1.2 CBT Pattern, Negative Marking & Score Normalization

CBT Pattern, Negative Marking & Score Normalization

Quick Answer: The RRB Group D CBT is a 100-question, 100-mark, 90-minute objective test split across four sections — Mathematics (25), General Intelligence & Reasoning (30), General Science (25), and General Awareness & Current Affairs (20). Each wrong answer costs 1/3 of a mark; unattempted questions cost nothing. Because the exam is held in multiple sessions, RRBs normalize scores using a statistical formula so that no candidate is disadvantaged by a harder paper.

The CBT Pattern at a Glance

The CBT is a single-stage Computer-Based Test. Questions are set at the 10th-standard (CBSE) level and the paper is available in English, Hindi, and 13 regional Indian languages. The section distribution is indicative, meaning RRBs may shuffle question counts marginally across cycles, but the totals stay at 100.

• Time: 90 minutes overall; 120 minutes for eligible PwBD candidates who opt to use a scribe.
• Marking: each correct answer carries +1 mark.
• Negative marking: −1/3 mark for each wrong answer.
• Unattempted: no penalty, no credit.

With 100 questions in 90 minutes, the average time per question is roughly 54 seconds. Reasoning (30 questions) consumes the most time, so plan to scan Mathematics and General Awareness quickly and reserve reasoning time for the harder puzzles.

Negative Marking — How It Works

The penalty is one-third (1/3) of a mark per wrong answer. Because each correct answer is worth exactly 1 mark, the penalty is one-third of the reward. Unattempted questions carry neither reward nor penalty — they simply do not contribute to your score.

Worked Example

Suppose a candidate attempts 80 questions: 55 are correct and 25 are wrong.

• Correct: 55 × (+1) = +55
• Wrong: 25 × (−1/3) = −8.33
• Raw score = 55 − 8.33 = 46.67 (out of 100)

Now consider an alternative: the candidate leaves all 25 risky guesses blank instead.

• Correct: 55 × (+1) = +55
• Wrong: 0 × (−1/3) = 0
• Raw score = 55.00

The 8.33 marks lost to wild guessing are more than what most candidates gain from blind play. Negative marking therefore rewards disciplined question selection — guess only when you can eliminate at least two options.

A Second Worked Example

Consider two candidates sitting the same session. Candidate A attempts 60 questions with 50 correct and 10 wrong; Candidate B attempts 90 questions with 60 correct and 30 wrong.

• Candidate A: 50 − 10/3 = 50 − 3.33 = 46.67
• Candidate B: 60 − 30/3 = 60 − 10.00 = 50.00

Candidate B's wider attempt earns a higher score despite more wrong answers, because the 10 extra correct answers (each +1) outweigh the 20 extra wrong answers (each −1/3, total −6.67). The lesson: extra attempts are worthwhile only when your accuracy on those extra attempts is better than 1 correct in 3 — the break-even ratio. At exactly 1-in-3 you break even; below it you lose marks; above it you gain.

Strategy Under Negative Marking

The break-even ratio of one correct in three wrong tells you the threshold, not the rule. In practice, exam-day strategy depends on your confidence on the specific question:

• High confidence (you know the answer): attempt. The expected value is strongly positive.
• Medium confidence (you can eliminate two of four options): attempt. With two options left, the expected value is (0.5 × +1) + (0.5 × −1/3) = +0.33 per question, so you gain over time.
• Low confidence (you can eliminate only one option or none): skip. The expected value is zero or negative, and you preserve your score.
• No time to read the question: skip and bubble nothing. A blank answer is always safer than a rushed guess.

The discipline of not attempting is as much a skill as solving. A realistic target for a well-prepared candidate is a high-accuracy attempt of roughly 75–85 questions, with the genuinely unknown items left blank rather than gambled on.

Minimum Qualifying Marks

Qualifying marks are the floor you must clear to be considered for shortlisting; they are not the cut-off for selection. The cut-off for PET shortlisting is almost always higher because it is set at 3× the vacancies on merit.

Score Normalization Across Sessions

The CBT is administered in multiple sessions with different question papers of comparable difficulty. Despite best efforts, papers are never identically difficult. To ensure fairness, RRBs apply a statistical normalization formula that adjusts raw scores to a common base.

Para 15.0 of CEN 09/2025 states only the rule: "Short listing of Candidates for various stages shall be based on the normalized marks obtained by them whenever CBT is conducted in multiple sessions for the same syllabus." The mechanics are set out separately, in the RRBs' Notice on Normalization Methodology for Computer Based Test (CBT) exams (for all CENs) — first issued in November 2024 and reissued in February 2026. That notice describes a percentile-based method, not a simple mean-and-standard-deviation rescaling:

1. Raw marks are computed for every candidate: (correct answers × 1) − (wrong answers × 1/3).
2. A percentile score is computed within each shift: Percentile = (number of candidates in that shift with a raw score less than or equal to yours ÷ total candidates in that shift) × 100. The topper of every shift therefore gets 100, and percentiles are carried to five decimal places to reduce ties.
3. A base shift is chosen — the shift with the highest mean marks, provided its attendance is at least 70% of the average attendance across shifts.
4. Your percentile is mapped onto the base shift's marks, by interpolation between the two nearest percentile points in that shift. The result is your normalized mark, rounded to two decimals.
5. Ties are broken exactly as CEN 09/2025 prescribes for merit: the older candidate ranks higher, and if ages are equal, alphabetical order (A to Z) of the given name applies.

What This Means for You

• Your normalized score, not your raw score, decides merit.
• A candidate who faced a harder session is not penalised: their raw score is lifted if their session's mean was low.
• Conversely, an easy session's raw scores are pulled down toward the base.
• You cannot compute your own normalized score at home, because it depends on every candidate's raw score in your shift and in the base shift; you only see the published normalized score.

The reason normalization matters is that RRB Group D is conducted across many sessions over several days, each with its own question paper drawn from the same syllabus. Even with careful setting, one session may lean heavier on a topic you find easy, while another may weight Reasoning puzzles you find hard. Without normalization, two equally capable candidates could receive very different raw scores purely because of the session they were assigned to — and assignment is random. Normalization removes that luck factor by mapping every session's raw scores onto the same statistical scale, so merit is decided by your performance relative to your own session's cohort, not by the absolute difficulty of your paper. This is also why RRBs publish only the normalized score, never the raw score, in the result.

The practical implication: aim for the highest raw score you can. Normalization corrects for session difficulty, but it does not rescue a low raw score. The candidates who clear the cut-off are those who maximise correct answers while controlling wrong answers through sensible guessing.