. It is a bad deal in time. Skip it because of time, not because of marks.
The time cost matters more than the penalty
New students focus on the penalty and ignore the clock. In a 60-minute SSC CGL Tier 1 paper with 100 questions, average time per question is 36 seconds. In NTPC CBT 1, it's 100 questions in 90 minutes, 54 seconds each. In IBPS prelims sections it is about 34 to 40 seconds.
A hard question that takes 2 minutes and gets you a 60 percent chance of a correct answer is a poor use of time compared to three easy questions that take 40 seconds each. Suppose SSC CGL Tier 1 with 2 marks per question. Three easy questions at 90 percent accuracy give 3 × (0.9 × 2 − 0.1 × 0.5) = 3 × 1.75 = 5.25 marks. The hard question gives 0.6 × 2 − 0.4 × 0.5 = 1.0 marks. There is no contest.
So the first thing to skip is not the question that scares you. It is the question that takes long. Then the next, the question you do not know. Only the last group is the one where the negative marking really matters.
A sample walk-through of an SSC CGL Tier 1 sitting
Take an actual sequence of choices.
You open quantitative aptitude. Question 3 is a percentage problem. You solve it in 30 seconds. Question 4 is a mensuration problem with a compound figure. You would need to compute two areas and subtract. Two minutes at least. Skip.
Question 9 is a data question, and two of the four options are clearly too large. You compute roughly and are between two values. It takes 20 seconds, you have a 50-50 chance. With +2 and −0.5, this is worth +0.75 in expectation. Answer.
Question 14 is a geometry question about an angle in a circle you do not know the theorem for. No elimination possible. Leave it. In Tier 1, a blind guess has expectation +0.125, which is tiny. It is not worth the four seconds when you might use them on the next easy item. But it is not harmful either. If you have 10 seconds at the very end, you can mark it.
At the end you have 18 questions marked for review out of 25 in the section. You now go back through them in order of easiest first. That is the difference between a good and average attempt: the second pass on questions that seemed hard but are solvable in 60 seconds with a fresh look.
Section by section, where does guessing make sense
This is my own opinion, drawn from the structure of these papers, and you should test it against your own mock data.
Quant. Best place to eliminate, because you can often bound the answer. Approximation and DI questions are good candidates. Never guess on multi-step word problems where the options are close together and you have not started the maths.
Reasoning. Mixed. Syllogisms with three or four statements can be eliminated by rejecting the wrong combinations. Puzzles with an arrangement half solved can also be guessed on with a reasonable basis. Coding-decoding is usually solvable or not. Non-verbal series and analogies are the ones people guess blind, and there the elimination is generally weak. See non-verbal reasoning for SSC and RRB.
English. Very good for elimination in error spotting and fill in the blanks, because grammar rules knock out options. Vocabulary questions are all-or-nothing. If you know the word you know it, and eliminations made because a word looks unfamiliar are unreliable. For para jumbles, if you can fix the first or last sentence you cut the options quickly.
General awareness. The most dangerous section for false confidence. A student who has read 60 percent of the syllabus feels like they know more than they do. Eliminate only options that you can positively rule out (an item is not in the correct state, it is from the wrong century, it is a person from another field). And in static GK, remember that examiners write options that look plausible. An unfamiliar name is not necessarily wrong.
For details on preparing this section see static GK for SSC and RRB and current affairs: how to study.
Testing your own guessing accuracy
A good exercise, and one that Pareeksha's mock reports are useful for: after each mock, sort your attempted questions into three groups.
- Confident (you knew the method or fact)
- Half-confident (you eliminated some options)
- Blind (you clicked without a basis)
Then, using the answer key, count how many of each group you got right. After five mocks you have real numbers. If your "half-confident" group runs 55 percent, guessing on those pays comfortably in every exam above. If it runs 30 percent, you have been over-trusting your own instincts and you should tighten the rule.
Do the same for the blind group. If it runs 25 percent or so, you already know blind guessing is a wash. Nothing wrong with it as a filler in the last ten seconds, but no reason to build a strategy around it.
This little audit does more than any article. It replaces "I think I am good at guessing" with a number. After a mock on Pareeksha, the review screen shows what you got wrong and what you left, so tagging takes a few minutes. Start with the free mini-mocks at /mock-tests and pull the answers into a notebook or spreadsheet.
Mistakes that cost more than the penalty
Changing correct answers at the last minute. Many students, when they revisit a question at the end, second-guess an answer that was right. Unless you have found a specific error in your working, leave it. Your first answer, made with a fresh reading, is right more often than a nervous revision.
Marking a question because you "have a feeling" for one option. A feeling is not an elimination. If you cannot say why the other options fail, it is a blind guess with a nice story attached.
Skipping too many. The opposite trap. A student who has been warned about negative marking begins leaving questions he could solve with a little more time. In a 100-question paper with 2 marks each, ten skipped questions you could have done costs you 20 marks. That is far more than typical penalties.
Attempting to hit a number. "I must attempt 90 questions." Your target is the marks, not the attempt count. If accuracy is 85 percent at 75 attempts, that beats 65 percent accuracy at 90 attempts. Compute it for SSC CGL Tier 1: 75 attempts at 85 percent is 63.75 correct and 11.25 wrong: 127.5 − 5.6 = 121.9 marks. Now 90 attempts at 65 percent is 58.5 correct and 31.5 wrong: 117 − 15.75 = 101.25 marks. The lower-attempt sitting wins by roughly twenty marks.
Ignoring the Tier 2 penalty. SSC CGL Tier 2 Paper I is reported to carry a heavier penalty than Tier 1. The break-even chance depends on the marks per question and the deduction, so read the current notification and recompute p = P / (M + P). If the deduction is one-third of the marks, the break-even is 25 percent, same as RRB. A blind guess on a four-option paper is then neutral, and you should lean on elimination even more.
Special cases
Normalisation and negative marking
In shift-based exams, your raw marks go through normalisation to compare across sessions. This does not change the guessing math, because your raw score is calculated first with penalties. For how this works see how normalization works and SSC CGL exam pattern and normalization.
Tie-breaks
In some exams, negative marks or the number of wrong answers are used as a tie-break at some stage. The rule differs by exam and year. If you are borderline you would rather have fewer wrong answers than more, all else equal. The notification tells you the tie-break order, so check it. Do not lean on this. It is a footnote, not a strategy.
Sectional bars
If the exam has sectional cutoffs (IBPS prelims and mains do), the penalty applies inside each section. That gives a second reason to be careful in your weakest section, and a reason to use guessing more actively when you are close to a bar. We cover this in clearing sectional cutoffs in bank exams.
Bilingual confusion
If you are reading a translated question and the two versions differ, do not guess between meanings. Choose the language you understand and stick to it. More on this in Hindi-medium preparation for SSC and RRB.
A practice routine to build the skill
Skipping and guessing well are learned, not decided. Here is a way to practise in about 4 weeks without taking extra time out of your day.
Week 1. In each mock, mark each attempted question with C, H or B (confident, half, blind) on rough paper as you go. Do not change your habits. Just record.
Week 2. After each mock, tally accuracy by group. Compute your net marks with and without the blind group. See what the blind ones added or took away.
Week 3. Apply a fixed rule: blind ones are never attempted until the last minute. Half-confident ones are attempted only if two options are eliminated. See what happens to net marks.
Week 4. Apply the rule in the timed mocks and compare to week 1. Most people see a small gain in marks and a bigger gain in calmness, because the decision has become routine.
Do this on full mocks, not sectional ones, because the pressure of a full paper is where the bad habits appear.
When you are far below the cutoff and need a gamble
Sometimes you are at 45 net at the end and the cutoff you hope for is 60. If there are 15 unattempted questions and 5 minutes, the situation is different from the calm case. You need variance, not expected value. Blind guessing with a neutral expectation has a chance, small, of adding enough. Not guessing has none.
That is a real reason to guess in the final two or three minutes when you are well below where you need to be. If you are comfortably above, the opposite applies: keep your lead and do not pile on wrong answers.
I would still not build this into a plan. A student who regularly needs the last-minute gamble is a student whose earlier 85 minutes were wasted, and the fix is upstream.
Before the exam: a checklist
- Confirm the negative marking scheme for your exam and stage on the official notification.
- Write the break-even chance for it on your rough sheet on the first line: 20 percent, 25 percent, whichever.
- Decide in advance your policy: two options eliminated means answer. No basis means leave.
- Practise the review pass in mocks: unattempted questions first, then confident answers you flagged, and never re-open a question you finished unless you have spotted a specific error.
What to do next
Take your last three mocks and count how many wrong answers came from blind guesses versus honest mistakes. Then check how many marks you lost to each. It will take twenty minutes and it will tell you which of the two problems to fix first.
If you are choosing among exams, the SSC CGL exam pattern and normalization and RRB NTPC CBT 1, CBT 2 and typing test posts have the full patterns. And section-wise time management goes deeper into how to split the clock.
FAQ
Is there negative marking for unattempted questions? No. In all the exams above, a blank costs nothing.
Is it better to attempt everything in SSC CGL Tier 1? Not automatically. A blind guess is slightly positive in expectation (about +0.125 marks) because of the 2 marks against a 0.5 penalty, but the time you spend can be better used. If you have seconds left at the end, marking the remaining questions is reasonable.
Should I guess in RRB NTPC? A blind guess is neutral (the break-even is 25 percent and a four-option guess is exactly 25 percent). Guess only when you can drop two options.
What if I am unsure about the penalty for my exam? Read the "scheme of examination" section of the official notification. Penalties are stated there. Do not rely on a coaching site for this one.
Does negative marking apply to the typing test or descriptive paper? No. It applies to objective sections. Typing tests and descriptive papers have their own qualifying rules.
Do more attempts always mean higher percentile? No. Accuracy matters as much as attempts. A 75-attempt sitting at high accuracy often beats a 90-attempt sitting at low accuracy, as the SSC example above shows.
Penalty schemes were checked against recent published exam patterns in September 2026. The SSC CGL Tier 2 marks per question and penalty are reported figures; confirm them in the current notification.
Comments (0)
No comments yet. Be the first to share your thoughts.