Exam Map, Scoring and Attempt Strategy
This chapter covers exam map, scoring and attempt strategy. Open with the UPSC NDA/NA written-test map exactly as the current notification states it: two papers, Paper I Mathematics and Paper II General Ability Test, with the GAT split into an English part and a General Knowledge part spanning physics, chemistry and general science, history and the freedom movement, geography, and current events.
KarmSakha preparation topics covered
- Understand and apply: UPSC NDA/NA Written Examination Pattern
- Understand and apply: Scoring Arithmetic: Expected Value and Break-Even
- Understand and apply: Working Accuracy vs Break-Even: The Safety Margin
- Understand and apply: Three-Pass Attempt Order, Leave-It Rule and Time Budgets
- Understand and apply: Confidence Zones and OMR Discipline
UPSC NDA/NA Written Examination Pattern
The UPSC NDA/NA written examination is a two-paper objective-type test. Paper I is Mathematics, and Paper II is the General Ability Test (GAT). The GAT is not a single homogeneous subject paper; it is split into an English part and a General Knowledge part. That two-paper, three-part skeleton is the one stable feature of the NDA pattern. What is not stable is the arithmetic: the number of questions, maximum marks, duration, and negative-marking fraction are all controlled by the notification for that particular examination cycle. Some cycles revise these numbers; coaching materials from previous years may be outdated.
Paper I Mathematics is drawn from the class XI–XII syllabi: algebra, matrices, trigonometry, analytic geometry, differential and integral calculus, vectors, statistics, and probability. Every question is objective, but the paper’s challenge is speed rather than concept recall alone. Paper II GAT consists of two distinct sections. The English section tests grammar, vocabulary, sentence structure, and comprehension. The General Knowledge section is further divided into physics, chemistry and general science; history and the freedom movement; geography; and current events. You should therefore treat GAT preparation as preparation for four or five separate sub-papers.
Do not begin a study plan by memorising question counts printed in any guide, including this one. UPSC revises specifications when it chooses, and a pattern table that is correct today may be wrong in the next notification. The honest and safe procedure is to download the official notice from the Commission’s website, locate the scheme-of-examination table, and copy the numbers into the blank table at the end of this section. Fill it before you allocate weekly hours; the numbers determine how many questions per minute you must target in each part.
The blank pattern table has one row for Paper I Mathematics, one for Paper II GAT English, one for Paper II GAT General Knowledge, and one for the GAT total. The English and General Knowledge rows share the same duration as the total GAT paper, so fill that duration in the total row. In some notifications the English and GK sections are listed with separate maximum marks; in others, only the GAT total appears. If the notification does not split them, leave the section-level cells blank rather than inventing numbers.
The most important cell is the penalty fraction. It is written in the notification as one-third of the marks assigned to that question or as a similar fraction. Record it in your own terms: if the penalty is one-third, write k = 1/3, not just −1/3. That fraction will feed directly into the scoring arithmetic in the next section. You should also note whether marks are awarded for correct answers and whether there is any provision for unanswered questions. Usually an unanswered question earns zero marks, but confirm this from the notice.
When you fill the table, use a pencil because you may need to update it when the official notice is issued. Do not trust anonymous online charts or latest-pattern images; verify against the PDF on the official UPSC website. The syllabus distribution — what topics appear — also comes from the notification appendix. In the GAT science and geography sections, the syllabus list tells you which physical phenomena, map skills, and current events are examinable. Keep the filled table pinned to your study desk; every practice-test target should be calculated against it.
One subtlety: the question count and duration determine not just total marks but your per-question average time. For example, if a paper has 120 questions in 150 minutes, the average is 75 seconds per question; if it has 120 questions in 180 minutes, it is 90 seconds. This average, not the raw count, is what you will later compare against the 45-second leave-it rule. But because the numbers are notification-controlled, any average you calculate now may be irrelevant. Calculate it only after you fill the blank table from the current official notice.
| Paper / Section | Number of Questions | Maximum Marks | Duration | Penalty Fraction (k) |
|---|---|---|---|---|
| Paper I — Mathematics | ||||
| Paper II — GAT — English | ||||
| Paper II — GAT — General Knowledge | ||||
| Paper II — GAT — Total |
Scoring Arithmetic: Expected Value and Break-Even
Every answer attempt is a bet with an expected value. Let n be the number of options per question, and let k be the penalty fraction subtracted for a wrong answer; a correct answer earns +1 mark and an incorrect answer earns −k marks. If you guess blindly, your probability of being correct is 1/n and of being wrong is (n−1)/n. Therefore the expected value of a blind guess is E = (1/n)(+1) + ((n−1)/n)(−k), which simplifies to 1/n − k(n−1)/n. This single formula is the foundation of NDA attempt strategy.
Apply it to the common NDA case of four options with a one-third penalty: n=4 and k=1/3. Then E = 1/4 − (1/3)(3/4) = 0.25 − 0.25 = 0. Blind guessing on four-option questions neither adds nor loses marks in the long run. That is why a one-third penalty is often described as a barrier to random guessing: on a single question you might gain or lose, but averaged over many questions the gain and loss are exactly balanced. If the penalty were reduced to 1/4, E would be 0.25 − 0.1875 = +0.0625, making blind guessing profitable; if increased to 1/2, E = 0.25 − 0.375 = −0.125, making it a systematic loss.
Set E = 0 and solve for k to find the zero-gain penalty. From 1/n = k(n−1)/n we get k = 1/(n−1). For four options, that value is 1/3, exactly the usual NDA penalty. For five options, the zero-gain penalty is 1/4; for three options, it is 1/2. This explains why a penalty that looks harsh in a four-option paper may actually be generous in a five-option paper. Always compare the notification’s k with 1/(n−1), not with a fixed rule such as one-third is always strict.
Blind guessing, however, is the worst form of attempted answer. A serious candidate has some degree of knowledge, so the relevant variable is your working accuracy p on the questions you attempt. If you attempt a question, the expected gain is p(1) + (1−p)(−k), which simplifies to p(1+k) − k. Leaving the question blank gives 0. Therefore attempting is worthwhile exactly when p(1+k) − k > 0, that is, when p > k/(1+k). The break-even accuracy is p* = k/(1+k).
For a one-third negative penalty, p* = (1/3)/(1+1/3) = (1/3)/(4/3) = 1/4 = 25%. This means: if you expect to be right on more than one question in four among those you attempt, attempting raises your expected score; if you expect to be right on fewer than one in four, those attempts cost you marks. Notice that p* does not depend on the number of options n; it depends only on the penalty k. The option count appears only in the blind-guess formula, because it determines your chance of a lucky hit.
Work through a concrete five-question mini-set with k=1/3 and p=25%. In expectation, you get 1.25 correct and 3.75 wrong, so the score is 1.25 − (3.75)/3 = 1.25 − 1.25 = 0. With p=50% on the same paper, 2 correct and 2 wrong give 2 − 2/3 = 1.33. Thus at exactly break-even you neither gain nor lose; above it, you gain linearly. This linearity is why small improvements in accuracy have large effects on total score, and why shifting from 25% to just 35% changes a 100-question paper from roughly zero to about 13.3 marks.
A common misinterpretation is that break-even of 25% means you should attempt every question for which your confidence is at least 25%. That is mathematically correct but psychologically useless. Your confidence is not the same as your true accuracy; candidates systematically overrate confidence on tricky questions. More importantly, a whole paper built near p=25% leaves no room for careless errors or wording surprises. Use break-even as a lower bound for marginal questions, not as a permission slip for broad guessing.
One further arithmetic nuance: the expected-value formula assumes a correct answer earns one full mark and a wrong answer loses k times that mark. If a two-mark question has a one-third penalty, a wrong answer loses 2/3 mark. After dividing all scores by the question’s weight, the problem is identical. So all derived inequalities remain valid regardless of whether questions carry one, two, or four marks. Keep this in mind when comparing practice-test scores from different sources.
Working Accuracy vs Break-Even: The Safety Margin
The break-even accuracy p* = k/(1+k) answers one question only: at what accuracy do attempted questions stop hurting your total score? It does not tell you what accuracy you should aim for. For the NDA one-third penalty, p* is 25%; but a candidate who genuinely knows only 25% of what they attempt and guesses the rest will score near zero after penalty. That is the score of someone who did not prepare. A realistic and safe target for classroom tests, mock papers, and the final examination is 70–80% working accuracy.
Define working accuracy carefully: it is the number of correct answers divided by the number of questions you actually attempted, not the number of questions in the paper. If you attempt 90 questions and get 72 right, your working accuracy is 80%. Leaving blank questions you do not know does not lower this percentage; attempting them and getting them wrong does. Working accuracy is therefore the quality of your attempts, and it is the number that connects directly to the scoring formula.
Why 70–80% rather than 90%? Because on a real NDA paper the questions are competitive and time pressure is high. Even a topper makes careless mistakes, misreads a word, or selects the right first calculation and the wrong final option. A safety margin means that a 10-percentage-point drop on a tough section still keeps you far above break-even. If you train at 75% and a difficult set drops you to 65%, your score still grows strongly. If you trained exactly at 25%, a dip to 15% turns your aggressive attempts into a net loss.
See the difference arithmetically. Suppose 100 attempted questions with k=1/3. At 25% accuracy: 25 correct and 75 wrong give 25 − 75/3 = 0. At 50%: 50 − 50/3 ≈ 33.3. At 70%: 70 − 30/3 = 60. At 80%: 80 − 20/3 ≈ 73.3. The gap between 25% and 70% is 60 marks — larger than the gap between 50% and 80%. This is why toppers do not sit near the break-even line; they create a large positive wedge between correct answers and penalised wrong answers.
The commonest attempt-strategy error is conflating the two numbers. Candidates read “one-third penalty breaks even at 25% accuracy” and conclude “I only need to know 25% of the syllabus and can guess the rest.” That is a misinterpretation on two levels. First, 25% is the minimum for not losing marks; it is a survival threshold, not a target. Second, the 25% is accuracy on attempted questions, not knowledge coverage of the syllabus. You may know 60% of the concepts yet score below 25% on attempted questions if your attempt strategy is poor.
Another way to think of the safety margin is as insurance against the examiner’s freedom. The NDA paper is not a random sample of easy questions; it deliberately contains tough items and trap options. If your true ability is 75%, a hard paper may reduce working accuracy to 65%, but you still gain marks. If your ability is 30%, the same paper may drop you to 20%, below break-even, and your attempts start losing marks. The margin absorbs variance in difficulty. This is why all three-pass strategies and leave-it rules are built around preserving accuracy, not chasing question count.
Aim above break-even, never at it. This principle also tells you how many questions to attempt. If your working accuracy in mocks is 80%, you can afford a higher attempt count, including some medium-confidence questions, because the average remains high. If your accuracy is 60%, you should be more selective: attempt only high-confidence questions sharply and leave borderline questions blank. The number of attempted questions is not a target in itself; it is a by-product of maintaining the 70–80% band. Let accuracy drive attempt count, not the reverse.
