Understanding UPSC Prelims Negative Marking (and How to Beat It)

Negative marking is the single most misunderstood part of UPSC Prelims. Aspirants either over-skip (and miss easy marks) or over-guess (and leak marks on questions they should have left alone). Both failures come from not having internalised the math. Once you actually work it out — and practice under the real marking scheme — the right strategy becomes obvious.

This post walks through the arithmetic, the break-even points, and the practice routines that will let you make the right call in real time on exam day.

The marking scheme, in plain terms

UPSC Civil Services Preliminary Examination — General Studies Paper 1 — has 100 multiple choice questions, four options each, 120 minutes total. The marking is:

• +2 marks for each correct answer
• −2/3 of 2 marks = −0.66 marks for each incorrect answer
• 0 marks for each unattempted question

That "minus two-thirds of two" is what makes the whole thing non-trivial. Two-thirds of 2 is about 0.6667, so every wrong answer costs you 0.66 marks. Multiply that by, say, 30 wrong answers and you've lost 20 marks. In a cutoff-driven exam, 20 marks is usually the difference between a call and a regret.

The break-even math

Here's the core question: if you guess randomly on a four-option question, what's your expected score?

You have a 1 in 4 chance of being right (+2) and a 3 in 4 chance of being wrong (−0.66).

Expected value = (1/4 × +2) + (3/4 × −0.66) = 0.5 − 0.5 = 0

Random guessing on a four-option question has an expected value of zero. So pure random guessing, over many questions, is statistically equivalent to leaving them blank. The risk is higher variance — you might land above zero, you might land below — but on average you neither gain nor lose.

(This is by design. UPSC calibrated the penalty at −1/3 of the correct-answer value specifically so that random guessing cancels out.)

Intelligent guessing: where the real edge lives

The math changes the moment you can eliminate even one option.

If you can eliminate 1 option (3 options remaining):

Expected value = (1/3 × +2) + (2/3 × −0.66) = 0.67 − 0.44 = +0.23

Now guessing has a positive expected value. Not huge, but real. Over 10 such questions, you'd gain ~2.3 marks on average.

If you can eliminate 2 options (2 options remaining):

Expected value = (1/2 × +2) + (1/2 × −0.66) = 1.0 − 0.33 = +0.67

A 50/50 guess is worth about two-thirds of a full correct mark, on average. You absolutely should attempt these.

If you can eliminate 3 options (1 option remaining): you're no longer guessing. Attempt.

The practical rule

The rule falls out cleanly from the math:

• Eliminate 0 options (you have no idea): leave it blank. Expected value is zero; variance is pointless risk.
• Eliminate 1 option: lean toward attempting, especially if the question is on a topic you've studied. Your elimination is probably better than random.
• Eliminate 2 options: always attempt.
• Eliminate 3 options: attempt, obviously.

The tricky case is the first one — "no idea." In practice, you rarely have truly zero information on a UPSC question. If you've prepared, you'll usually have at least a weak signal that rules one option out. The discipline is being honest with yourself about whether your "feeling" is actually an elimination or just a coin flip.

The hidden cost most aspirants ignore: time

All the math above assumes each question is independent. In reality, attempting a question costs you time, and time is the other currency of Prelims. 120 minutes / 100 questions = 72 seconds per question. If you spend 3 minutes agonising over a "50/50" that's worth +0.67 expected marks, you've used ~2 minutes of time that could have fetched you a full +2 elsewhere.

This is why per-question timing data is so valuable during practice. A platform that records seconds spent on each question (and how many times you revisited it) gives you the raw material to figure out where your time actually goes. craqdIAS tracks per-question timing and displays it in the post-exam review so you can see the pattern.

How to practice with negative marking

Here's the problem: if you practice without the real penalty, you'll develop a bad guessing habit. You'll attempt questions you shouldn't and end exam day losing 15–20 marks you never needed to lose.

Concrete suggestions:

1. Always practice under real UPSC marking (+2 / −0.66 / 0). If your platform uses simplified scoring, its "accuracy" number is lying to you.
2. After each paper, count three categories separately: attempted & correct, attempted & wrong, left blank. Look at the ratio of wrong-to-right. If you got 60 right and 25 wrong, your net score is 60 × 2 − 25 × 0.66 = 120 − 16.5 = 103.5. Every wrong answer cost you a third of the value of a right one.
3. For every wrong answer, mark whether it was a real guess or a confident attempt. If your "confident attempts" are consistently wrong, the issue is concept strength, not guessing strategy. If your "guesses" are mostly wrong, your elimination skills are weaker than you thought — tighten the rule, skip more.
4. Track the trend over attempts. If your wrong-to-right ratio is climbing, you're guessing too much. If your blank count is very high, you're being too conservative and leaving marks on the table.

craqdIAS uses the authentic UPSC marking scheme in every mock test and exposes all three counts (correct / incorrect / skipped) plus net accuracy in every report card — so this calibration happens as a byproduct of practicing.

What not to do

• Don't blanket-attempt. "I'll attempt all 100" is not a strategy. It guarantees you leak marks on the questions you truly don't know.
• Don't blanket-skip. If you're skipping questions where you've eliminated two options, you're leaving positive expected value on the table. In a competitive cutoff exam, that's expensive.
• Don't change your rule on exam day. Whatever strategy you practiced with is the one you'll execute under stress. Exam day is not the time to experiment.

The summary

Negative marking is not a hazard; it's a tool UPSC uses to reward discipline. Random guessing is neutral. Informed guessing is profitable. The aspirants who do best are the ones who have practiced enough under real marking to know — in under a second — which category each question falls into.

The single most practical step you can take today is to practice a full paper under authentic +2/−0.66 marking. Pick any year from 2013 to 2025, set the 120-minute timer, and attempt it. Then review. Then do it again next week. That's the whole thing.