RBI Grade‑B – Foundation & Exam Strategy
Topic: Marking Scheme
Audience – Candidates preparing for the RBI Grade‑B Foundation test.
Goal – Understand the exact scoring algorithm, derive optimal answer‑taking strategies and prepare for edge‑case scenarios.
Level – MIT‑style, first‑principles derivation, rigorous probability reasoning, and a challenging applied problem.
1. Overview of the RBI Grade‑B Foundation Exam
| Feature | Detail |
|---|---|
| Sections | 1. General Aptitude – 40 MCQs (10 each of Quantitative, Reasoning, English, Data Interpretation) 2. Document‑Based Question (DBQ) – 1 multi‑part question (4 sub‑parts) |
| Time | 1 hour 45 minutes (105 min) |
| Answer Sheet | Optical mark‑sheet (OMR) |
| Scoring | 4 marks for each correct answer, –1 mark for each wrong answer, 0 for un‑answered. |
| Total Marks | 160 (120 MCQs + 40 DBQ) |
| Negative Marking | Only on MCQs; DBQ has no negative marks (partial credit is awarded). |
Note – The DBQ is not a multiple‑choice question; it is a short‑answer/essay type scored by examiners. The marking scheme described below applies only to the 120 MCQs.
2. Marking Scheme – First Principles
2.1 Score Function
Let
- (N = 120) – number of MCQs.
- (c_i \in {0,1}) – indicator that question (i) is answered.
- (s_i \in {0,1}) – indicator that answer to question (i) is correct.
- (M = 4) – marks for a correct answer.
- (L = 1) – marks deducted for a wrong answer.
The raw score (R) is
[ R = \sum_{i=1}^N \bigl( M, c_i, s_i - L, c_i, (1-s_i) \bigr) = \sum_{i=1}^N c_i (M s_i - L(1-s_i)) ]
Simplifying:
[ R = \sum_{i=1}^N c_i \bigl( (M+L)s_i - L \bigr) ]
Thus each answered question contributes either (+M) or (-L); unanswered questions contribute (0).
2.2 Expected Score under Random Guessing
Assume a candidate guesses the answer to a question with probability (p) of being correct (e.g., (p = 0.25) if the MCQ is four‑option). The expected contribution of a guessed question is
[ E[\text{contrib}] = p \cdot M + (1-p)(-L) = M p - L (1-p) ]
For (M=4, L=1, p=0.25):
[ E = 4(0.25) - 1(0.75) = 1 - 0.75 = 0.25 ]
Hence guessing a 4‑option MCQ yields a positive expected value of 0.25 marks per question. This is the key reason why answering all questions is statistically optimal.
3. Advanced Analysis – Optimal Strategy
3.1 Threshold Probability
Generally, for a question with (k) options, the probability of a random guess being correct is (p = 1/k). The expected value of guessing is
[ E = M p - L (1-p) = M\frac{1}{k} - L\left(1-\frac{1}{k}\right) ]
Set (E \ge 0) to find the critical (k):
[ M\frac{1}{k} \ge L\left(1-\frac{1}{k}\right) ;;\Longrightarrow;; k \le \frac{M+L}{L} ]